diff --git a/README.md b/README.md
index 27f4f35..c3186a5 100644
--- a/README.md
+++ b/README.md
@@ -5,7 +5,7 @@
 
 ![Python 3.11+](https://img.shields.io/badge/Python-3.11%2B-2563eb)
 ![Research status](https://img.shields.io/badge/status-paper--live_research-d97706)
-![Tests](https://img.shields.io/badge/tests-218_passing-0f766e)
+![Tests](https://img.shields.io/badge/tests-221_passing-0f766e)
 ![License](https://img.shields.io/badge/license-MIT-64748b)
 
 The project asks a narrow investment question:
@@ -227,7 +227,14 @@ visible from the shared decision-close baseline. It reports net return,
 SPY-relative return, replacement-only turnover, and rebalance counts under
 identical execution assumptions.
 
-[Open the deployed live-forward dashboard →](https://hansun103.github.io/AI-Driven-Portfolio-Optimization-Backtesting-System/)
+The deployed monitor now has two linked pages with one shared visual shell:
+
+- [Live Forward Monitor — default view →](https://hansun103.github.io/AI-Driven-Portfolio-Optimization-Backtesting-System/)
+- [Horizon Comparison — 5/10/20-session research →](https://hansun103.github.io/AI-Driven-Portfolio-Optimization-Backtesting-System/horizons.html)
+
+The original Stock-level view remains canonical rather than being duplicated
+on the horizon page. The two-page contract and evidence boundaries are recorded
+in the [dashboard merge design](docs/design/DASHBOARD_TWO_PAGE_MERGE.md).
 
 This monitoring track does not replace the primary horizon-matched 5/10/20
 experiment described above. The former one-session fitting workflow remains
@@ -378,20 +385,29 @@ because they weaken the preferred narrative.
 ## Known limitations
 
 1. No redesigned 20-session paper-live forecast has matured yet.
-2. The strongest portfolio results remain development evidence, not a genuine
+2. The strongest repaired historical test covers roughly 2018–2023, not a
+   full decade. It therefore spans too few independent market regimes for a
+   durable claim. Extending point-in-time testing toward at least ten years is
+   a research target, not a guarantee of adequacy: long-window backtests still
+   require regime analysis, untouched confirmation, and correction for the
+   number of strategies tried. See
+   [CFA Institute’s backtesting guidance](https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/backtesting-and-simulation)
+   and the
+   [deflated-Sharpe discussion of sample length and multiple testing](https://academic.oup.com/jrssig/article/18/6/22/7038278).
+3. The strongest portfolio results remain development evidence, not a genuine
    untouched confirmation.
-3. Public data does not fully solve delisted-security coverage or historical
+4. Public data does not fully solve delisted-security coverage or historical
    sector classification.
-4. The average ranking edge is small and varies materially by year and regime.
-5. Sentiment is promising, but coverage, timestamp quality, and causal
+5. The average ranking edge is small and varies materially by year and regime.
+6. Sentiment is promising, but coverage, timestamp quality, and causal
    attribution remain incomplete.
-6. The high-recall survival model has not passed its activation gates and
+7. The high-recall survival model has not passed its activation gates and
    remains a shadow watchlist.
-7. Same-close tactical execution is unvalidated without intraday quotes and
+8. Same-close tactical execution is unvalidated without intraday quotes and
    actual paper or broker fills.
-8. Scheduled full top-40 replacement remains under audit; the six-session
+9. Scheduled full top-40 replacement remains under audit; the six-session
    ablation is too short to activate a freeze or buffer.
-9. Static cost assumptions do not establish capacity, taxes, borrow
+10. Static cost assumptions do not establish capacity, taxes, borrow
    availability, market impact, or partial-fill behavior.
 
 ## Research disclaimer
diff --git a/docs/design/DASHBOARD_TWO_PAGE_MERGE.md b/docs/design/DASHBOARD_TWO_PAGE_MERGE.md
new file mode 100644
index 0000000..280180b
--- /dev/null
+++ b/docs/design/DASHBOARD_TWO_PAGE_MERGE.md
@@ -0,0 +1,88 @@
+# Two-page dashboard merge design
+
+Status: local proposal awaiting approval. Do not merge or deploy yet.
+
+## Decision
+
+Publish one dashboard product with two real static pages:
+
+| Page | Proposed URL | Responsibility |
+| --- | --- | --- |
+| Live Forward Monitor | `dashboard.html` | Daily operational evidence, benchmark comparison, stability, the canonical stock-level view, and experiment history |
+| Horizon Comparison | `horizons.html` | Horizon-matched 5/10/20-session progress, validation context, model specifications, SHAP explanations, and pipeline health |
+
+Both headers use the same two-option page switcher. The selected page is visually
+active. Ordinary links are preferable to a hidden single-page application because
+browser back/forward works, each page can fail independently, and GitHub Pages can
+serve the files without a runtime.
+
+## Shared stock-level view
+
+The Horizon Comparison page does not implement another portfolio table.
+
+The existing Stock-level view remains the canonical component. Its model selector
+will be extended to understand both:
+
+- legacy daily live-forward snapshots; and
+- v2 Ridge/XGBoost arms for the 5-, 10-, and 20-session horizons.
+
+An `Inspect stocks` action on a horizon model will link to a URL such as:
+
+```text
+dashboard.html?decision=2026-07-29&family=xgb_ranker&horizon=20#stock-level
+```
+
+The original page reads those parameters, selects the requested frozen portfolio,
+and scrolls to the shared Stock-level view. This avoids divergent filtering,
+sorting, return formatting, and holding-level evaluation logic.
+
+## Partial and mature evidence
+
+The horizon page reports three different evidence states without mixing them:
+
+1. **Awaiting entry** — the portfolio is frozen but its next-open entry has not
+   occurred.
+2. **Interim mark** — one or more open-to-open observations exist. The page shows
+   session progress and the compounded partial return immediately.
+3. **Matured label** — the complete 5-, 10-, or 20-session evaluation is available
+   and may enter the formal prospective comparison.
+
+The page never displays validation return as live return. A zero-session portfolio
+is shown as `0 / horizon` and `Awaiting entry`, not as a zero realized return.
+
+## Explainability
+
+Every newly frozen v2 forecast writes `shap_summary.csv`.
+
+- Ridge uses linear SHAP contributions around the current eligible cohort as the
+  background.
+- XGBoost uses exact TreeSHAP contributions from the fitted booster.
+- The dashboard publishes mean absolute SHAP, signed mean SHAP, positive
+  contribution share, method, and explanation-row count.
+
+Each model card contains a collapsed `Show SHAP feature importance` control.
+Opening it displays a horizontal importance chart. Old immutable forecasts that
+predate this artifact state that SHAP was not captured; coefficients or XGBoost
+gain are not relabelled as SHAP.
+
+## Daily refresh contract
+
+The daily pipeline should:
+
+1. refresh data and generate the current snapshots;
+2. evaluate all available interim open-to-open observations;
+3. evaluate newly matured forecasts;
+4. generate SHAP summaries for new horizon snapshots;
+5. rebuild `dashboard.html` and `horizons.html`;
+6. run public-payload and browser smoke checks; and
+7. publish both pages together through the existing GitHub Pages action.
+
+Publishing must be atomic: a cycle should never expose one page with a newer data
+contract than the other.
+
+## Approval boundary
+
+The current local Horizon Comparison preview demonstrates the second page. After
+approval, the merge work will add the page switcher and v2 arm support to the
+original dashboard, generate the final `horizons.html`, update the refresh and
+deployment workflows, and then request approval before pushing.
diff --git a/site/assets/dashboard.css b/site/assets/dashboard.css
index 325f1ca..7a2b8b5 100644
--- a/site/assets/dashboard.css
+++ b/site/assets/dashboard.css
@@ -1,11 +1,12 @@
 
 :root{--ink:#102033;--muted:#637083;--line:#dce3ec;--surface:#fff;--canvas:#f4f7fb;--blue:#3156d3;--teal:#0f9f8f;--amber:#d89614;--red:#c94c5b;--shadow:0 14px 35px rgba(24,39,75,.08)}
 *{box-sizing:border-box}body{margin:0;color:var(--ink);background:var(--canvas);font-family:Inter,ui-sans-serif,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif}
-header{position:relative;overflow:hidden;padding:38px max(28px,calc((100vw - 1440px)/2));color:#fff;background:linear-gradient(120deg,#081525,#12294b 70%,#173c57)}
-header:after{content:"";position:absolute;right:-90px;top:-160px;width:420px;height:420px;border:1px solid #ffffff21;border-radius:50%;box-shadow:0 0 0 55px #ffffff0a,0 0 0 115px #ffffff08}
-.eyebrow{color:#7dd3c7;font-size:11px;font-weight:800;letter-spacing:.16em;text-transform:uppercase}h1{margin:7px 0 8px;font-size:clamp(28px,3vw,42px);letter-spacing:-.035em}header p{max-width:760px;margin:0;color:#bdc9da;line-height:1.55}
-main{width:min(1440px,calc(100% - 40px));margin:auto;padding:26px 0 54px}.panel{background:var(--surface);border:1px solid #d8e0eb;border-radius:18px;box-shadow:var(--shadow)}
-.toolbar{display:flex;justify-content:space-between;align-items:end;gap:20px;padding:18px 20px;margin-top:-50px;position:relative;z-index:2}.toolbar-group,.segmented{display:flex;align-items:end;gap:10px;flex-wrap:wrap}
+header{position:relative;overflow:hidden;padding:42px max(28px,calc((100vw - 1440px)/2)) 82px;color:#fff;background:linear-gradient(120deg,#081525,#12294b 68%,#173c57)}
+header:after{content:"";position:absolute;right:-90px;top:-170px;width:440px;height:440px;border:1px solid #ffffff21;border-radius:50%;box-shadow:0 0 0 55px #ffffff0a,0 0 0 115px #ffffff08}
+.hero-row{display:flex;justify-content:space-between;align-items:end;gap:28px;position:relative;z-index:1}.eyebrow{color:#7dd3c7;font-size:11px;font-weight:850;letter-spacing:.16em;text-transform:uppercase}h1{margin:8px 0 9px;font-size:clamp(30px,3.2vw,46px);letter-spacing:-.04em}header p{max-width:780px;margin:0;color:#bdc9da;line-height:1.55}.hero-badge{display:grid;gap:4px;min-width:190px;padding:14px 16px;border:1px solid #ffffff24;border-radius:14px;background:#ffffff0d;backdrop-filter:blur(6px)}.hero-badge span{color:#9fb0c8;font-size:10px;text-transform:uppercase;letter-spacing:.08em}.hero-badge strong{font-size:15px}
+.page-switcher{display:flex;gap:9px;margin-top:22px;position:relative;z-index:2}.page-link{display:grid;gap:3px;min-width:210px;padding:12px 15px;border:1px solid #ffffff28;border-radius:12px;color:#c5d2e3;background:#ffffff0a;text-decoration:none;transition:.18s ease}.page-link:hover{transform:translateY(-1px);border-color:#7dd3c7;background:#ffffff12}.page-link.active{color:#fff;border-color:#7dd3c7;background:#7dd3c719;box-shadow:inset 0 -3px 0 #7dd3c7}.page-link strong{font-size:12px}.page-link small{color:#95a8c0;font-size:9px}.page-link.active small{color:#b9d8d3}
+main{width:min(1440px,calc(100% - 40px));margin:-52px auto 0;padding-bottom:58px;position:relative}.panel{background:var(--surface);border:1px solid #d8e0eb;border-radius:18px;box-shadow:var(--shadow)}
+.toolbar{display:flex;justify-content:space-between;align-items:end;gap:18px;padding:18px 20px;position:relative;z-index:2}.toolbar-group,.segmented{display:flex;align-items:end;gap:10px;flex-wrap:wrap}
 label{display:grid;gap:6px;color:var(--muted);font-size:11px;font-weight:750;letter-spacing:.04em;text-transform:uppercase}input,select{min-height:39px;padding:8px 11px;color:var(--ink);background:#fbfcfe;border:1px solid #cfd8e5;border-radius:9px;font:inherit;text-transform:none;letter-spacing:0}
 button{min-height:39px;padding:8px 15px;border:1px solid #cfd8e5;border-radius:9px;color:#34445a;background:#fff;cursor:pointer;font-weight:700}button:hover,button.active{color:#fff;border-color:var(--blue);background:var(--blue)}
 .status{display:flex;align-items:center;gap:8px;color:var(--muted);font-size:12px;white-space:nowrap}.status:before{content:"";width:8px;height:8px;border-radius:50%;background:#24b47e;box-shadow:0 0 0 4px #24b47e1f}
@@ -19,4 +20,4 @@ button{min-height:39px;padding:8px 15px;border:1px solid #cfd8e5;border-radius:9
 .stability-grid{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:10px}.stability-card{padding:14px;border:1px solid var(--line);border-radius:12px;background:#fbfcfe}.stability-card span{display:block;color:var(--muted);font-size:10px}.stability-card strong{display:block;margin-top:6px;font-size:20px}.stability-card small{display:block;margin-top:5px;color:var(--muted)}.stock-summary{margin-bottom:12px}.consensus{margin-top:13px;padding:12px 14px;border-radius:11px;background:#edf9f7;color:#245b54;font-size:12px;line-height:1.6}.consensus b{color:#087a6d}
 .stability-details{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:12px;margin-top:12px}.stability-detail{padding:15px;border:1px solid var(--line);border-radius:12px}.stability-detail h3{margin:0 0 11px;font-size:14px}.stat-row{display:grid;grid-template-columns:repeat(3,1fr);gap:7px}.stat{padding:8px;border-radius:8px;background:#f4f7fa}.stat span{display:block;color:var(--muted);font-size:9px}.stat b{display:block;margin-top:4px;font-size:12px}.ticker-row{margin-top:10px;color:var(--muted);font-size:10px;line-height:1.55}.ticker-row b{color:var(--ink)}details{margin-top:8px;color:var(--muted);font-size:10px}summary{cursor:pointer;color:#425268;font-weight:700}.rule-insight{margin-top:12px;padding:12px 14px;border-left:4px solid var(--blue);border-radius:8px;background:#f2f5ff;color:#34445a;font-size:11px;line-height:1.6}
 .table-wrap{overflow-x:auto;border:1px solid var(--line);border-radius:12px}table{width:100%;border-collapse:collapse;background:#fff}th,td{padding:11px 13px;border-bottom:1px solid #edf1f5;font-size:12px;text-align:right;white-space:nowrap}th:first-child,td:first-child{text-align:left}th{color:#526075;background:#f7f9fc;font-size:10px;letter-spacing:.045em;text-transform:uppercase}tr:last-child td{border-bottom:0}.pos{color:#078071;font-weight:700}.neg{color:var(--red);font-weight:700}.muted,.empty{color:var(--muted)}.empty{padding:42px;text-align:center}.history-controls{display:flex;align-items:end;justify-content:flex-end;gap:10px;flex-wrap:wrap}.sort-button{min-height:0;padding:0;border:0;border-radius:0;color:inherit;background:transparent;font-size:inherit;font-weight:800;letter-spacing:inherit;text-transform:inherit}.sort-button:hover,.sort-button.active{color:var(--blue);background:transparent}.sort-arrow{display:inline-block;min-width:11px;margin-left:4px}.evaluation-note{margin:0 0 12px;padding:10px 12px;border-radius:9px;color:#536176;background:#f5f7fa;font-size:11px}.evaluation-note strong{color:var(--ink)}
-@media(max-width:980px){.kpi-grid,.stability-grid{grid-template-columns:repeat(2,1fr)}.model-grid,.stability-details{grid-template-columns:1fr}.toolbar,.pending-panel{align-items:stretch;flex-direction:column}}@media(max-width:620px){main{width:calc(100% - 20px)}header{padding-left:20px}.kpi-grid,.metric-grid,.parameter-grid,.stability-grid,.stat-row{grid-template-columns:1fr}}
+@media(max-width:980px){.kpi-grid,.stability-grid{grid-template-columns:repeat(2,1fr)}.model-grid,.stability-details{grid-template-columns:1fr}.toolbar,.pending-panel{align-items:stretch;flex-direction:column}.hero-row{align-items:start;flex-direction:column}.hero-badge{min-width:0}}@media(max-width:620px){main{width:calc(100% - 20px)}header{padding-left:20px;padding-right:20px}.kpi-grid,.metric-grid,.parameter-grid,.stability-grid,.stat-row{grid-template-columns:1fr}.page-switcher{display:grid}.page-link{min-width:0}.hero-badge{display:none}}
diff --git a/site/data/manifest.json b/site/data/manifest.json
index b95bbc8..28b9456 100644
--- a/site/data/manifest.json
+++ b/site/data/manifest.json
@@ -1,10 +1,13 @@
 {
-  "schema_version": 2,
-  "generated_at_utc": "2026-07-29T16:55:17.967215+00:00",
+  "schema_version": 3,
+  "generated_at_utc": "2026-07-29T22:29:54.012984+00:00",
   "latest_decision_date": "2026-07-28",
   "snapshot_count": 17,
   "evaluation_count": 14,
   "dashboard_sha256": "9d15420d6e5370038763505be97432e381f7a0932cc1aeac953399c030b2185c",
   "dashboard_bytes": 546502,
+  "horizon_dashboard_sha256": "20d6bdaa4d619ecca77441801a4f2e63b93c665822979b278cb490999bd4695a",
+  "horizon_dashboard_bytes": 66861,
+  "horizon_forecast_arms": 6,
   "source_policy": "sanitized_live_forward_outputs_only"
 }
\ No newline at end of file
diff --git a/site/horizons.html b/site/horizons.html
new file mode 100644
index 0000000..afff2dd
--- /dev/null
+++ b/site/horizons.html
@@ -0,0 +1,65 @@
+<!doctype html>
+<html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1">
+<title>Horizon-Matched Live Forward Monitor</title>
+<style>
+:root{--ink:#17172c;--muted:#68677d;--line:#dfddea;--surface:#fff;--canvas:#f6f4fa;--blue:#5c4bc4;--blue2:#8c7cdf;--teal:#0f9f8f;--teal2:#70c8bd;--amber:#d89614;--red:#c94c5b;--violet:#7559c7;--page-accent:#bca8ff;--shadow:0 14px 35px rgba(43,35,86,.09)}
+*{box-sizing:border-box}html{scroll-behavior:smooth}body{margin:0;color:var(--ink);background:var(--canvas);font-family:Inter,ui-sans-serif,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif}
+header{position:relative;overflow:hidden;padding:42px max(28px,calc((100vw - 1440px)/2)) 82px;color:#fff;background:linear-gradient(120deg,#171329,#2b2455 66%,#493b73)}
+header:after{content:"";position:absolute;right:-90px;top:-170px;width:440px;height:440px;border:1px solid #d7cbff2b;border-radius:50%;box-shadow:0 0 0 55px #bca8ff0c,0 0 0 115px #bca8ff09}
+.eyebrow{color:#cbbcff;font-size:11px;font-weight:850;letter-spacing:.16em;text-transform:uppercase}.hero-row{display:flex;justify-content:space-between;align-items:end;gap:28px;position:relative;z-index:1}
+h1{margin:8px 0 9px;font-size:clamp(30px,3.2vw,46px);letter-spacing:-.04em}.hero-copy{max-width:780px;margin:0;color:#bdc9da;line-height:1.55}
+.hero-badge{display:grid;gap:4px;min-width:190px;padding:14px 16px;border:1px solid #ffffff24;border-radius:14px;background:#ffffff0d;backdrop-filter:blur(6px)}.hero-badge span{color:#9fb0c8;font-size:10px;text-transform:uppercase;letter-spacing:.08em}.hero-badge strong{font-size:15px}
+.page-switcher{display:flex;gap:9px;margin-top:22px;position:relative;z-index:2}.page-link{display:grid;gap:3px;min-width:210px;padding:12px 15px;border:1px solid #ffffff28;border-radius:12px;color:#d5d0e5;background:#ffffff0a;text-decoration:none;transition:.18s ease}.page-link:hover{transform:translateY(-1px);border-color:var(--page-accent);background:#ffffff12}.page-link.active{color:#fff;border-color:var(--page-accent);background:#bca8ff17;box-shadow:inset 0 -3px 0 var(--page-accent)}.page-link strong{font-size:12px}.page-link small{color:#aaa4c0;font-size:9px}.page-link.active small{color:#d7cff1}
+main{width:min(1440px,calc(100% - 40px));margin:-52px auto 0;padding-bottom:58px;position:relative}.panel{background:var(--surface);border:1px solid #d8e0eb;border-radius:18px;box-shadow:var(--shadow)}
+.toolbar{display:flex;justify-content:space-between;align-items:end;gap:18px;padding:18px 20px}.toolbar-group,.segmented{display:flex;align-items:end;gap:10px;flex-wrap:wrap}
+label{display:grid;gap:6px;color:var(--muted);font-size:10px;font-weight:800;letter-spacing:.055em;text-transform:uppercase}select,input{min-height:39px;padding:8px 11px;color:var(--ink);background:#fbfcfe;border:1px solid #cfd8e5;border-radius:9px;font:inherit;text-transform:none;letter-spacing:0}
+button{min-height:39px;padding:8px 14px;border:1px solid #cfd8e5;border-radius:9px;color:#34445a;background:#fff;cursor:pointer;font-weight:750}button:hover,button.active{color:#fff;border-color:var(--blue);background:var(--blue)}
+.status{display:flex;align-items:center;gap:8px;color:var(--muted);font-size:12px;white-space:nowrap}.status:before{content:"";width:8px;height:8px;border-radius:50%;background:#24b47e;box-shadow:0 0 0 4px #24b47e1f}
+.notice{display:flex;justify-content:space-between;align-items:center;gap:18px;margin-top:16px;padding:15px 18px;border-color:#b9ddd7;background:linear-gradient(105deg,#f3fffc,#fff)}.notice-main{display:flex;align-items:center;gap:12px}.notice-icon{display:grid;place-items:center;flex:0 0 38px;height:38px;border-radius:12px;color:#fff;background:var(--teal);font-weight:900}.notice strong{display:block;font-size:13px}.notice small{display:block;margin-top:4px;color:var(--muted);line-height:1.45}.policy{padding:8px 11px;border:1px solid #f0d49c;border-radius:9px;color:#8b5a05;background:#fff8e8;font-size:10px;font-weight:800;text-transform:uppercase;letter-spacing:.04em}
+.kpi-grid{display:grid;grid-template-columns:repeat(4,minmax(155px,1fr));gap:14px;margin:16px 0}.kpi{position:relative;overflow:hidden;padding:17px 18px;border-top:4px solid var(--signal,var(--line))}.kpi span{color:var(--muted);font-size:11px;font-weight:700}.kpi strong{display:block;margin-top:9px;font-size:27px;letter-spacing:-.035em}.kpi small{display:block;margin-top:5px;color:var(--muted);font-size:10px;line-height:1.4}
+.section{margin-top:16px;padding:22px}.section-head{display:flex;justify-content:space-between;align-items:start;gap:18px;margin-bottom:17px}.section-head h2{margin:0;font-size:18px;letter-spacing:-.018em}.section-head p{max-width:780px;margin:5px 0 0;color:var(--muted);font-size:11px;line-height:1.5}
+.legend{display:flex;gap:7px;flex-wrap:wrap;justify-content:flex-end}.legend button{display:inline-flex;align-items:center;gap:7px;min-height:31px;padding:6px 10px;border-radius:999px;font-size:10px}.legend button i{width:16px;height:3px;border-radius:999px;background:var(--series)}.legend button.off{opacity:.43;text-decoration:line-through}
+.chart svg{display:block;width:100%;height:auto;min-height:330px}.empty{display:grid;place-items:center;min-height:210px;padding:36px;color:var(--muted);text-align:center}.empty strong{display:block;color:var(--ink);font-size:15px}.empty span{display:block;max-width:520px;margin-top:7px;font-size:11px;line-height:1.55}
+.progress-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:9px;margin-bottom:15px}.progress-card{padding:12px;border:1px solid var(--line);border-radius:11px;background:#fbfcfe}.progress-head{display:flex;justify-content:space-between;gap:10px;align-items:center}.progress-head strong{font-size:11px}.progress-head span{color:var(--muted);font-size:9px}.progress-track{height:6px;margin:9px 0 7px;overflow:hidden;border-radius:999px;background:#e7ecf3}.progress-track i{display:block;height:100%;border-radius:inherit;background:var(--series)}.progress-card small{color:var(--muted);font-size:9px;line-height:1.4}.interim-note{margin-top:10px;padding:9px 11px;border-left:3px solid var(--blue);border-radius:7px;color:#536176;background:#f2f5ff;font-size:9px;line-height:1.45}
+.horizon-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:13px}.horizon-card{padding:18px;border:1px solid var(--line);border-radius:14px;background:linear-gradient(180deg,#fff,#fbfcfe)}.horizon-title{display:flex;justify-content:space-between;align-items:center}.horizon-title h3{margin:0;font-size:16px}.clock{padding:5px 8px;border-radius:999px;color:#325278;background:#edf3ff;font-size:9px;font-weight:850;text-transform:uppercase}.pair{display:grid;grid-template-columns:1fr 1fr;gap:8px;margin-top:14px}.model-slice{padding:12px;border-left:4px solid var(--series);border-radius:10px;background:#f4f7fa}.model-slice span{display:block;color:var(--muted);font-size:10px}.model-slice strong{display:block;margin-top:6px;font-size:20px}.model-slice small{display:block;margin-top:5px;color:var(--muted);font-size:9px;line-height:1.45}.maturity{margin-top:12px;padding-top:11px;border-top:1px solid var(--line);color:var(--muted);font-size:10px}
+.explorer-controls{display:flex;gap:10px;align-items:end;flex-wrap:wrap}.search{min-width:220px}.holding-summary{display:grid;grid-template-columns:repeat(4,1fr);gap:9px;margin-bottom:12px}.mini{padding:12px;border:1px solid var(--line);border-radius:11px;background:#fbfcfe}.mini span{display:block;color:var(--muted);font-size:9px}.mini strong{display:block;margin-top:5px;font-size:16px}
+.table-wrap{overflow:auto;border:1px solid var(--line);border-radius:12px;max-height:560px}table{width:100%;border-collapse:collapse;background:#fff}th,td{padding:11px 13px;border-bottom:1px solid #edf1f5;font-size:11px;text-align:right;white-space:nowrap}th:first-child,td:first-child{text-align:left}th{position:sticky;top:0;z-index:1;color:#526075;background:#f7f9fc;font-size:9px;letter-spacing:.055em;text-transform:uppercase}tr:last-child td{border-bottom:0}tbody tr:hover{background:#f8faff}.sort{min-height:0;padding:0;border:0;border-radius:0;color:inherit;background:transparent;font-size:inherit;text-transform:inherit}.sort:hover{color:var(--blue);background:transparent}
+.model-grid{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:13px}.model-card{padding:18px;border:1px solid var(--line);border-radius:14px;background:#fff}.model-title{display:flex;justify-content:space-between;gap:12px;align-items:center}.model-title h3{margin:0;font-size:15px}.badge{padding:5px 8px;border-radius:999px;color:#0a7467;background:#e7f7f4;font-size:9px;font-weight:850;text-transform:uppercase}.metrics{display:grid;grid-template-columns:repeat(3,1fr);gap:7px;margin:13px 0}.metric{padding:9px;border-radius:9px;background:#f3f6fa}.metric span{display:block;color:var(--muted);font-size:8px}.metric strong{display:block;margin-top:5px;font-size:13px}.features{display:flex;gap:5px;flex-wrap:wrap;margin-top:10px}.feature{padding:4px 7px;border-radius:999px;color:#536176;background:#eef2f7;font-size:8px}.model-meta{color:var(--muted);font-size:9px;line-height:1.55}.explain-toggle{width:100%;margin-top:13px;color:#3156d3;background:#f5f7ff;border-color:#d9e0fa;font-size:10px}.explain-toggle:hover{color:#fff}.explain-panel{margin-top:10px;padding:12px;border:1px solid var(--line);border-radius:10px;background:#fbfcfe}.explain-panel[hidden]{display:none}.shap-row{display:grid;grid-template-columns:minmax(95px,1.3fr) 2fr 58px;gap:8px;align-items:center;margin:7px 0}.shap-row span{overflow:hidden;color:#526075;font-size:8px;text-overflow:ellipsis;white-space:nowrap}.shap-bar{height:8px;overflow:hidden;border-radius:999px;background:#e7ecf3}.shap-bar i{display:block;height:100%;border-radius:inherit;background:linear-gradient(90deg,#3156d3,#0f9f8f)}.shap-row strong{text-align:right;font-size:8px}.shap-empty{color:var(--muted);font-size:9px;line-height:1.55}
+.health-grid{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:9px}.health{padding:14px;border:1px solid var(--line);border-radius:11px;background:#fbfcfe}.health-head{display:flex;justify-content:space-between;align-items:center}.health-head span{font-size:10px;font-weight:800}.dot{width:8px;height:8px;border-radius:50%;background:#24b47e}.dot.warn{background:var(--amber)}.health strong{display:block;margin-top:9px;font-size:17px}.health small{display:block;margin-top:5px;color:var(--muted);font-size:9px;line-height:1.45}.warning-list{display:grid;gap:7px;margin-top:12px}.warning{padding:10px 12px;border-left:4px solid var(--amber);border-radius:8px;color:#665126;background:#fff8e8;font-size:10px;line-height:1.45}
+.pos{color:#078071}.neg{color:var(--red)}.muted{color:var(--muted)}
+@media(max-width:1050px){.horizon-grid{grid-template-columns:1fr}.model-grid{grid-template-columns:1fr}.health-grid,.holding-summary{grid-template-columns:repeat(2,1fr)}.hero-row{align-items:start;flex-direction:column}.hero-badge{min-width:0}}
+@media(max-width:700px){main{width:calc(100% - 20px)}header{padding-left:20px;padding-right:20px}.toolbar,.notice,.section-head{align-items:stretch;flex-direction:column}.kpi-grid,.health-grid,.holding-summary,.pair,.metrics,.progress-grid{grid-template-columns:1fr}.search{min-width:0;width:100%}.hero-badge{display:none}.page-switcher{display:grid}.page-link{min-width:0}.shap-row{grid-template-columns:92px 1fr 48px}}
+</style></head><body>
+<header><div class="hero-row"><div><div class="eyebrow">AI Portfolio Research</div><h1>Horizon Comparison</h1><p class="hero-copy">Six independently tuned portfolios tracked on their natural 5, 10, and 20-session clocks. Interim marks, matured outcomes, and validation evidence remain explicitly separated.</p><nav class="page-switcher" aria-label="Monitor pages"><a class="page-link" href="./index.html"><strong>Live Forward Monitor</strong><small>Daily evidence and shared stock-level view</small></a><a class="page-link active" href="./horizons.html" aria-current="page"><strong>Horizon Comparison</strong><small>5 / 10 / 20-session experiment</small></a></nav></div><div class="hero-badge"><span>Evidence status</span><strong id="heroStatus">Paper-live · collecting</strong><span id="heroDate"></span></div></div></header>
+<main>
+<div class="toolbar panel"><div class="toolbar-group"><label>Decision date<select id="decisionSelect"></select></label><label>Family<select id="familyFilter"><option value="all">All families</option><option value="ridge">Ridge</option><option value="xgb_ranker">XGBoost</option></select></label><label>Horizon<select id="horizonFilter"><option value="all">All horizons</option><option value="5">5 sessions</option><option value="10">10 sessions</option><option value="20">20 sessions</option></select></label></div><div class="toolbar-group"><button id="refreshData">Refresh data</button><div class="status" id="dataStatus">Frozen evidence loaded</div></div></div>
+<section class="notice panel"><div class="notice-main"><div class="notice-icon">→</div><div><strong id="noticeTitle"></strong><small id="noticeCopy"></small></div></div><div class="policy" id="policyBadge"></div></section>
+<div class="kpi-grid" id="kpis"></div>
+<section class="section panel"><div class="section-head"><div><h2>Prospective strategy evidence</h2><p>Partial mark-to-market outcomes appear as soon as an entered portfolio has a new observable open. They remain interim until the complete 5, 10, or 20-session window matures.</p></div><div class="legend" id="strategyLegend"></div></div><div class="progress-grid" id="strategyProgress"></div><div class="chart" id="strategyChart"></div><div class="interim-note">The decision session is not a holding session. A July 29 close signal enters at the July 30 open, so July 29 must remain 0/5 rather than borrowing a pre-entry return. Interim marks answer “what has happened so far?” and are never mixed with validation estimates.</div></section>
+<section class="section panel"><div class="section-head"><div><h2>Horizon comparison</h2><p>Paired Ridge and XGBoost development evidence with each arm’s own purge, feature subset, refit interval, and maturity date.</p></div></div><div class="horizon-grid" id="horizonGrid"></div></section>
+<section class="section panel"><div class="section-head"><div><h2>Selected model specifications</h2><p>The winning development configuration for each visible arm. Open a model’s explainability panel to inspect genuine SHAP importance when that frozen vintage captured it.</p></div></div><div class="model-grid" id="modelGrid"></div></section>
+<section class="section panel"><div class="section-head"><div><h2>Pipeline and risk controls</h2><p>Freshness, universe, NLP, tactical routing, survival guardrails, and artifact integrity for the selected decision date.</p></div></div><div class="health-grid" id="healthGrid"></div><div class="warning-list" id="warnings"></div></section>
+<section class="section panel"><div class="section-head"><div><h2>Matured forecast history</h2><p>Prospective outcomes appear only after their full label interval matures. Pending forecasts are not scored as zero.</p></div></div><div id="maturedTable"></div></section>
+</main>
+<script id="dashboard-data" type="application/json">{"schema_version": 1, "experiment": {"id": "live_forward_horizon_matched_v2", "status": "paper_live_research", "decision_date": "2026-07-29", "primary_horizon_sessions": 20, "comparison_design": "horizon_matched_refit_and_rebalance", "invested_weight": 1.0}, "rebalance_policy": {"active": "full_top40_refresh_on_horizon_clock", "status": "under_audit", "buffer_enabled": false, "next_comparison": ["full_refresh", "frozen_holdings", "rank_margin_buffer", "score_percentile_buffer"]}, "refit_sessions_by_horizon": {"20": 20, "10": 10, "5": 5}, "input_audit": {"mode": "live_refresh", "decision_date": "2026-07-29", "market_start": "2021-01-04", "market_end": "2026-07-29", "requested_stock_tickers": 621, "downloaded_stock_tickers": 564, "raw_news_rows_before": 316114, "raw_news_rows_after": 316264, "recent_news_rows": 900, "processed_news_rows": 900, "latest_news_publication": "2026-07-29T19:52:48+00:00", "latest_news_age_hours": 0.12, "finbert_model": "ProsusAI/finbert", "finbert_fallback_permitted": false, "nlp_feature_rows": 377318, "nlp_feature_tickers": 433}, "universe": {"maximum_names": 250, "current_liquid_names": 250, "hard_exclusions": 0, "ml_exclusions_activated": false, "ml_exclusions": 0, "final_names": 250}, "survival_validation": {"activated": false, "threshold": null, "recall": null, "exclusion_fraction": null, "minimum_yearly_recall": null, "reason": "no_threshold_passed_all_recall_and_coverage_gates", "training_rows": 25250, "validation_rows": 12750, "actual_exits": 14, "label_horizon_sessions": 63, "validation_start": "2025-04-28", "validation_end": "2026-04-27"}, "arms": [{"id": "2026-07-29:ridge:h5", "decision_date": "2026-07-29", "family": "ridge", "horizon_sessions": 5, "status": "pending", "expected_entry_date": "2026-07-30", "expected_maturity_date": "2026-08-06", "refit_interval_sessions": 5, "model_refit_date": "2026-07-29", "refitted_today": true, "strategy_rebalance": true, "strategic_rebalance": false, "eligible_names": 250, "selected_names": 40, "sentiment_coverage": 1.0, "feature_count": 18, "features": ["nlp_event_novelty", "nlp_sentiment_3d", "nlp_mention_count", "nlp_mention_spike", "sector_confidence", "nlp_recency_decay", "macd_signal", "macd_hist", "sector_margin", "log_volume", "nlp_sentiment_7d", "momentum_63", "nlp_sentiment_1d", "momentum_5", "macd", "sector_changed_21d", "returns_1d", "momentum_10"], "shap_summary": [], "explainability": {"status": "not_captured"}, "selected": [{"ticker": "AMD", "sector": "Information Technology", "rank": 1, "score": 0.0239320259508228, "weight": 0.025}, {"ticker": "MU", "sector": "Information Technology", "rank": 2, "score": 0.0168331371458403, "weight": 0.025}, {"ticker": "INTC", "sector": "Information Technology", "rank": 3, "score": 0.0164974099080078, "weight": 0.025}, {"ticker": "BAC", "sector": "Financials", "rank": 4, "score": 0.0162571878339871, "weight": 0.025}, {"ticker": "USB", "sector": "Financials", "rank": 5, "score": 0.0148509098068091, "weight": 0.025}, {"ticker": "DELL", "sector": "Information Technology", "rank": 6, "score": 0.0148172305805603, "weight": 0.025}, {"ticker": "HPE", "sector": "Information Technology", "rank": 7, "score": 0.0148033476515849, "weight": 0.025}, {"ticker": "WFC", "sector": "Financials", "rank": 8, "score": 0.0137293657284132, "weight": 0.025}, {"ticker": "AVGO", "sector": "Information Technology", "rank": 9, "score": 0.013713285388337, "weight": 0.025}, {"ticker": "AMAT", "sector": "Information Technology", "rank": 10, "score": 0.0136059337612303, "weight": 0.025}, {"ticker": "JPM", "sector": "Financials", "rank": 11, "score": 0.0120806132992142, "weight": 0.025}, {"ticker": "CRWD", "sector": "Information Technology", "rank": 12, "score": 0.0120091003208392, "weight": 0.025}, {"ticker": "ANET", "sector": "Information Technology", "rank": 13, "score": 0.011904621867478, "weight": 0.025}, {"ticker": "MRVL", "sector": "Information Technology", "rank": 14, "score": 0.0116829169932977, "weight": 0.025}, {"ticker": "BNY", "sector": "Financials", "rank": 20, "score": 0.0106962299045973, "weight": 0.025}, {"ticker": "CARR", "sector": "Industrials", "rank": 23, "score": 0.0100719969505096, "weight": 0.025}, {"ticker": "VLO", "sector": "Energy", "rank": 24, "score": 0.0097905479107003, "weight": 0.025}, {"ticker": "GE", "sector": "Industrials", "rank": 25, "score": 0.0097700330488633, "weight": 0.025}, {"ticker": "SCHW", "sector": "Financials", "rank": 27, "score": 0.0091428116969801, "weight": 0.025}, {"ticker": "HWM", "sector": "Industrials", "rank": 28, "score": 0.0090420291768451, "weight": 0.025}, {"ticker": "AXP", "sector": "Financials", "rank": 29, "score": 0.0090357436440269, "weight": 0.025}, {"ticker": "CAT", "sector": "Industrials", "rank": 30, "score": 0.0087415838024146, "weight": 0.025}, {"ticker": "MS", "sector": "Financials", "rank": 32, "score": 0.0085156825453713, "weight": 0.025}, {"ticker": "MPC", "sector": "Energy", "rank": 33, "score": 0.0083045724534949, "weight": 0.025}, {"ticker": "AEP", "sector": "Utilities", "rank": 35, "score": 0.0080568947765432, "weight": 0.025}, {"ticker": "PNC", "sector": "Financials", "rank": 36, "score": 0.0079131771701305, "weight": 0.025}, {"ticker": "PSX", "sector": "Energy", "rank": 37, "score": 0.0079010926566455, "weight": 0.025}, {"ticker": "C", "sector": "Financials", "rank": 38, "score": 0.0078578280946911, "weight": 0.025}, {"ticker": "CSX", "sector": "Industrials", "rank": 39, "score": 0.0077991473922432, "weight": 0.025}, {"ticker": "XEL", "sector": "Utilities", "rank": 41, "score": 0.0075692116501817, "weight": 0.025}, {"ticker": "MRK", "sector": "Health Care", "rank": 43, "score": 0.0072670150015648, "weight": 0.025}, {"ticker": "ABBV", "sector": "Health Care", "rank": 44, "score": 0.0070355476524056, "weight": 0.025}, {"ticker": "PG", "sector": "Consumer Staples", "rank": 46, "score": 0.0068951041738261, "weight": 0.025}, {"ticker": "CVS", "sector": "Health Care", "rank": 50, "score": 0.0062954448495779, "weight": 0.025}, {"ticker": "ETN", "sector": "Industrials", "rank": 51, "score": 0.0062204757530309, "weight": 0.025}, {"ticker": "TT", "sector": "Industrials", "rank": 53, "score": 0.0060471309639172, "weight": 0.025}, {"ticker": "D", "sector": "Utilities", "rank": 54, "score": 0.0060285369021248, "weight": 0.025}, {"ticker": "SO", "sector": "Utilities", "rank": 55, "score": 0.0057423162289187, "weight": 0.025}, {"ticker": "NEE", "sector": "Utilities", "rank": 57, "score": 0.0057031491206284, "weight": 0.025}, {"ticker": "BMY", "sector": "Health Care", "rank": 58, "score": 0.0056377631046637, "weight": 0.025}], "training_start": "2023-07-17", "training_end": "2026-07-21", "training_rows": 188931, "model_config": {"alpha": 7036.786707594506}, "validation": {"rank_ic": 0.0134896851215978, "positive_ic_rate": 0.5428571428571428, "top40_raw_forward_return": 0.0046637559868577, "universe_raw_forward_return": 0.0021617933714112, "top40_raw_excess": 0.0025019626154464, "resolved_feature_count": 18, "trial": 12, "selection_metric": "top40_raw_excess", "completed_trials": 30}}, {"id": "2026-07-29:ridge:h10", "decision_date": "2026-07-29", "family": "ridge", "horizon_sessions": 10, "status": "pending", "expected_entry_date": "2026-07-30", "expected_maturity_date": "2026-08-13", "refit_interval_sessions": 10, "model_refit_date": "2026-07-29", "refitted_today": true, "strategy_rebalance": true, "strategic_rebalance": false, "eligible_names": 250, "selected_names": 40, "sentiment_coverage": 1.0, "feature_count": 18, "features": ["nlp_event_novelty", "nlp_mention_count", "sector_confidence", "log_volume", "sector_margin", "macd_signal", "nlp_sentiment_3d", "macd", "nlp_sentiment_1d", "nlp_recency_decay", "sector_changed_21d", "nlp_mention_spike", "momentum_63", "residual_return", "rsi_14", "idiosyncratic_vol", "macd_hist", "momentum_5"], "shap_summary": [], "explainability": {"status": "not_captured"}, "selected": [{"ticker": "BAC", "sector": "Financials", "rank": 1, "score": 0.0207984700099512, "weight": 0.025}, {"ticker": "USB", "sector": "Financials", "rank": 2, "score": 0.0178121725924493, "weight": 0.025}, {"ticker": "OXY", "sector": "Energy", "rank": 3, "score": 0.0146553064657122, "weight": 0.025}, {"ticker": "WFC", "sector": "Financials", "rank": 4, "score": 0.014618806416624, "weight": 0.025}, {"ticker": "V", "sector": "Financials", "rank": 5, "score": 0.0144681033826645, "weight": 0.025}, {"ticker": "SCHW", "sector": "Financials", "rank": 6, "score": 0.0129495304921317, "weight": 0.025}, {"ticker": "JPM", "sector": "Financials", "rank": 7, "score": 0.0128865803015888, "weight": 0.025}, {"ticker": "NEE", "sector": "Utilities", "rank": 8, "score": 0.0128309871715624, "weight": 0.025}, {"ticker": "ABBV", "sector": "Health Care", "rank": 9, "score": 0.0125612159511164, "weight": 0.025}, {"ticker": "KO", "sector": "Consumer Staples", "rank": 10, "score": 0.0124407692062622, "weight": 0.025}, {"ticker": "ANET", "sector": "Information Technology", "rank": 11, "score": 0.0120578154994243, "weight": 0.025}, {"ticker": "PSX", "sector": "Energy", "rank": 12, "score": 0.011945835577228, "weight": 0.025}, {"ticker": "TFC", "sector": "Financials", "rank": 13, "score": 0.0119314467124573, "weight": 0.025}, {"ticker": "DVN", "sector": "Energy", "rank": 14, "score": 0.0117079545647763, "weight": 0.025}, {"ticker": "BMY", "sector": "Health Care", "rank": 15, "score": 0.0114887999834569, "weight": 0.025}, {"ticker": "AMD", "sector": "Information Technology", "rank": 16, "score": 0.0113703325574135, "weight": 0.025}, {"ticker": "MPC", "sector": "Energy", "rank": 17, "score": 0.0111834856637579, "weight": 0.025}, {"ticker": "EXC", "sector": "Utilities", "rank": 18, "score": 0.0111465766775146, "weight": 0.025}, {"ticker": "CL", "sector": "Consumer Staples", "rank": 19, "score": 0.0110682040329699, "weight": 0.025}, {"ticker": "VLO", "sector": "Energy", "rank": 20, "score": 0.0106631926951098, "weight": 0.025}, {"ticker": "SO", "sector": "Utilities", "rank": 21, "score": 0.0104884272995704, "weight": 0.025}, {"ticker": "PG", "sector": "Consumer Staples", "rank": 22, "score": 0.0104770915122394, "weight": 0.025}, {"ticker": "CSX", "sector": "Industrials", "rank": 23, "score": 0.0102370312388958, "weight": 0.025}, {"ticker": "XOM", "sector": "Energy", "rank": 24, "score": 0.0101063642458133, "weight": 0.025}, {"ticker": "AEP", "sector": "Utilities", "rank": 25, "score": 0.0099116565569126, "weight": 0.025}, {"ticker": "XEL", "sector": "Utilities", "rank": 26, "score": 0.0098892585923869, "weight": 0.025}, {"ticker": "MDLZ", "sector": "Consumer Staples", "rank": 27, "score": 0.0098777893818319, "weight": 0.025}, {"ticker": "HOOD", "sector": "Information Technology", "rank": 28, "score": 0.0097800143652501, "weight": 0.025}, {"ticker": "CVX", "sector": "Energy", "rank": 29, "score": 0.0097637615386033, "weight": 0.025}, {"ticker": "AVGO", "sector": "Information Technology", "rank": 30, "score": 0.009240193299289, "weight": 0.025}, {"ticker": "DUK", "sector": "Utilities", "rank": 31, "score": 0.0091882669713098, "weight": 0.025}, {"ticker": "JCI", "sector": "Industrials", "rank": 32, "score": 0.0086503793041933, "weight": 0.025}, {"ticker": "KLAC", "sector": "Information Technology", "rank": 33, "score": 0.0084712698823372, "weight": 0.025}, {"ticker": "EA", "sector": "Information Technology", "rank": 34, "score": 0.0083124514754122, "weight": 0.025}, {"ticker": "AXP", "sector": "Financials", "rank": 35, "score": 0.0082142512042838, "weight": 0.025}, {"ticker": "D", "sector": "Utilities", "rank": 36, "score": 0.0079778847493339, "weight": 0.025}, {"ticker": "CARR", "sector": "Industrials", "rank": 37, "score": 0.0078754113897004, "weight": 0.025}, {"ticker": "WMT", "sector": "Consumer Staples", "rank": 38, "score": 0.0078692625468846, "weight": 0.025}, {"ticker": "HPE", "sector": "Information Technology", "rank": 39, "score": 0.0078328611441567, "weight": 0.025}, {"ticker": "EOG", "sector": "Energy", "rank": 40, "score": 0.0077804427414052, "weight": 0.025}], "training_start": "2023-07-10", "training_end": "2026-07-14", "training_rows": 188931, "model_config": {"alpha": 3806.8472041686705}, "validation": {"rank_ic": 0.0288920308181152, "positive_ic_rate": 0.6238095238095239, "top40_raw_forward_return": 0.0089838495470688, "universe_raw_forward_return": 0.0045338297709489, "top40_raw_excess": 0.0044500197761198, "resolved_feature_count": 18, "trial": 14, "selection_metric": "top40_raw_excess", "completed_trials": 30}}, {"id": "2026-07-29:ridge:h20", "decision_date": "2026-07-29", "family": "ridge", "horizon_sessions": 20, "status": "pending", "expected_entry_date": "2026-07-30", "expected_maturity_date": "2026-08-27", "refit_interval_sessions": 20, "model_refit_date": "2026-07-29", "refitted_today": true, "strategy_rebalance": true, "strategic_rebalance": true, "eligible_names": 250, "selected_names": 40, "sentiment_coverage": 1.0, "feature_count": 18, "features": ["nlp_event_novelty", "nlp_mention_count", "sector_confidence", "log_volume", "sector_margin", "macd", "macd_signal", "rsi_14", "residual_return", "sector_changed_21d", "nlp_mention_spike", "sector_rel_rank", "cs_return_rank", "momentum_21", "idiosyncratic_vol", "nlp_sentiment_3d", "bb_pct", "rolling_beta"], "shap_summary": [], "explainability": {"status": "not_captured"}, "selected": [{"ticker": "BAC", "sector": "Financials", "rank": 1, "score": 0.0274824246754274, "weight": 0.025}, {"ticker": "USB", "sector": "Financials", "rank": 2, "score": 0.0204615060640226, "weight": 0.025}, {"ticker": "JPM", "sector": "Financials", "rank": 3, "score": 0.0197150783867123, "weight": 0.025}, {"ticker": "WFC", "sector": "Financials", "rank": 4, "score": 0.0158444841331232, "weight": 0.025}, {"ticker": "KO", "sector": "Consumer Staples", "rank": 5, "score": 0.0154105757578541, "weight": 0.025}, {"ticker": "NEE", "sector": "Utilities", "rank": 6, "score": 0.0154080217586129, "weight": 0.025}, {"ticker": "JCI", "sector": "Industrials", "rank": 7, "score": 0.0142506295032849, "weight": 0.025}, {"ticker": "V", "sector": "Financials", "rank": 8, "score": 0.0141506426501567, "weight": 0.025}, {"ticker": "CSX", "sector": "Industrials", "rank": 9, "score": 0.0140468493320577, "weight": 0.025}, {"ticker": "BMY", "sector": "Health Care", "rank": 10, "score": 0.0138940346089407, "weight": 0.025}, {"ticker": "XEL", "sector": "Utilities", "rank": 11, "score": 0.0137726090758553, "weight": 0.025}, {"ticker": "ABBV", "sector": "Health Care", "rank": 12, "score": 0.0133095150018597, "weight": 0.025}, {"ticker": "F", "sector": "Consumer Discretionary", "rank": 13, "score": 0.013132363125918, "weight": 0.025}, {"ticker": "GM", "sector": "Consumer Discretionary", "rank": 14, "score": 0.0129347668239911, "weight": 0.025}, {"ticker": "TFC", "sector": "Financials", "rank": 15, "score": 0.0128315075879097, "weight": 0.025}, {"ticker": "SO", "sector": "Utilities", "rank": 16, "score": 0.0127871024209264, "weight": 0.025}, {"ticker": "HPE", "sector": "Information Technology", "rank": 17, "score": 0.0125650442330199, "weight": 0.025}, {"ticker": "EXC", "sector": "Utilities", "rank": 18, "score": 0.0123573697073142, "weight": 0.025}, {"ticker": "SCHW", "sector": "Financials", "rank": 19, "score": 0.0117664368673144, "weight": 0.025}, {"ticker": "COF", "sector": "Financials", "rank": 20, "score": 0.0117474077636563, "weight": 0.025}, {"ticker": "BX", "sector": "Financials", "rank": 21, "score": 0.0116973083736034, "weight": 0.025}, {"ticker": "GE", "sector": "Industrials", "rank": 22, "score": 0.0116611157726241, "weight": 0.025}, {"ticker": "MDLZ", "sector": "Consumer Staples", "rank": 23, "score": 0.0113829888999397, "weight": 0.025}, {"ticker": "AMGN", "sector": "Health Care", "rank": 24, "score": 0.0107879082348509, "weight": 0.025}, {"ticker": "MRK", "sector": "Health Care", "rank": 25, "score": 0.0106069185341007, "weight": 0.025}, {"ticker": "DUK", "sector": "Utilities", "rank": 26, "score": 0.0105548025806047, "weight": 0.025}, {"ticker": "MU", "sector": "Information Technology", "rank": 27, "score": 0.0103532253898229, "weight": 0.025}, {"ticker": "AEP", "sector": "Utilities", "rank": 28, "score": 0.010332408036889, "weight": 0.025}, {"ticker": "PG", "sector": "Consumer Staples", "rank": 29, "score": 0.0100527792330357, "weight": 0.025}, {"ticker": "CL", "sector": "Consumer Staples", "rank": 30, "score": 0.0097139936596626, "weight": 0.025}, {"ticker": "CCL", "sector": "Consumer Discretionary", "rank": 31, "score": 0.0096929728049463, "weight": 0.025}, {"ticker": "ANET", "sector": "Information Technology", "rank": 32, "score": 0.0094307822347214, "weight": 0.025}, {"ticker": "AVGO", "sector": "Information Technology", "rank": 33, "score": 0.0091798589570561, "weight": 0.025}, {"ticker": "HWM", "sector": "Industrials", "rank": 34, "score": 0.0091405993462457, "weight": 0.025}, {"ticker": "D", "sector": "Utilities", "rank": 35, "score": 0.009087014056649, "weight": 0.025}, {"ticker": "JNJ", "sector": "Health Care", "rank": 36, "score": 0.0089380928576857, "weight": 0.025}, {"ticker": "XOM", "sector": "Energy", "rank": 37, "score": 0.0089228552335242, "weight": 0.025}, {"ticker": "WMT", "sector": "Consumer Staples", "rank": 38, "score": 0.0085535391386736, "weight": 0.025}, {"ticker": "ABT", "sector": "Health Care", "rank": 39, "score": 0.0084174948924252, "weight": 0.025}, {"ticker": "C", "sector": "Financials", "rank": 40, "score": 0.0077799212490036, "weight": 0.025}], "training_start": "2023-06-23", "training_end": "2026-06-29", "training_rows": 188931, "model_config": {"alpha": 18231.75165326811}, "validation": {"rank_ic": 0.0239729148698637, "positive_ic_rate": 0.6380952380952382, "top40_raw_forward_return": 0.0128884359249551, "universe_raw_forward_return": 0.0104353925100403, "top40_raw_excess": 0.0024530434149148, "resolved_feature_count": 18, "trial": 25, "selection_metric": "top40_raw_excess", "completed_trials": 30}}, {"id": "2026-07-29:xgb_ranker:h5", "decision_date": "2026-07-29", "family": "xgb_ranker", "horizon_sessions": 5, "status": "pending", "expected_entry_date": "2026-07-30", "expected_maturity_date": "2026-08-06", "refit_interval_sessions": 5, "model_refit_date": "2026-07-29", "refitted_today": true, "strategy_rebalance": true, "strategic_rebalance": false, "eligible_names": 250, "selected_names": 40, "sentiment_coverage": 1.0, "feature_count": 18, "features": ["nlp_event_novelty", "nlp_sentiment_3d", "nlp_mention_count", "nlp_mention_spike", "sector_confidence", "nlp_recency_decay", "macd_signal", "macd_hist", "sector_margin", "log_volume", "nlp_sentiment_7d", "momentum_63", "nlp_sentiment_1d", "momentum_5", "macd", "sector_changed_21d", "returns_1d", "momentum_10"], "shap_summary": [], "explainability": {"status": "not_captured"}, "selected": [{"ticker": "AMAT", "sector": "Information Technology", "rank": 1, "score": 0.0035098828, "weight": 0.025}, {"ticker": "LRCX", "sector": "Information Technology", "rank": 2, "score": 0.0035098828, "weight": 0.025}, {"ticker": "C", "sector": "Financials", "rank": 3, "score": 0.0034551718, "weight": 0.025}, {"ticker": "AMZN", "sector": "Consumer Discretionary", "rank": 4, "score": 0.0034551718, "weight": 0.025}, {"ticker": "KLAC", "sector": "Information Technology", "rank": 5, "score": 0.003231783, "weight": 0.025}, {"ticker": "CARR", "sector": "Industrials", "rank": 6, "score": 0.003231783, "weight": 0.025}, {"ticker": "WDC", "sector": "Information Technology", "rank": 7, "score": 0.003231783, "weight": 0.025}, {"ticker": "MRVL", "sector": "Information Technology", "rank": 8, "score": 0.003231783, "weight": 0.025}, {"ticker": "FTNT", "sector": "Information Technology", "rank": 9, "score": 0.003167998, "weight": 0.025}, {"ticker": "AMD", "sector": "Information Technology", "rank": 10, "score": 0.0029616794, "weight": 0.025}, {"ticker": "INTC", "sector": "Information Technology", "rank": 11, "score": 0.0029616794, "weight": 0.025}, {"ticker": "MU", "sector": "Information Technology", "rank": 12, "score": 0.0029616794, "weight": 0.025}, {"ticker": "NVDA", "sector": "Information Technology", "rank": 13, "score": 0.0029069688, "weight": 0.025}, {"ticker": "BMY", "sector": "Health Care", "rank": 15, "score": 0.0026477901, "weight": 0.025}, {"ticker": "WFC", "sector": "Financials", "rank": 16, "score": 0.002624696, "weight": 0.025}, {"ticker": "HAL", "sector": "Energy", "rank": 18, "score": 0.0024433506, "weight": 0.025}, {"ticker": "EXC", "sector": "Utilities", "rank": 19, "score": 0.0023776868, "weight": 0.025}, {"ticker": "KO", "sector": "Consumer Staples", "rank": 20, "score": 0.0023776868, "weight": 0.025}, {"ticker": "PEP", "sector": "Consumer Staples", "rank": 21, "score": 0.0023776868, "weight": 0.025}, {"ticker": "WMT", "sector": "Consumer Staples", "rank": 22, "score": 0.0023776868, "weight": 0.025}, {"ticker": "PG", "sector": "Consumer Staples", "rank": 23, "score": 0.0022445074, "weight": 0.025}, {"ticker": "DHR", "sector": "Health Care", "rank": 25, "score": 0.0022246523, "weight": 0.025}, {"ticker": "MDLZ", "sector": "Consumer Staples", "rank": 26, "score": 0.0021850823, "weight": 0.025}, {"ticker": "BAC", "sector": "Financials", "rank": 27, "score": 0.0021610777, "weight": 0.025}, {"ticker": "CSX", "sector": "Industrials", "rank": 28, "score": 0.0021295913, "weight": 0.025}, {"ticker": "NFLX", "sector": "Communication Services", "rank": 29, "score": 0.0019776432, "weight": 0.025}, {"ticker": "VZ", "sector": "Consumer Staples", "rank": 34, "score": 0.0017850387, "weight": 0.025}, {"ticker": "XOM", "sector": "Energy", "rank": 35, "score": 0.0017214642, "weight": 0.025}, {"ticker": "CVX", "sector": "Energy", "rank": 38, "score": 0.0015107861, "weight": 0.025}, {"ticker": "OXY", "sector": "Energy", "rank": 39, "score": 0.0015107861, "weight": 0.025}, {"ticker": "V", "sector": "Financials", "rank": 40, "score": 0.0014294401, "weight": 0.025}, {"ticker": "ABT", "sector": "Health Care", "rank": 41, "score": 0.0013613479, "weight": 0.025}, {"ticker": "ACN", "sector": "Financials", "rank": 44, "score": 0.0012093998, "weight": 0.025}, {"ticker": "BKR", "sector": "Energy", "rank": 45, "score": 0.0011617815, "weight": 0.025}, {"ticker": "META", "sector": "Communication Services", "rank": 47, "score": 0.0006128809, "weight": 0.025}, {"ticker": "CVNA", "sector": "Consumer Discretionary", "rank": 49, "score": 8.3599094e-05, "weight": 0.025}, {"ticker": "PFE", "sector": "Health Care", "rank": 50, "score": 8.3599094e-05, "weight": 0.025}, {"ticker": "F", "sector": "Consumer Discretionary", "rank": 51, "score": 8.3599094e-05, "weight": 0.025}, {"ticker": "WDAY", "sector": "Financials", "rank": 53, "score": -0.00034009168, "weight": 0.025}, {"ticker": "CTSH", "sector": "Financials", "rank": 54, "score": -0.00049203984, "weight": 0.025}], "training_start": "2023-07-17", "training_end": "2026-07-21", "training_rows": 188931, "model_config": {"colsample_bytree": 0.7324244130856432, "gamma": 1.2055081153486717, "learning_rate": 0.007171504310432957, "max_depth": 3, "min_child_weight": 23.944112374919623, "n_estimators": 10, "n_jobs": 4, "reg_alpha": 0.2223947739271475, "reg_lambda": 7.528657637770702, "subsample": 0.6644885149016018, "tree_method": "hist"}, "validation": {"rank_ic": 0.0254034726983806, "positive_ic_rate": 0.6047619047619048, "top40_raw_forward_return": 0.0058696047851267, "universe_raw_forward_return": 0.0021617933714112, "top40_raw_excess": 0.0037078114137154, "resolved_feature_count": 18, "trial": 8, "median_effective_n_estimators": 10.0, "selection_metric": "top40_raw_excess", "completed_trials": 25}}, {"id": "2026-07-29:xgb_ranker:h10", "decision_date": "2026-07-29", "family": "xgb_ranker", "horizon_sessions": 10, "status": "pending", "expected_entry_date": "2026-07-30", "expected_maturity_date": "2026-08-13", "refit_interval_sessions": 10, "model_refit_date": "2026-07-29", "refitted_today": true, "strategy_rebalance": true, "strategic_rebalance": false, "eligible_names": 250, "selected_names": 40, "sentiment_coverage": 1.0, "feature_count": 32, "features": ["nlp_event_novelty", "nlp_mention_count", "sector_confidence", "log_volume", "sector_margin", "macd_signal", "nlp_sentiment_3d", "macd", "nlp_sentiment_1d", "nlp_recency_decay", "sector_changed_21d", "nlp_mention_spike", "momentum_63", "residual_return", "rsi_14", "idiosyncratic_vol", "macd_hist", "momentum_5", "returns_1d", "nlp_sentiment_7d", "sector_rel_rank", "rolling_beta", "cs_return_rank", "momentum_21", "overnight_gap", "atr_14", "volume_ratio_20d", "vol_10d", "vol_63d", "vol_21d", "bb_pct", "momentum_10"], "shap_summary": [], "explainability": {"status": "not_captured"}, "selected": [{"ticker": "INTC", "sector": "Information Technology", "rank": 1, "score": 0.005496101, "weight": 0.025}, {"ticker": "AVGO", "sector": "Information Technology", "rank": 2, "score": 0.0048822686, "weight": 0.025}, {"ticker": "QCOM", "sector": "Information Technology", "rank": 3, "score": 0.0048822686, "weight": 0.025}, {"ticker": "ORCL", "sector": "Information Technology", "rank": 4, "score": 0.0048822686, "weight": 0.025}, {"ticker": "TXN", "sector": "Industrials", "rank": 5, "score": 0.004417613, "weight": 0.025}, {"ticker": "CARR", "sector": "Industrials", "rank": 6, "score": 0.0042929277, "weight": 0.025}, {"ticker": "META", "sector": "Communication Services", "rank": 7, "score": 0.0042929277, "weight": 0.025}, {"ticker": "FCX", "sector": "Information Technology", "rank": 8, "score": 0.004133967, "weight": 0.025}, {"ticker": "MCHP", "sector": "Information Technology", "rank": 9, "score": 0.004133967, "weight": 0.025}, {"ticker": "CAT", "sector": "Industrials", "rank": 10, "score": 0.0039514303, "weight": 0.025}, {"ticker": "JBL", "sector": "Information Technology", "rank": 11, "score": 0.0039514303, "weight": 0.025}, {"ticker": "NXPI", "sector": "Information Technology", "rank": 12, "score": 0.0039514303, "weight": 0.025}, {"ticker": "PWR", "sector": "Industrials", "rank": 13, "score": 0.0039514303, "weight": 0.025}, {"ticker": "ANET", "sector": "Information Technology", "rank": 14, "score": 0.0039514303, "weight": 0.025}, {"ticker": "ETN", "sector": "Industrials", "rank": 15, "score": 0.0039514303, "weight": 0.025}, {"ticker": "TSLA", "sector": "Consumer Discretionary", "rank": 16, "score": 0.0038778677, "weight": 0.025}, {"ticker": "AMD", "sector": "Information Technology", "rank": 19, "score": 0.0038667105, "weight": 0.025}, {"ticker": "CVNA", "sector": "Consumer Discretionary", "rank": 21, "score": 0.003747523, "weight": 0.025}, {"ticker": "CCL", "sector": "Consumer Discretionary", "rank": 22, "score": 0.003747523, "weight": 0.025}, {"ticker": "AMZN", "sector": "Consumer Discretionary", "rank": 25, "score": 0.0033972152, "weight": 0.025}, {"ticker": "ADI", "sector": "Industrials", "rank": 27, "score": 0.0033620892, "weight": 0.025}, {"ticker": "CMI", "sector": "Industrials", "rank": 28, "score": 0.0033620892, "weight": 0.025}, {"ticker": "MS", "sector": "Financials", "rank": 29, "score": 0.0033620892, "weight": 0.025}, {"ticker": "GOOGL", "sector": "Communication Services", "rank": 30, "score": 0.0031581817, "weight": 0.025}, {"ticker": "GOOG", "sector": "Communication Services", "rank": 31, "score": 0.0031581817, "weight": 0.025}, {"ticker": "FIX", "sector": "Industrials", "rank": 35, "score": 0.0029358722, "weight": 0.025}, {"ticker": "GEV", "sector": "Industrials", "rank": 36, "score": 0.0029358722, "weight": 0.025}, {"ticker": "DASH", "sector": "Financials", "rank": 43, "score": 0.0028166845, "weight": 0.025}, {"ticker": "CEG", "sector": "Utilities", "rank": 47, "score": 0.0024663769, "weight": 0.025}, {"ticker": "C", "sector": "Financials", "rank": 49, "score": 0.0022624694, "weight": 0.025}, {"ticker": "DIS", "sector": "Communication Services", "rank": 50, "score": 0.0022624694, "weight": 0.025}, {"ticker": "XYZ", "sector": "Consumer Discretionary", "rank": 52, "score": 0.0022273434, "weight": 0.025}, {"ticker": "TEL", "sector": "Industrials", "rank": 54, "score": 0.0022273434, "weight": 0.025}, {"ticker": "NEM", "sector": "Materials", "rank": 55, "score": 0.0022273434, "weight": 0.025}, {"ticker": "UAL", "sector": "Consumer Discretionary", "rank": 59, "score": 0.0020683834, "weight": 0.025}, {"ticker": "BNY", "sector": "Financials", "rank": 61, "score": 0.0017363721, "weight": 0.025}, {"ticker": "DLR", "sector": "Real Estate", "rank": 62, "score": 0.0014512584, "weight": 0.025}, {"ticker": "BAC", "sector": "Financials", "rank": 63, "score": 0.0014428811, "weight": 0.025}, {"ticker": "BKNG", "sector": "Consumer Discretionary", "rank": 76, "score": 0.0013316311, "weight": 0.025}, {"ticker": "ABNB", "sector": "Consumer Discretionary", "rank": 79, "score": 0.0013316311, "weight": 0.025}], "training_start": "2023-07-10", "training_end": "2026-07-14", "training_rows": 188931, "model_config": {"colsample_bytree": 0.6391712616169326, "gamma": 1.0815238388714232, "learning_rate": 0.008868626294351352, "max_depth": 3, "min_child_weight": 42.84822736083466, "n_estimators": 9, "n_jobs": 4, "reg_alpha": 0.17432822729399894, "reg_lambda": 0.6430114013111514, "subsample": 0.8782947246006406, "tree_method": "hist"}, "validation": {"rank_ic": 0.0253241020991792, "positive_ic_rate": 0.6904761904761905, "top40_raw_forward_return": 0.0127963076855659, "universe_raw_forward_return": 0.0045338297709489, "top40_raw_excess": 0.008262477914617, "resolved_feature_count": 32, "trial": 24, "median_effective_n_estimators": 9.0, "selection_metric": "top40_raw_excess", "completed_trials": 25}}, {"id": "2026-07-29:xgb_ranker:h20", "decision_date": "2026-07-29", "family": "xgb_ranker", "horizon_sessions": 20, "status": "pending", "expected_entry_date": "2026-07-30", "expected_maturity_date": "2026-08-27", "refit_interval_sessions": 20, "model_refit_date": "2026-07-29", "refitted_today": true, "strategy_rebalance": true, "strategic_rebalance": true, "eligible_names": 250, "selected_names": 40, "sentiment_coverage": 1.0, "feature_count": 32, "features": ["nlp_event_novelty", "nlp_mention_count", "sector_confidence", "log_volume", "sector_margin", "macd", "macd_signal", "rsi_14", "residual_return", "sector_changed_21d", "nlp_mention_spike", "sector_rel_rank", "cs_return_rank", "momentum_21", "idiosyncratic_vol", "nlp_sentiment_3d", "bb_pct", "rolling_beta", "momentum_63", "momentum_10", "nlp_sentiment_1d", "nlp_recency_decay", "atr_14", "momentum_5", "vol_63d", "vol_10d", "volume_ratio_20d", "vol_21d", "returns_1d", "overnight_gap", "nlp_sentiment_7d", "macd_hist"], "shap_summary": [], "explainability": {"status": "not_captured"}, "selected": [{"ticker": "ANET", "sector": "Information Technology", "rank": 1, "score": 0.008621738, "weight": 0.025}, {"ticker": "CVNA", "sector": "Consumer Discretionary", "rank": 2, "score": 0.008571395, "weight": 0.025}, {"ticker": "ETN", "sector": "Industrials", "rank": 3, "score": 0.008564489, "weight": 0.025}, {"ticker": "CAT", "sector": "Industrials", "rank": 4, "score": 0.008336625, "weight": 0.025}, {"ticker": "HOOD", "sector": "Information Technology", "rank": 5, "score": 0.008259865, "weight": 0.025}, {"ticker": "AMAT", "sector": "Information Technology", "rank": 6, "score": 0.00824884, "weight": 0.025}, {"ticker": "TSLA", "sector": "Consumer Discretionary", "rank": 7, "score": 0.008079288, "weight": 0.025}, {"ticker": "CCL", "sector": "Consumer Discretionary", "rank": 8, "score": 0.007856587, "weight": 0.025}, {"ticker": "ORCL", "sector": "Information Technology", "rank": 9, "score": 0.007790266, "weight": 0.025}, {"ticker": "NXPI", "sector": "Information Technology", "rank": 10, "score": 0.007790266, "weight": 0.025}, {"ticker": "HPE", "sector": "Information Technology", "rank": 11, "score": 0.007702481, "weight": 0.025}, {"ticker": "STX", "sector": "Information Technology", "rank": 12, "score": 0.007524962, "weight": 0.025}, {"ticker": "SNDK", "sector": "Information Technology", "rank": 13, "score": 0.007504287, "weight": 0.025}, {"ticker": "GEV", "sector": "Industrials", "rank": 14, "score": 0.0074359216, "weight": 0.025}, {"ticker": "MRVL", "sector": "Information Technology", "rank": 15, "score": 0.0071952613, "weight": 0.025}, {"ticker": "LITE", "sector": "Information Technology", "rank": 16, "score": 0.0071014687, "weight": 0.025}, {"ticker": "TXN", "sector": "Industrials", "rank": 18, "score": 0.0068784133, "weight": 0.025}, {"ticker": "CEG", "sector": "Utilities", "rank": 27, "score": 0.0060239425, "weight": 0.025}, {"ticker": "DASH", "sector": "Financials", "rank": 28, "score": 0.0058616763, "weight": 0.025}, {"ticker": "FIX", "sector": "Industrials", "rank": 30, "score": 0.005499245, "weight": 0.025}, {"ticker": "PWR", "sector": "Industrials", "rank": 32, "score": 0.0054728254, "weight": 0.025}, {"ticker": "ADI", "sector": "Industrials", "rank": 37, "score": 0.003811608, "weight": 0.025}, {"ticker": "NEM", "sector": "Materials", "rank": 38, "score": 0.0036516208, "weight": 0.025}, {"ticker": "INTU", "sector": "Financials", "rank": 39, "score": 0.0036308824, "weight": 0.025}, {"ticker": "CARR", "sector": "Industrials", "rank": 40, "score": 0.0032851202, "weight": 0.025}, {"ticker": "CMI", "sector": "Industrials", "rank": 46, "score": 0.0026795757, "weight": 0.025}, {"ticker": "RCL", "sector": "Consumer Discretionary", "rank": 48, "score": 0.0024461322, "weight": 0.025}, {"ticker": "HPQ", "sector": "Communication Services", "rank": 52, "score": 0.0018332893, "weight": 0.025}, {"ticker": "F", "sector": "Consumer Discretionary", "rank": 54, "score": 0.0016223685, "weight": 0.025}, {"ticker": "BA", "sector": "Industrials", "rank": 57, "score": 0.0012341901, "weight": 0.025}, {"ticker": "TRV", "sector": "Real Estate", "rank": 59, "score": 0.0009192403, "weight": 0.025}, {"ticker": "VRSK", "sector": "Financials", "rank": 60, "score": 0.000865735, "weight": 0.025}, {"ticker": "CRH", "sector": "Materials", "rank": 63, "score": 0.00074302754, "weight": 0.025}, {"ticker": "XYZ", "sector": "Consumer Discretionary", "rank": 64, "score": 0.0007260397, "weight": 0.025}, {"ticker": "MRNA", "sector": "Consumer Discretionary", "rank": 65, "score": 0.00054263533, "weight": 0.025}, {"ticker": "DAL", "sector": "Consumer Discretionary", "rank": 66, "score": 0.00054162974, "weight": 0.025}, {"ticker": "GOOGL", "sector": "Communication Services", "rank": 67, "score": 0.0003181083, "weight": 0.025}, {"ticker": "GOOG", "sector": "Communication Services", "rank": 68, "score": 0.0003181083, "weight": 0.025}, {"ticker": "BX", "sector": "Financials", "rank": 76, "score": -0.00034842803, "weight": 0.025}, {"ticker": "GM", "sector": "Consumer Discretionary", "rank": 78, "score": -0.000590802, "weight": 0.025}], "training_start": "2023-06-23", "training_end": "2026-06-29", "training_rows": 188931, "model_config": {"colsample_bytree": 0.5348734651630105, "gamma": 1.1777639420895203, "learning_rate": 0.007192888541657487, "max_depth": 4, "min_child_weight": 2.0248445091398266, "n_estimators": 14, "n_jobs": 4, "reg_alpha": 0.04490198072606135, "reg_lambda": 0.3202585241999403, "subsample": 0.8447411578889518, "tree_method": "hist"}, "validation": {"rank_ic": 0.0245857780753604, "positive_ic_rate": 0.6142857142857142, "top40_raw_forward_return": 0.0293820254375538, "universe_raw_forward_return": 0.0104353925100403, "top40_raw_excess": 0.0189466329275135, "resolved_feature_count": 32, "trial": 1, "median_effective_n_estimators": 14.0, "selection_metric": "top40_raw_excess", "completed_trials": 25}}], "matured": [], "strategy_returns": [], "runs": [{"decision_date": "2026-07-29", "status": "live_forward_20d_cycle_complete", "matured_forecasts": 0, "strategy_daily_return_rows": 0, "tactical_orders": 0, "model_failures": 0}], "tactical_orders": []}</script>
+<script>
+const data=JSON.parse(document.getElementById('dashboard-data').textContent),decisionSelect=document.getElementById('decisionSelect'),familyFilter=document.getElementById('familyFilter'),horizonFilter=document.getElementById('horizonFilter'),STATE_KEY='live-forward-horizon-dashboard-v1';
+const COLORS={ridge:'#3156d3',xgb_ranker:'#0f9f8f',ridge_h5:'#7290eb',ridge_h10:'#4f70dc',ridge_h20:'#3156d3',xgb_ranker_h5:'#70c8bd',xgb_ranker_h10:'#3eb2a4',xgb_ranker_h20:'#0f9f8f',spy:'#d89614'},names={ridge:'Ridge',xgb_ranker:'XGBoost'},fmtPct=v=>v==null||!Number.isFinite(Number(v))?'n/a':(Number(v)*100).toFixed(2)+'%',fmtNum=(v,d=3)=>v==null||!Number.isFinite(Number(v))?'n/a':Number(v).toFixed(d),human=v=>String(v??'').replaceAll('_',' ').replace(/\b\w/g,c=>c.toUpperCase()),tone=v=>Number(v)>=0?'pos':'neg',armColor=a=>COLORS[`${a.family}_h${a.horizon_sessions}`]||COLORS[a.family],armLabel=a=>`${names[a.family]||human(a.family)} · ${a.horizon_sessions} sessions`;
+let saved={};try{saved=JSON.parse(localStorage.getItem(STATE_KEY)||'{}')}catch{}let hiddenSeries=new Set(saved.hiddenSeries||[]);
+function save(){try{localStorage.setItem(STATE_KEY,JSON.stringify({decisionDate:decisionSelect.value,family:familyFilter.value,horizon:horizonFilter.value,hiddenSeries:[...hiddenSeries]}))}catch{}}
+function option(value,label){const node=document.createElement('option');node.value=value;node.textContent=label;return node}
+const dates=[...new Set(data.arms.map(a=>a.decision_date))].sort().reverse();dates.forEach(d=>decisionSelect.appendChild(option(d,d)));decisionSelect.value=dates.includes(saved.decisionDate)?saved.decisionDate:dates[0]||'';familyFilter.value=['ridge','xgb_ranker'].includes(saved.family)?saved.family:'all';horizonFilter.value=['5','10','20'].includes(saved.horizon)?saved.horizon:'all';
+function decisionArms(){return data.arms.filter(a=>a.decision_date===decisionSelect.value)}
+function visibleArms(){return decisionArms().filter(a=>(familyFilter.value==='all'||a.family===familyFilter.value)&&(horizonFilter.value==='all'||String(a.horizon_sessions)===horizonFilter.value))}
+function table(rows,columns,sortState){if(!rows.length)return'<div class="empty"><div><strong>No matching records</strong><span>Adjust the filters or wait for the required outcome window.</span></div></div>';const head=columns.map(([key,label])=>`<th><button class="sort" data-sort="${key}">${label}${sortState?.key===key?(sortState.direction==='asc'?' ↑':' ↓'):''}</button></th>`).join(''),body=rows.map(row=>`<tr>${columns.map(([key])=>{const value=row[key],shown=key.includes('return')||key==='weight'||key==='sentiment_coverage'?fmtPct(value):key==='score'?fmtNum(value,5):value??'—';return`<td class="${key.includes('return')&&value!=null?tone(value):''}">${shown}</td>`}).join('')}</tr>`).join('');return`<div class="table-wrap"><table><thead><tr>${head}</tr></thead><tbody>${body}</tbody></table></div>`}
+function renderNotice(){const arms=decisionArms(),pending=arms.filter(a=>a.status==='pending').length,next=[...new Set(arms.map(a=>a.expected_entry_date).filter(Boolean))].sort()[0],policy=data.rebalance_policy||{};document.getElementById('noticeTitle').textContent=`${pending} frozen forecasts ready for the next session`;document.getElementById('noticeCopy').textContent=`Decision ${decisionSelect.value} · expected entry ${next||'pending'} · partial marks update session by session; final labels wait for full maturity.`;document.getElementById('policyBadge').textContent=`${human(policy.status||'under audit')} · buffer ${policy.buffer_enabled?'on':'off'}`;document.getElementById('heroDate').textContent=`Latest decision · ${dates[0]||'n/a'}`}
+function renderKpis(){const arms=visibleArms(),matured=data.matured.filter(x=>!decisionSelect.value||x.decision_date<=decisionSelect.value),audit=data.input_audit||{},universe=data.universe||{},items=[['Active arms',arms.length,'#3156d3',`${decisionArms().length} total for this decision`],['Matured forecasts',matured.length,'#7559c7',matured.length?'Prospective outcomes available':'First outcomes still pending'],['Eligible universe',universe.final_names,'#0f9f8f',`${universe.current_liquid_names??0} liquid names`],['Sentiment coverage',arms[0]?.sentiment_coverage,'#d89614',`${audit.processed_news_rows??0} recent articles processed`]];document.getElementById('kpis').innerHTML=items.map(([label,value,color,detail],i)=>`<article class="kpi panel" style="--signal:${color}"><span>${label}</span><strong>${i===3?fmtPct(value):value??'n/a'}</strong><small>${detail}</small></article>`).join('');document.getElementById('dataStatus').textContent=`${dates.length} decision date${dates.length===1?'':'s'} · ${data.arms.length} frozen forecasts`}
+function strategyGroups(){const by=new Map,spyByDate=new Map;data.strategy_returns.forEach(row=>{const id=`${row.family}_h${row.horizon_sessions}`;if(!by.has(id))by.set(id,[]);by.get(id).push(row);if(Number.isFinite(Number(row.spy_return))&&!spyByDate.has(row.return_date))spyByDate.set(row.return_date,{return_date:row.return_date,gross_return:Number(row.spy_return)})});if(spyByDate.size)by.set('spy',[...spyByDate.values()]);return [...by.entries()].map(([id,rows])=>({id,rows:rows.sort((a,b)=>String(a.return_date).localeCompare(String(b.return_date)))}))}
+function renderStrategy(){const arms=visibleArms(),visibleIds=new Set([...arms.map(a=>`${a.family}_h${a.horizon_sessions}`),'spy']),groups=strategyGroups().filter(g=>visibleIds.has(g.id)),legend=document.getElementById('strategyLegend');document.getElementById('strategyProgress').innerHTML=arms.map(a=>{const rows=data.strategy_returns.filter(r=>r.family===a.family&&Number(r.horizon_sessions)===a.horizon_sessions&&r.rebalance_decision_date===a.decision_date).sort((x,y)=>String(x.return_date).localeCompare(String(y.return_date))),observed=rows.length,cumulative=rows.reduce((value,row)=>value*(1+Number(row.gross_return)),1)-1,spyCumulative=rows.reduce((value,row)=>Number.isFinite(Number(row.spy_return))?value*(1+Number(row.spy_return)):value,1)-1,asOf=data.experiment.decision_date,status=observed?`${fmtPct(cumulative)} interim · ${fmtPct(cumulative-spyCumulative)} vs SPY`:(asOf<a.expected_entry_date?`Awaiting ${a.expected_entry_date} open`:'Entered · first mark pending');return`<article class="progress-card"><div class="progress-head"><strong>${armLabel(a)}</strong><span>${observed}/${a.horizon_sessions} sessions held</span></div><div class="progress-track"><i style="width:${Math.min(100,observed/a.horizon_sessions*100)}%;--series:${armColor(a)}"></i></div><small>${status} · complete label ${a.expected_maturity_date}</small></article>`}).join('');legend.innerHTML=groups.map(g=>`<button class="${hiddenSeries.has(g.id)?'off':'active'}" data-series="${g.id}" style="--series:${COLORS[g.id]||'#7559c7'}"><i></i>${g.id==='spy'?'SPY':human(g.id.replace('_h',' · h'))}</button>`).join('');legend.querySelectorAll('[data-series]').forEach(b=>b.addEventListener('click',()=>{hiddenSeries.has(b.dataset.series)?hiddenSeries.delete(b.dataset.series):hiddenSeries.add(b.dataset.series);renderStrategy();save()}));const visible=groups.filter(g=>!hiddenSeries.has(g.id));if(!visible.length){const entries=arms.map(a=>a.expected_entry_date).filter(Boolean).sort();document.getElementById('strategyChart').innerHTML=`<div class="empty"><div><strong>No observable post-entry mark yet</strong><span>The ${decisionSelect.value} close decision first becomes invested at the ${entries[0]||'next configured'} open. The chart begins with the first post-entry open-to-open return and does not wait for full maturity.</span></div></div>`;return}const lines=visible.map(g=>{let cumulative=1;return{id:g.id,points:g.rows.map(r=>({date:r.return_date,value:(cumulative*=1+Number(r.gross_return))-1}))}}),stamps=[...new Set(lines.flatMap(l=>l.points.map(p=>p.date)))].sort(),vals=lines.flatMap(l=>l.points.map(p=>p.value)),min=Math.min(0,...vals),max=Math.max(0,...vals),pad=Math.max((max-min)*.14,.001),w=1180,h=360,left=66,right=22,top=18,bottom=48,pw=w-left-right,ph=h-top-bottom,x=s=>left+stamps.indexOf(s)/Math.max(1,stamps.length-1)*pw,y=v=>top+(1-(v-(min-pad))/((max+pad)-(min-pad)||1))*ph,ticks=[0,.25,.5,.75,1].map(f=>max+pad-f*((max+pad)-(min-pad))),grid=ticks.map(v=>`<line x1="${left}" x2="${left+pw}" y1="${y(v)}" y2="${y(v)}" stroke="#e6ebf2"/><text x="${left-9}" y="${y(v)+4}" text-anchor="end" fill="#69768a" font-size="10">${fmtPct(v)}</text>`).join(''),paths=lines.map(line=>`<path d="${line.points.map((p,i)=>`${i?'L':'M'}${x(p.date)},${y(p.value)}`).join(' ')}" fill="none" stroke="${COLORS[line.id]}" stroke-width="${line.id==='spy'?2:3}" stroke-dasharray="${line.id==='spy'?'6 5':''}" stroke-linecap="round"/>`).join('');document.getElementById('strategyChart').innerHTML=`<svg viewBox="0 0 ${w} ${h}">${grid}<line x1="${left}" x2="${left+pw}" y1="${y(0)}" y2="${y(0)}" stroke="#9aa7b8" stroke-dasharray="5 5"/>${paths}</svg>`}
+function renderHorizons(){const all=decisionArms(),cards=[5,10,20].map(h=>{const pair=all.filter(a=>a.horizon_sessions===h).sort((a,b)=>a.family.localeCompare(b.family)),slices=pair.map(a=>`<div class="model-slice" style="--series:${armColor(a)}"><span>${names[a.family]}</span><strong class="${tone(a.validation?.top40_raw_excess)}">${fmtPct(a.validation?.top40_raw_excess)}</strong><small>Validation top-40 excess<br>Rank IC ${fmtNum(a.validation?.rank_ic)} · ${a.feature_count} features</small></div>`).join(''),maturity=pair[0]?.expected_maturity_date||'pending';return`<article class="horizon-card"><div class="horizon-title"><h3>${h}-session horizon</h3><span class="clock">Refit every ${h}</span></div><div class="pair">${slices}</div><div class="maturity">Expected first maturity · <strong>${maturity}</strong></div></article>`});document.getElementById('horizonGrid').innerHTML=cards.join('')}
+function renderModels(){const node=document.getElementById('modelGrid');node.innerHTML=visibleArms().map(a=>{const v=a.validation||{},features=a.features.slice(0,12).map(f=>`<span class="feature">${human(f)}</span>`).join(''),extra=Math.max(0,a.features.length-12),shap=a.shap_summary||[],max=Math.max(...shap.map(x=>Number(x.mean_abs_shap)||0),0),chart=shap.length?`<div class="model-meta">${human(shap[0].method)} · ${shap[0].explanation_rows} current-cohort rows · mean absolute SHAP</div>${shap.slice(0,12).map(x=>`<div class="shap-row"><span title="${human(x.feature)}">${human(x.feature)}</span><div class="shap-bar"><i style="width:${max?Number(x.mean_abs_shap)/max*100:0}%"></i></div><strong>${fmtNum(x.mean_abs_shap,4)}</strong></div>`).join('')}`:`<div class="shap-empty"><strong>SHAP was not captured for this frozen vintage.</strong><br>The July 29 artifacts remain immutable. Genuine ${a.family==='ridge'?'linear SHAP':'TreeSHAP'} summaries will be saved with each new forecast; coefficients or XGBoost gain are not substituted or mislabeled as SHAP.</div>`;return`<article class="model-card"><div class="model-title"><h3>${armLabel(a)}</h3><span class="badge">${a.refitted_today?'Refit today':'Frozen model'}</span></div><div class="model-meta">Training ${a.training_start||'n/a'} to ${a.training_end||'n/a'} · ${Number(a.training_rows||0).toLocaleString()} rows · ${v.completed_trials||'n/a'} eligible trials</div><div class="metrics"><div class="metric"><span>Top-40 excess</span><strong class="${tone(v.top40_raw_excess)}">${fmtPct(v.top40_raw_excess)}</strong></div><div class="metric"><span>Rank IC</span><strong>${fmtNum(v.rank_ic)}</strong></div><div class="metric"><span>Positive IC</span><strong>${fmtPct(v.positive_ic_rate)}</strong></div></div><div class="model-meta">Selected ${a.feature_count} features · ${human(v.selection_metric||'n/a')}</div><div class="features">${features}${extra?`<span class="feature">+${extra} more</span>`:''}</div><button class="explain-toggle" type="button" data-explain="${a.id}" aria-expanded="false">Show SHAP feature importance</button><div class="explain-panel" data-explain-panel="${a.id}" hidden>${chart}</div></article>`}).join('')||'<div class="empty"><div><strong>No visible model specifications</strong><span>Change the family or horizon filter.</span></div></div>';node.querySelectorAll('[data-explain]').forEach(button=>button.addEventListener('click',()=>{const panel=node.querySelector(`[data-explain-panel="${button.dataset.explain}"]`),opening=panel.hidden;panel.hidden=!opening;button.setAttribute('aria-expanded',String(opening));button.textContent=opening?'Hide SHAP feature importance':'Show SHAP feature importance'}))}
+function renderHealth(){const audit=data.input_audit||{},universe=data.universe||{},survival=data.survival_validation||{},tactical=data.tactical_orders||[],cards=[['Market data',audit.market_end===data.experiment.decision_date?'Current':'Review',`${audit.downloaded_stock_tickers??0}/${audit.requested_stock_tickers??0} histories · through ${audit.market_end||'n/a'}`,audit.market_end===data.experiment.decision_date],['News + NLP',audit.finbert_fallback_permitted===false?'Strict FinBERT':'Review',`${audit.processed_news_rows??0} articles · ${audit.nlp_feature_tickers??0} tickers`,audit.finbert_fallback_permitted===false],['Universe',`${universe.final_names??0} eligible`,`Hard exclusions ${universe.hard_exclusions??0} · ML exclusions ${universe.ml_exclusions??0}`,Number(universe.final_names)>=40],['Tactical routing',`${tactical.length} orders`,tactical.length?'Review routed orders':'No event passed every frozen gate',true],['Survival ML',survival.activated?'Active':'Shadow only',survival.activated?'Activation gates passed':human(survival.reason||'not activated'),!survival.activated],['Artifacts',`${decisionArms().length} forecasts`,`${decisionArms().filter(a=>a.status==='pending').length} pending · immutable manifests`,decisionArms().length===6],['Sentiment freshness',audit.latest_news_age_hours!=null?`${Number(audit.latest_news_age_hours).toFixed(2)} hours`:'n/a',`Latest publication ${audit.latest_news_publication||'n/a'}`,Number(audit.latest_news_age_hours)<=72],['Failures',`${(data.runs.at(-1)?.model_failures)||0} model failures`,`Latest cycle ${data.runs.at(-1)?.status||'n/a'}`,!(data.runs.at(-1)?.model_failures)]],node=document.getElementById('healthGrid');node.innerHTML=cards.map(([label,value,detail,ok])=>`<article class="health"><div class="health-head"><span>${label}</span><i class="dot ${ok?'':'warn'}"></i></div><strong>${value}</strong><small>${detail}</small></article>`).join('');const warnings=[];if(!survival.activated)warnings.push('Survival ML remains shadow-only; no stock is excluded by an unvalidated threshold.');if(data.rebalance_policy?.status==='under_audit')warnings.push('Full top-40 refresh remains under audit; the rank/score buffer has not been activated.');if(!data.matured.length)warnings.push('No v2 forecast has matured. Today’s successful run is operational evidence, not realized alpha.');document.getElementById('warnings').innerHTML=warnings.map(x=>`<div class="warning">${x}</div>`).join('')}
+function renderMatured(){const rows=data.matured.filter(x=>(familyFilter.value==='all'||x.family===familyFilter.value)&&(horizonFilter.value==='all'||String(x.horizon_sessions)===horizonFilter.value));document.getElementById('maturedTable').innerHTML=table(rows,[['decision_date','Decision'],['family','Model'],['horizon_sessions','Horizon'],['selected_return','Selected return'],['active_return_vs_universe','Excess vs universe'],['active_return_vs_spy','Excess vs SPY']],null)}
+function renderAll(){renderNotice();renderKpis();renderStrategy();renderHorizons();renderModels();renderHealth();renderMatured();save()}
+[decisionSelect,familyFilter,horizonFilter].forEach(node=>node.addEventListener('change',renderAll));document.getElementById('refreshData').addEventListener('click',()=>{save();location.reload()});renderAll();
+</script></body></html>
\ No newline at end of file
diff --git a/site/index.html b/site/index.html
index fe49769..6375b63 100644
--- a/site/index.html
+++ b/site/index.html
@@ -2,7 +2,7 @@
 <html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1">
 <title>Live Forward Research Monitor</title>
 <link rel="stylesheet" href="./assets/dashboard.css"></head><body>
-<header><div class="eyebrow">AI Portfolio Research</div><h1>Live Forward Monitor</h1><p>Out-of-sample model portfolios tracked against SPY. Every session uses the portfolio frozen before its evaluation window.</p></header>
+<header><div class="hero-row"><div><div class="eyebrow">AI Portfolio Research</div><h1>Live Forward Monitor</h1><p>Out-of-sample model portfolios tracked against SPY. Every session uses the portfolio frozen before its evaluation window.</p><nav class="page-switcher" aria-label="Monitor pages"><a class="page-link active" href="dashboard.html" aria-current="page"><strong>Live Forward Monitor</strong><small>Default · daily evidence and stock-level view</small></a><a class="page-link" href="./horizons.html"><strong>Horizon Comparison</strong><small>5 / 10 / 20-session experiment</small></a></nav></div><div class="hero-badge"><span>Default view</span><strong>Daily evidence</strong><span>Frozen-cohort monitor</span></div></div></header>
 <main>
 <div class="toolbar panel"><div class="toolbar-group"><label>Start date<input id="startDate" type="date"></label><label>End date<input id="endDate" type="date"></label><div class="segmented"><button data-range="latest">Latest session</button><button data-range="5">5 sessions</button><button data-range="all" class="active">All available</button></div></div><div class="toolbar-group"><button id="refreshData" title="Reload deployed dashboard data">Refresh data</button><div class="status" id="dataStatus">Frozen evidence loaded</div></div></div>
 <section class="pending-panel panel" id="nextSession"></section>
@@ -10,7 +10,7 @@
 <section class="section panel"><div class="section-head"><div><h2>Portfolio vs Benchmark</h2><p>Compounded return across selected live sessions. Models may begin on different dates; use the controls to show or hide individual portfolios.</p></div><div class="legend" id="legend"></div></div><div id="chart"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Holding-period laboratory</h2><p>Compare both models while keeping a dated top-40 list or replaying a chosen rebalance interval. SPY is always shown from the same decision-close baseline.</p></div><div class="cohort-controls"><label>Analysis<select id="cohortMode"><option value="cohort">Keep one stock list</option><option value="policy">Rebalance horizon</option></select></label><label id="cohortChoiceLabel">Starting list<select id="cohortChoice"></select></label><label id="rebalanceHorizonLabel">Rebalance every<input id="rebalanceHorizon" class="horizon-input" type="number" min="1" max="60" step="1" value="5"></label><div class="horizon-presets"><button type="button" data-cohort-horizon="5">5</button><button type="button" data-cohort-horizon="10">10</button><button type="button" data-cohort-horizon="20">20</button></div></div></div><div class="cohort-models" id="cohortModelToggles"></div><div id="cohortLab"></div><div id="horizonComparison"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Selection stability</h2><p>Daily model agreement, retained holdings, rank consistency, and implied equal-weight replacement.</p></div></div><div id="stability"></div></section>
-<section class="section panel"><div class="section-head"><div><h2>Stock-level view</h2><p>Inspect a frozen portfolio for one decision date. Realized returns appear after a holding-level evaluation is available.</p></div><div class="toolbar-group"><label>Decision date<select id="stockDate"></select></label><label>Model<select id="stockModel"></select></label></div></div><div id="holdingTable"></div></section>
+<section class="section panel" id="stock-level"><div class="section-head"><div><h2>Stock-level view</h2><p>Inspect a frozen portfolio for one decision date. Realized returns appear after a holding-level evaluation is available.</p></div><div class="toolbar-group"><label>Decision date<select id="stockDate"></select></label><label>Model<select id="stockModel"></select></label></div></div><div id="holdingTable"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Selected models</h2><p>Only the winning specification refit for each live candidate is shown; candidate trials are omitted.</p></div></div><div class="model-grid" id="modelCards"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Experiment history</h2><p>Newest evaluated evidence is shown first. Filter the archive and sort any column as it grows.</p></div><div class="history-controls"><label>Browse by<select id="historyGroup"><option value="evaluation_date">Evaluation date</option><option value="model">Selected model</option></select></label><label id="historyValueLabel">Evaluation date<select id="historyValue"></select></label><label>Rows<select id="historyLimit"><option value="6">Latest 6</option><option value="12">Latest 12</option><option value="all">All matching</option></select></label></div></div><div id="historyTable"></div></section>
 </main><script id="dashboard-data" type="application/json">{"snapshots": [], "cohort_monitor": {}}</script><script src="./assets/loader.js" defer></script></body></html>
\ No newline at end of file
diff --git a/src/experiments/live_dashboard.py b/src/experiments/live_dashboard.py
index 1cf5bd8..6d9d907 100644
--- a/src/experiments/live_dashboard.py
+++ b/src/experiments/live_dashboard.py
@@ -71,11 +71,12 @@ def dashboard_site_assets() -> dict[str, str]:
 <style>
 :root{--ink:#102033;--muted:#637083;--line:#dce3ec;--surface:#fff;--canvas:#f4f7fb;--blue:#3156d3;--teal:#0f9f8f;--amber:#d89614;--red:#c94c5b;--shadow:0 14px 35px rgba(24,39,75,.08)}
 *{box-sizing:border-box}body{margin:0;color:var(--ink);background:var(--canvas);font-family:Inter,ui-sans-serif,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif}
-header{position:relative;overflow:hidden;padding:38px max(28px,calc((100vw - 1440px)/2));color:#fff;background:linear-gradient(120deg,#081525,#12294b 70%,#173c57)}
-header:after{content:"";position:absolute;right:-90px;top:-160px;width:420px;height:420px;border:1px solid #ffffff21;border-radius:50%;box-shadow:0 0 0 55px #ffffff0a,0 0 0 115px #ffffff08}
-.eyebrow{color:#7dd3c7;font-size:11px;font-weight:800;letter-spacing:.16em;text-transform:uppercase}h1{margin:7px 0 8px;font-size:clamp(28px,3vw,42px);letter-spacing:-.035em}header p{max-width:760px;margin:0;color:#bdc9da;line-height:1.55}
-main{width:min(1440px,calc(100% - 40px));margin:auto;padding:26px 0 54px}.panel{background:var(--surface);border:1px solid #d8e0eb;border-radius:18px;box-shadow:var(--shadow)}
-.toolbar{display:flex;justify-content:space-between;align-items:end;gap:20px;padding:18px 20px;margin-top:-50px;position:relative;z-index:2}.toolbar-group,.segmented{display:flex;align-items:end;gap:10px;flex-wrap:wrap}
+header{position:relative;overflow:hidden;padding:42px max(28px,calc((100vw - 1440px)/2)) 82px;color:#fff;background:linear-gradient(120deg,#081525,#12294b 68%,#173c57)}
+header:after{content:"";position:absolute;right:-90px;top:-170px;width:440px;height:440px;border:1px solid #ffffff21;border-radius:50%;box-shadow:0 0 0 55px #ffffff0a,0 0 0 115px #ffffff08}
+.hero-row{display:flex;justify-content:space-between;align-items:end;gap:28px;position:relative;z-index:1}.eyebrow{color:#7dd3c7;font-size:11px;font-weight:850;letter-spacing:.16em;text-transform:uppercase}h1{margin:8px 0 9px;font-size:clamp(30px,3.2vw,46px);letter-spacing:-.04em}header p{max-width:780px;margin:0;color:#bdc9da;line-height:1.55}.hero-badge{display:grid;gap:4px;min-width:190px;padding:14px 16px;border:1px solid #ffffff24;border-radius:14px;background:#ffffff0d;backdrop-filter:blur(6px)}.hero-badge span{color:#9fb0c8;font-size:10px;text-transform:uppercase;letter-spacing:.08em}.hero-badge strong{font-size:15px}
+.page-switcher{display:flex;gap:9px;margin-top:22px;position:relative;z-index:2}.page-link{display:grid;gap:3px;min-width:210px;padding:12px 15px;border:1px solid #ffffff28;border-radius:12px;color:#c5d2e3;background:#ffffff0a;text-decoration:none;transition:.18s ease}.page-link:hover{transform:translateY(-1px);border-color:#7dd3c7;background:#ffffff12}.page-link.active{color:#fff;border-color:#7dd3c7;background:#7dd3c719;box-shadow:inset 0 -3px 0 #7dd3c7}.page-link strong{font-size:12px}.page-link small{color:#95a8c0;font-size:9px}.page-link.active small{color:#b9d8d3}
+main{width:min(1440px,calc(100% - 40px));margin:-52px auto 0;padding-bottom:58px;position:relative}.panel{background:var(--surface);border:1px solid #d8e0eb;border-radius:18px;box-shadow:var(--shadow)}
+.toolbar{display:flex;justify-content:space-between;align-items:end;gap:18px;padding:18px 20px;position:relative;z-index:2}.toolbar-group,.segmented{display:flex;align-items:end;gap:10px;flex-wrap:wrap}
 label{display:grid;gap:6px;color:var(--muted);font-size:11px;font-weight:750;letter-spacing:.04em;text-transform:uppercase}input,select{min-height:39px;padding:8px 11px;color:var(--ink);background:#fbfcfe;border:1px solid #cfd8e5;border-radius:9px;font:inherit;text-transform:none;letter-spacing:0}
 button{min-height:39px;padding:8px 15px;border:1px solid #cfd8e5;border-radius:9px;color:#34445a;background:#fff;cursor:pointer;font-weight:700}button:hover,button.active{color:#fff;border-color:var(--blue);background:var(--blue)}
 .status{display:flex;align-items:center;gap:8px;color:var(--muted);font-size:12px;white-space:nowrap}.status:before{content:"";width:8px;height:8px;border-radius:50%;background:#24b47e;box-shadow:0 0 0 4px #24b47e1f}
@@ -89,9 +90,9 @@ def dashboard_site_assets() -> dict[str, str]:
 .stability-grid{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:10px}.stability-card{padding:14px;border:1px solid var(--line);border-radius:12px;background:#fbfcfe}.stability-card span{display:block;color:var(--muted);font-size:10px}.stability-card strong{display:block;margin-top:6px;font-size:20px}.stability-card small{display:block;margin-top:5px;color:var(--muted)}.stock-summary{margin-bottom:12px}.consensus{margin-top:13px;padding:12px 14px;border-radius:11px;background:#edf9f7;color:#245b54;font-size:12px;line-height:1.6}.consensus b{color:#087a6d}
 .stability-details{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:12px;margin-top:12px}.stability-detail{padding:15px;border:1px solid var(--line);border-radius:12px}.stability-detail h3{margin:0 0 11px;font-size:14px}.stat-row{display:grid;grid-template-columns:repeat(3,1fr);gap:7px}.stat{padding:8px;border-radius:8px;background:#f4f7fa}.stat span{display:block;color:var(--muted);font-size:9px}.stat b{display:block;margin-top:4px;font-size:12px}.ticker-row{margin-top:10px;color:var(--muted);font-size:10px;line-height:1.55}.ticker-row b{color:var(--ink)}details{margin-top:8px;color:var(--muted);font-size:10px}summary{cursor:pointer;color:#425268;font-weight:700}.rule-insight{margin-top:12px;padding:12px 14px;border-left:4px solid var(--blue);border-radius:8px;background:#f2f5ff;color:#34445a;font-size:11px;line-height:1.6}
 .table-wrap{overflow-x:auto;border:1px solid var(--line);border-radius:12px}table{width:100%;border-collapse:collapse;background:#fff}th,td{padding:11px 13px;border-bottom:1px solid #edf1f5;font-size:12px;text-align:right;white-space:nowrap}th:first-child,td:first-child{text-align:left}th{color:#526075;background:#f7f9fc;font-size:10px;letter-spacing:.045em;text-transform:uppercase}tr:last-child td{border-bottom:0}.pos{color:#078071;font-weight:700}.neg{color:var(--red);font-weight:700}.muted,.empty{color:var(--muted)}.empty{padding:42px;text-align:center}.history-controls{display:flex;align-items:end;justify-content:flex-end;gap:10px;flex-wrap:wrap}.sort-button{min-height:0;padding:0;border:0;border-radius:0;color:inherit;background:transparent;font-size:inherit;font-weight:800;letter-spacing:inherit;text-transform:inherit}.sort-button:hover,.sort-button.active{color:var(--blue);background:transparent}.sort-arrow{display:inline-block;min-width:11px;margin-left:4px}.evaluation-note{margin:0 0 12px;padding:10px 12px;border-radius:9px;color:#536176;background:#f5f7fa;font-size:11px}.evaluation-note strong{color:var(--ink)}
-@media(max-width:980px){.kpi-grid,.stability-grid{grid-template-columns:repeat(2,1fr)}.model-grid,.stability-details{grid-template-columns:1fr}.toolbar,.pending-panel{align-items:stretch;flex-direction:column}}@media(max-width:620px){main{width:calc(100% - 20px)}header{padding-left:20px}.kpi-grid,.metric-grid,.parameter-grid,.stability-grid,.stat-row{grid-template-columns:1fr}}
+@media(max-width:980px){.kpi-grid,.stability-grid{grid-template-columns:repeat(2,1fr)}.model-grid,.stability-details{grid-template-columns:1fr}.toolbar,.pending-panel{align-items:stretch;flex-direction:column}.hero-row{align-items:start;flex-direction:column}.hero-badge{min-width:0}}@media(max-width:620px){main{width:calc(100% - 20px)}header{padding-left:20px;padding-right:20px}.kpi-grid,.metric-grid,.parameter-grid,.stability-grid,.stat-row{grid-template-columns:1fr}.page-switcher{display:grid}.page-link{min-width:0}.hero-badge{display:none}}
 </style></head><body>
-<header><div class="eyebrow">AI Portfolio Research</div><h1>Live Forward Monitor</h1><p>Out-of-sample model portfolios tracked against SPY. Every session uses the portfolio frozen before its evaluation window.</p></header>
+<header><div class="hero-row"><div><div class="eyebrow">AI Portfolio Research</div><h1>Live Forward Monitor</h1><p>Out-of-sample model portfolios tracked against SPY. Every session uses the portfolio frozen before its evaluation window.</p><nav class="page-switcher" aria-label="Monitor pages"><a class="page-link active" href="dashboard.html" aria-current="page"><strong>Live Forward Monitor</strong><small>Default · daily evidence and stock-level view</small></a><a class="page-link" href="../live_forward_horizon_matched_v2/dashboard.html"><strong>Horizon Comparison</strong><small>5 / 10 / 20-session experiment</small></a></nav></div><div class="hero-badge"><span>Default view</span><strong>Daily evidence</strong><span>Frozen-cohort monitor</span></div></div></header>
 <main>
 <div class="toolbar panel"><div class="toolbar-group"><label>Start date<input id="startDate" type="date"></label><label>End date<input id="endDate" type="date"></label><div class="segmented"><button data-range="latest">Latest session</button><button data-range="5">5 sessions</button><button data-range="all" class="active">All available</button></div></div><div class="toolbar-group"><button id="refreshData" title="Reload deployed dashboard data">Refresh data</button><div class="status" id="dataStatus">Frozen evidence loaded</div></div></div>
 <section class="pending-panel panel" id="nextSession"></section>
@@ -99,7 +100,7 @@ def dashboard_site_assets() -> dict[str, str]:
 <section class="section panel"><div class="section-head"><div><h2>Portfolio vs Benchmark</h2><p>Compounded return across selected live sessions. Models may begin on different dates; use the controls to show or hide individual portfolios.</p></div><div class="legend" id="legend"></div></div><div id="chart"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Holding-period laboratory</h2><p>Compare both models while keeping a dated top-40 list or replaying a chosen rebalance interval. SPY is always shown from the same decision-close baseline.</p></div><div class="cohort-controls"><label>Analysis<select id="cohortMode"><option value="cohort">Keep one stock list</option><option value="policy">Rebalance horizon</option></select></label><label id="cohortChoiceLabel">Starting list<select id="cohortChoice"></select></label><label id="rebalanceHorizonLabel">Rebalance every<input id="rebalanceHorizon" class="horizon-input" type="number" min="1" max="60" step="1" value="5"></label><div class="horizon-presets"><button type="button" data-cohort-horizon="5">5</button><button type="button" data-cohort-horizon="10">10</button><button type="button" data-cohort-horizon="20">20</button></div></div></div><div class="cohort-models" id="cohortModelToggles"></div><div id="cohortLab"></div><div id="horizonComparison"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Selection stability</h2><p>Daily model agreement, retained holdings, rank consistency, and implied equal-weight replacement.</p></div></div><div id="stability"></div></section>
-<section class="section panel"><div class="section-head"><div><h2>Stock-level view</h2><p>Inspect a frozen portfolio for one decision date. Realized returns appear after a holding-level evaluation is available.</p></div><div class="toolbar-group"><label>Decision date<select id="stockDate"></select></label><label>Model<select id="stockModel"></select></label></div></div><div id="holdingTable"></div></section>
+<section class="section panel" id="stock-level"><div class="section-head"><div><h2>Stock-level view</h2><p>Inspect a frozen portfolio for one decision date. Realized returns appear after a holding-level evaluation is available.</p></div><div class="toolbar-group"><label>Decision date<select id="stockDate"></select></label><label>Model<select id="stockModel"></select></label></div></div><div id="holdingTable"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Selected models</h2><p>Only the winning specification refit for each live candidate is shown; candidate trials are omitted.</p></div></div><div class="model-grid" id="modelCards"></div></section>
 <section class="section panel"><div class="section-head"><div><h2>Experiment history</h2><p>Newest evaluated evidence is shown first. Filter the archive and sort any column as it grows.</p></div><div class="history-controls"><label>Browse by<select id="historyGroup"><option value="evaluation_date">Evaluation date</option><option value="model">Selected model</option></select></label><label id="historyValueLabel">Evaluation date<select id="historyValue"></select></label><label>Rows<select id="historyLimit"><option value="6">Latest 6</option><option value="12">Latest 12</option><option value="all">All matching</option></select></label></div></div><div id="historyTable"></div></section>
 </main><script id="dashboard-data" type="application/json">__DASHBOARD_PAYLOAD__</script><script>
diff --git a/src/experiments/live_dashboard_publish.py b/src/experiments/live_dashboard_publish.py
index 44196b1..9b90033 100644
--- a/src/experiments/live_dashboard_publish.py
+++ b/src/experiments/live_dashboard_publish.py
@@ -7,7 +7,13 @@
 from pathlib import Path
 from typing import Any
 
+import pandas as pd
+
 from .live_dashboard import dashboard_site_assets
+from .live_forward_20d_dashboard import (
+    build_horizon_dashboard_payload,
+    horizon_dashboard_html,
+)
 from .live_forward_test import _collect_dashboard_snapshots
 from .public_dashboard_validation import dashboard_sha256, validate_public_dashboard_bundle
 
@@ -237,9 +243,65 @@ def sanitize_cohort_monitor(monitor: dict[str, Any] | None) -> dict[str, Any]:
     return public
 
 
+def _public_horizon_page(
+    horizon_root: Path,
+    latest_decision: str,
+) -> tuple[str, int]:
+    latest_run = horizon_root / "latest_run.json"
+    if latest_run.exists():
+        summary = json.loads(latest_run.read_text(encoding="utf-8"))
+        matured_path = horizon_root / "matured_forecasts.csv"
+        matured = (
+            pd.read_csv(matured_path)
+            if matured_path.exists()
+            else pd.DataFrame()
+        )
+        payload = build_horizon_dashboard_payload(
+            horizon_root,
+            summary,
+            matured,
+        )
+    else:
+        payload = {
+            "schema_version": 1,
+            "experiment": {
+                "id": "live_forward_horizon_matched_v2",
+                "status": "awaiting_first_public_snapshot",
+                "decision_date": latest_decision,
+                "primary_horizon_sessions": 20,
+                "comparison_design": "horizon_matched",
+                "invested_weight": 1.0,
+            },
+            "rebalance_policy": {
+                "status": "under_audit",
+                "buffer_enabled": False,
+            },
+            "refit_sessions_by_horizon": {"5": 5, "10": 10, "20": 20},
+            "input_audit": {},
+            "universe": {},
+            "survival_validation": {},
+            "arms": [],
+            "matured": [],
+            "strategy_returns": [],
+            "runs": [],
+            "tactical_orders": [],
+        }
+    html = horizon_dashboard_html(payload)
+    html = html.replace(
+        'href="../live_forward/dashboard.html"',
+        'href="./index.html"',
+    ).replace(
+        'href="dashboard.html" aria-current="page"',
+        'href="./horizons.html" aria-current="page"',
+    )
+    return html, len(payload.get("arms", []))
+
+
 def publish_live_dashboard(
     live_root: str | Path = "reports/live_forward",
     output_dir: str | Path = "site",
+    *,
+    horizon_root: str | Path | None = None,
 ) -> dict[str, Any]:
     """Build and validate the public GitHub Pages dashboard bundle."""
     source = Path(live_root)
@@ -269,22 +331,40 @@ def publish_live_dashboard(
     ).encode("utf-8")
     (data_dir / "dashboard.json").write_bytes(dashboard_bytes)
 
+    latest_decision = max(item["decision_date"] for item in snapshots)
     assets = dashboard_site_assets()
+    assets["index.html"] = assets["index.html"].replace(
+        'href="../live_forward_horizon_matched_v2/dashboard.html"',
+        'href="./horizons.html"',
+    )
     for relative, contents in assets.items():
         target = destination / relative
         target.parent.mkdir(parents=True, exist_ok=True)
         target.write_text(contents, encoding="utf-8")
+    horizon_source = (
+        Path(horizon_root)
+        if horizon_root is not None
+        else source.parent / "live_forward_horizon_matched_v2"
+    )
+    horizon_html, horizon_arms = _public_horizon_page(
+        horizon_source,
+        latest_decision,
+    )
+    horizon_bytes = horizon_html.encode("utf-8")
+    (destination / "horizons.html").write_bytes(horizon_bytes)
     (destination / ".nojekyll").write_text("", encoding="utf-8")
 
-    latest_decision = max(item["decision_date"] for item in snapshots)
     public_manifest = {
-        "schema_version": 2,
+        "schema_version": 3,
         "generated_at_utc": datetime.now(UTC).isoformat(),
         "latest_decision_date": latest_decision,
         "snapshot_count": len(snapshots),
         "evaluation_count": sum(len(item["evaluations"]) for item in snapshots),
         "dashboard_sha256": dashboard_sha256(dashboard_bytes),
         "dashboard_bytes": len(dashboard_bytes),
+        "horizon_dashboard_sha256": dashboard_sha256(horizon_bytes),
+        "horizon_dashboard_bytes": len(horizon_bytes),
+        "horizon_forecast_arms": horizon_arms,
         "source_policy": "sanitized_live_forward_outputs_only",
     }
     (data_dir / "manifest.json").write_text(
diff --git a/src/experiments/live_forward_20d.py b/src/experiments/live_forward_20d.py
index efa2b58..ca5ed15 100644
--- a/src/experiments/live_forward_20d.py
+++ b/src/experiments/live_forward_20d.py
@@ -1089,6 +1089,62 @@ def _sentiment_coverage(frame: pd.DataFrame) -> float:
     return float(numeric.abs().sum(axis=1).gt(0.0).mean())
 
 
+def build_shap_feature_summary(
+    current: pd.DataFrame,
+    model: RidgeRankModel | XGBRankModel,
+    feature_names: list[str],
+    *,
+    family: str,
+) -> pd.DataFrame:
+    """Summarize current-cohort SHAP contributions without a plotting dependency."""
+    numeric = current.reindex(columns=feature_names).apply(
+        pd.to_numeric,
+        errors="coerce",
+    )
+    if numeric.empty:
+        return pd.DataFrame()
+    if family == "ridge":
+        pipeline = model.model
+        imputed = pipeline.named_steps["imputer"].transform(numeric)
+        scaled = pipeline.named_steps["scaler"].transform(imputed)
+        background = np.asarray(scaled, dtype=float).mean(axis=0)
+        coefficients = np.asarray(
+            pipeline.named_steps["ridge"].coef_,
+            dtype=float,
+        ).reshape(-1)
+        contributions = (np.asarray(scaled) - background) * coefficients
+        method = "linear_shap_current_cohort_background"
+    elif family == "xgb_ranker":
+        import xgboost as xgb
+
+        matrix = xgb.DMatrix(numeric, feature_names=feature_names)
+        contributions = np.asarray(
+            model.model.get_booster().predict(
+                matrix,
+                pred_contribs=True,
+            ),
+            dtype=float,
+        )[:, : len(feature_names)]
+        method = "tree_shap"
+    else:
+        raise ValueError(f"Unsupported SHAP family: {family}")
+    summary = pd.DataFrame(
+        {
+            "feature": feature_names,
+            "mean_abs_shap": np.mean(np.abs(contributions), axis=0),
+            "mean_shap": np.mean(contributions, axis=0),
+            "positive_share": np.mean(contributions > 0.0, axis=0),
+            "method": method,
+            "explanation_rows": int(len(numeric)),
+        }
+    )
+    summary = summary.sort_values("mean_abs_shap", ascending=False).reset_index(
+        drop=True
+    )
+    summary["importance_rank"] = np.arange(1, len(summary) + 1)
+    return summary
+
+
 def freeze_forecast_snapshot(
     current: pd.DataFrame,
     model: RidgeRankModel | XGBRankModel,
@@ -1146,6 +1202,14 @@ def freeze_forecast_snapshot(
     output_dir.mkdir(parents=True, exist_ok=True)
     ranked.reset_index().to_parquet(output_dir / "eligible_scores.parquet", index=False)
     selected_frame.reset_index().to_csv(output_dir / "selected_holdings.csv", index=False)
+    shap_summary = build_shap_feature_summary(
+        scored,
+        model,
+        feature_names,
+        family=family,
+    )
+    if not shap_summary.empty:
+        shap_summary.to_csv(output_dir / "shap_summary.csv", index=False)
     manifest = {
         "status": "pending",
         "experiment_version": experiment_version,
@@ -1171,6 +1235,24 @@ def freeze_forecast_snapshot(
         "sentiment_coverage": _sentiment_coverage(current),
         "feature_count": int(len(feature_names)),
         "features": feature_names,
+        "explainability": {
+            "status": "available" if not shap_summary.empty else "unavailable",
+            "artifact": (
+                str(output_dir / "shap_summary.csv")
+                if not shap_summary.empty
+                else None
+            ),
+            "method": (
+                str(shap_summary["method"].iloc[0])
+                if not shap_summary.empty
+                else None
+            ),
+            "rows": (
+                int(shap_summary["explanation_rows"].iloc[0])
+                if not shap_summary.empty
+                else 0
+            ),
+        },
         "created_utc": pd.Timestamp.now(tz=UTC).isoformat(),
         "output": str(output_dir),
     }
@@ -1627,6 +1709,15 @@ def evaluate_horizon_matched_strategies(
                 min_count=1,
             )
             returns = values.pct_change(fill_method=None).iloc[1:]
+            spy_returns = pd.Series(index=returns.index, dtype=float)
+            if "SPY" in opened.columns:
+                spy_returns = (
+                    opened["SPY"]
+                    .reindex(period_dates)
+                    .pct_change(fill_method=None)
+                    .iloc[1:]
+                    .reindex(returns.index)
+                )
             for return_date, gross_return in returns.items():
                 if not np.isfinite(gross_return):
                     continue
@@ -1639,6 +1730,11 @@ def evaluate_horizon_matched_strategies(
                         "entry_date": str(price_dates[entry_pos].date()),
                         "return_date": str(pd.Timestamp(return_date).date()),
                         "gross_return": float(gross_return),
+                        "spy_return": (
+                            float(spy_returns.get(return_date))
+                            if np.isfinite(spy_returns.get(return_date, np.nan))
+                            else float("nan")
+                        ),
                         "initial_holdings": int(len(weights)),
                     }
                 )
@@ -1656,85 +1752,14 @@ def build_live_20d_dashboard(
     run_summary: dict[str, Any],
     matured: pd.DataFrame,
 ) -> Path:
-    """Write a compact deterministic dashboard for the new experiment."""
-    rebalance_policy = run_summary.get("rebalance_policy", {})
-    active_rebalance_policy = str(
-        rebalance_policy.get(
-            "active",
-            "full_top40_refresh_on_horizon_clock",
-        )
-    )
-    rebalance_policy_status = str(
-        rebalance_policy.get("status", "under_audit")
+    """Write the self-contained horizon-matched research dashboard."""
+    from .live_forward_20d_dashboard import (
+        build_horizon_dashboard_payload,
+        horizon_dashboard_html,
     )
-    latest_rows = "".join(
-        (
-            "<tr>"
-            f"<td>{item['family']}</td>"
-            f"<td>{item['horizon_sessions']}</td>"
-            f"<td>{item['eligible_names']}</td>"
-            f"<td>{item['selected_names']}</td>"
-            f"<td>{item['sentiment_coverage']:.1%}</td>"
-            f"<td>{'yes' if item['refitted_today'] else 'no'}</td>"
-            "</tr>"
-        )
-        for item in run_summary.get("forecasts", [])
-    )
-    matured_rows = ""
-    if not matured.empty:
-        for item in matured.tail(30).to_dict("records"):
-            matured_rows += (
-                "<tr>"
-                f"<td>{item['decision_date']}</td>"
-                f"<td>{item['family']}</td>"
-                f"<td>{item['horizon_sessions']}</td>"
-                f"<td>{item['selected_return']:.2%}</td>"
-                f"<td>{item['active_return_vs_universe']:.2%}</td>"
-                f"<td>{item['active_return_vs_spy']:.2%}</td>"
-                "</tr>"
-            )
-    strategy_rows = "".join(
-        (
-            "<tr>"
-            f"<td>{item['family']}</td>"
-            f"<td>{item['horizon_sessions']}</td>"
-            f"<td>{item['daily_observations']}</td>"
-            f"<td>{item['gross_cumulative_return']:.2%}</td>"
-            f"<td>{item['first_return_date']}</td>"
-            f"<td>{item['last_return_date']}</td>"
-            "</tr>"
-        )
-        for item in run_summary.get("strategy_performance", [])
-    )
-    html = f"""<!doctype html>
-<html><head><meta charset="utf-8"><title>20-session live forward</title>
-<style>
-body{{font-family:Arial,sans-serif;max-width:1180px;margin:32px auto;color:#172033}}
-h1{{margin-bottom:4px}} .muted{{color:#667085}} .card{{border:1px solid #d0d5dd;
-border-radius:10px;padding:18px;margin:18px 0}} table{{border-collapse:collapse;width:100%}}
-th,td{{padding:9px;border-bottom:1px solid #eaecf0;text-align:left}}
-.warn{{background:#fff7e8;border-color:#f5c26b}}
-</style></head><body>
-<h1>20-session live-forward research</h1>
-<p class="muted">Decision {run_summary['decision_date']} | 100% research exposure |
-horizon-matched 5/10/20-session strategy clocks</p>
-<div class="card warn"><strong>Research only.</strong> Same-close sentiment orders fail
-closed to next-open unless exact intraday data and a broker cutoff are verified.</div>
-<div class="card warn"><strong>Rebalance policy under audit.</strong>
-{active_rebalance_policy} | {rebalance_policy_status} | buffer disabled until validated.</div>
-<div class="card"><h2>Latest frozen forecasts</h2><table><thead><tr>
-<th>Model</th><th>Horizon</th><th>Eligible</th><th>Selected</th>
-<th>Sentiment coverage</th><th>Refit today</th></tr></thead>
-<tbody>{latest_rows}</tbody></table></div>
-<div class="card"><h2>Horizon-matched zero-cost strategies</h2><table><thead><tr>
-<th>Model</th><th>Horizon</th><th>Daily observations</th>
-<th>Gross cumulative return</th><th>First return</th><th>Last return</th>
-</tr></thead><tbody>{strategy_rows}</tbody></table></div>
-<div class="card"><h2>Matured forecasts</h2><table><thead><tr>
-<th>Decision</th><th>Model</th><th>Horizon</th><th>Selected return</th>
-<th>Active vs universe</th><th>Active vs SPY</th></tr></thead>
-<tbody>{matured_rows}</tbody></table></div>
-</body></html>"""
+
+    payload = build_horizon_dashboard_payload(output_root, run_summary, matured)
+    html = horizon_dashboard_html(payload)
     path = output_root / "dashboard.html"
     path.write_text(html, encoding="utf-8")
     return path
diff --git a/src/experiments/live_forward_20d_dashboard.py b/src/experiments/live_forward_20d_dashboard.py
new file mode 100644
index 0000000..ad1be17
--- /dev/null
+++ b/src/experiments/live_forward_20d_dashboard.py
@@ -0,0 +1,325 @@
+# ruff: noqa: E501
+"""Self-contained dashboard for the horizon-matched live-forward experiment."""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+import pandas as pd
+
+
+def _json_value(value: Any) -> Any:
+    if isinstance(value, dict):
+        return {str(key): _json_value(item) for key, item in value.items()}
+    if isinstance(value, (list, tuple)):
+        return [_json_value(item) for item in value]
+    if isinstance(value, (pd.Timestamp,)):
+        return value.isoformat()
+    if isinstance(value, (np.integer,)):
+        return int(value)
+    if isinstance(value, (np.floating,)):
+        return None if not np.isfinite(value) else float(value)
+    if pd.isna(value):
+        return None
+    return value
+
+
+def _best_validation(model_manifest: dict[str, Any], root: Path) -> dict[str, Any]:
+    selection = model_manifest.get("model_selection", {})
+    report = selection.get("report")
+    if not report:
+        return {}
+    path = Path(report)
+    if not path.is_absolute() and not path.exists():
+        path = root.parent.parent / path
+    if not path.exists():
+        return {}
+    table = pd.read_csv(path)
+    if table.empty or "row_type" not in table:
+        return {}
+    eligible = table.loc[table["row_type"].eq("aggregate")].copy()
+    if "selection_eligible" in eligible:
+        eligible = eligible.loc[
+            eligible["selection_eligible"].astype(str).str.lower().eq("true")
+        ]
+    metric = str(selection.get("selection_metric", "top40_raw_excess"))
+    if eligible.empty or metric not in eligible:
+        return {}
+    eligible[metric] = pd.to_numeric(eligible[metric], errors="coerce")
+    eligible = eligible.dropna(subset=[metric])
+    if eligible.empty:
+        return {}
+    row = eligible.sort_values(metric, ascending=False).iloc[0]
+    fields = [
+        "rank_ic",
+        "positive_ic_rate",
+        "top40_raw_forward_return",
+        "universe_raw_forward_return",
+        "top40_raw_excess",
+        "resolved_feature_count",
+        "trial",
+        "median_effective_n_estimators",
+    ]
+    result = {
+        field: _json_value(row.get(field))
+        for field in fields
+        if field in row and pd.notna(row.get(field))
+    }
+    result["selection_metric"] = metric
+    result["completed_trials"] = int(len(eligible))
+    return result
+
+
+def _collect_arms(output_root: Path) -> list[dict[str, Any]]:
+    arms: list[dict[str, Any]] = []
+    forecast_root = output_root / "forecasts"
+    if not forecast_root.exists():
+        return arms
+    for manifest_path in sorted(forecast_root.glob("*/*/manifest.json")):
+        forecast = json.loads(manifest_path.read_text(encoding="utf-8"))
+        family = str(forecast["family"])
+        horizon = int(forecast["horizon_sessions"])
+        decision_date = str(forecast["decision_date"])
+        refit_date = str(forecast["model_refit_date"])
+        model_path = (
+            output_root
+            / "model_versions"
+            / family
+            / f"h{horizon}"
+            / refit_date
+            / "manifest.json"
+        )
+        model_manifest = (
+            json.loads(model_path.read_text(encoding="utf-8"))
+            if model_path.exists()
+            else {}
+        )
+        selected_path = manifest_path.parent / "selected_holdings.csv"
+        shap_path = manifest_path.parent / "shap_summary.csv"
+        selected: list[dict[str, Any]] = []
+        shap_summary: list[dict[str, Any]] = []
+        if selected_path.exists():
+            holdings = pd.read_csv(selected_path)
+            public_columns = [
+                column
+                for column in ("ticker", "sector", "rank", "score", "weight")
+                if column in holdings
+            ]
+            selected = _json_value(
+                holdings[public_columns].replace({np.nan: None}).to_dict("records")
+            )
+        if shap_path.exists():
+            shap = pd.read_csv(shap_path).sort_values(
+                "mean_abs_shap",
+                ascending=False,
+            )
+            public_shap_columns = [
+                column
+                for column in (
+                    "feature",
+                    "mean_abs_shap",
+                    "mean_shap",
+                    "positive_share",
+                    "method",
+                    "explanation_rows",
+                    "importance_rank",
+                )
+                if column in shap
+            ]
+            shap_summary = _json_value(
+                shap[public_shap_columns].head(20).to_dict("records")
+            )
+        raw_explainability = forecast.get("explainability", {})
+        explainability = {
+            key: raw_explainability.get(key)
+            for key in ("status", "method", "rows")
+            if key in raw_explainability
+        }
+        if not explainability:
+            explainability = {
+                "status": "available" if shap_summary else "not_captured"
+            }
+        arms.append(
+            {
+                "id": f"{decision_date}:{family}:h{horizon}",
+                "decision_date": decision_date,
+                "family": family,
+                "horizon_sessions": horizon,
+                "status": forecast.get("status"),
+                "expected_entry_date": forecast.get("expected_entry_date"),
+                "expected_maturity_date": forecast.get("expected_maturity_date"),
+                "refit_interval_sessions": forecast.get("refit_interval_sessions"),
+                "model_refit_date": refit_date,
+                "refitted_today": bool(forecast.get("refitted_today")),
+                "strategy_rebalance": bool(forecast.get("strategy_rebalance")),
+                "strategic_rebalance": bool(forecast.get("strategic_rebalance")),
+                "eligible_names": int(forecast.get("eligible_names", 0)),
+                "selected_names": int(forecast.get("selected_names", 0)),
+                "sentiment_coverage": forecast.get("sentiment_coverage"),
+                "feature_count": int(forecast.get("feature_count", 0)),
+                "features": list(forecast.get("features", [])),
+                "shap_summary": shap_summary,
+                "explainability": explainability,
+                "selected": selected,
+                "training_start": model_manifest.get("training_start"),
+                "training_end": model_manifest.get("training_end"),
+                "training_rows": model_manifest.get("training_rows"),
+                "model_config": model_manifest.get("model_config", {}),
+                "validation": _best_validation(model_manifest, output_root),
+            }
+        )
+    return sorted(
+        arms,
+        key=lambda arm: (
+            arm["decision_date"],
+            arm["family"],
+            arm["horizon_sessions"],
+        ),
+    )
+
+
+def build_horizon_dashboard_payload(
+    output_root: str | Path,
+    run_summary: dict[str, Any],
+    matured: pd.DataFrame,
+) -> dict[str, Any]:
+    """Collect a public-safe dashboard payload from frozen v2 artifacts."""
+    root = Path(output_root)
+    strategy_path = root / "strategy_daily_returns.csv"
+    strategy = (
+        pd.read_csv(strategy_path).to_dict("records")
+        if strategy_path.exists()
+        else []
+    )
+    runs: list[dict[str, Any]] = []
+    for path in sorted((root / "forecasts").glob("*/run_summary.json")):
+        summary = json.loads(path.read_text(encoding="utf-8"))
+        runs.append(
+            {
+                "decision_date": summary.get("decision_date"),
+                "status": summary.get("status"),
+                "matured_forecasts": summary.get("matured_forecasts", 0),
+                "strategy_daily_return_rows": summary.get(
+                    "strategy_daily_return_rows", 0
+                ),
+                "tactical_orders": summary.get("tactical_orders", 0),
+                "model_failures": len(summary.get("model_failures", [])),
+            }
+        )
+    tactical_path = (
+        root
+        / "forecasts"
+        / str(run_summary["decision_date"])
+        / "tactical_orders.csv"
+    )
+    tactical = (
+        pd.read_csv(tactical_path).to_dict("records")
+        if tactical_path.exists() and tactical_path.stat().st_size > 0
+        else []
+    )
+    payload = {
+        "schema_version": 1,
+        "experiment": {
+            "id": run_summary.get("experiment_version"),
+            "status": "paper_live_research",
+            "decision_date": run_summary.get("decision_date"),
+            "primary_horizon_sessions": run_summary.get(
+                "primary_horizon_sessions"
+            ),
+            "comparison_design": run_summary.get("comparison_design"),
+            "invested_weight": run_summary.get("invested_weight"),
+        },
+        "rebalance_policy": run_summary.get("rebalance_policy", {}),
+        "refit_sessions_by_horizon": run_summary.get(
+            "refit_sessions_by_horizon", {}
+        ),
+        "input_audit": run_summary.get("input_audit", {}),
+        "universe": run_summary.get("universe", {}),
+        "survival_validation": run_summary.get("survival_validation", {}),
+        "arms": _collect_arms(root),
+        "matured": _json_value(
+            matured.replace({np.nan: None}).to_dict("records")
+            if not matured.empty
+            else []
+        ),
+        "strategy_returns": _json_value(strategy),
+        "runs": runs,
+        "tactical_orders": _json_value(tactical),
+    }
+    return _json_value(payload)
+
+
+def horizon_dashboard_html(payload: dict[str, Any]) -> str:
+    """Render the horizon-matched dashboard with no external dependencies."""
+    encoded = json.dumps(payload, ensure_ascii=False).replace("</", "<\\/")
+    return _TEMPLATE.replace("__DASHBOARD_PAYLOAD__", encoded)
+
+
+_TEMPLATE = r"""<!doctype html>
+<html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1">
+<title>Horizon-Matched Live Forward Monitor</title>
+<style>
+:root{--ink:#17172c;--muted:#68677d;--line:#dfddea;--surface:#fff;--canvas:#f6f4fa;--blue:#5c4bc4;--blue2:#8c7cdf;--teal:#0f9f8f;--teal2:#70c8bd;--amber:#d89614;--red:#c94c5b;--violet:#7559c7;--page-accent:#bca8ff;--shadow:0 14px 35px rgba(43,35,86,.09)}
+*{box-sizing:border-box}html{scroll-behavior:smooth}body{margin:0;color:var(--ink);background:var(--canvas);font-family:Inter,ui-sans-serif,-apple-system,BlinkMacSystemFont,"Segoe UI",sans-serif}
+header{position:relative;overflow:hidden;padding:42px max(28px,calc((100vw - 1440px)/2)) 82px;color:#fff;background:linear-gradient(120deg,#171329,#2b2455 66%,#493b73)}
+header:after{content:"";position:absolute;right:-90px;top:-170px;width:440px;height:440px;border:1px solid #d7cbff2b;border-radius:50%;box-shadow:0 0 0 55px #bca8ff0c,0 0 0 115px #bca8ff09}
+.eyebrow{color:#cbbcff;font-size:11px;font-weight:850;letter-spacing:.16em;text-transform:uppercase}.hero-row{display:flex;justify-content:space-between;align-items:end;gap:28px;position:relative;z-index:1}
+h1{margin:8px 0 9px;font-size:clamp(30px,3.2vw,46px);letter-spacing:-.04em}.hero-copy{max-width:780px;margin:0;color:#bdc9da;line-height:1.55}
+.hero-badge{display:grid;gap:4px;min-width:190px;padding:14px 16px;border:1px solid #ffffff24;border-radius:14px;background:#ffffff0d;backdrop-filter:blur(6px)}.hero-badge span{color:#9fb0c8;font-size:10px;text-transform:uppercase;letter-spacing:.08em}.hero-badge strong{font-size:15px}
+.page-switcher{display:flex;gap:9px;margin-top:22px;position:relative;z-index:2}.page-link{display:grid;gap:3px;min-width:210px;padding:12px 15px;border:1px solid #ffffff28;border-radius:12px;color:#d5d0e5;background:#ffffff0a;text-decoration:none;transition:.18s ease}.page-link:hover{transform:translateY(-1px);border-color:var(--page-accent);background:#ffffff12}.page-link.active{color:#fff;border-color:var(--page-accent);background:#bca8ff17;box-shadow:inset 0 -3px 0 var(--page-accent)}.page-link strong{font-size:12px}.page-link small{color:#aaa4c0;font-size:9px}.page-link.active small{color:#d7cff1}
+main{width:min(1440px,calc(100% - 40px));margin:-52px auto 0;padding-bottom:58px;position:relative}.panel{background:var(--surface);border:1px solid #d8e0eb;border-radius:18px;box-shadow:var(--shadow)}
+.toolbar{display:flex;justify-content:space-between;align-items:end;gap:18px;padding:18px 20px}.toolbar-group,.segmented{display:flex;align-items:end;gap:10px;flex-wrap:wrap}
+label{display:grid;gap:6px;color:var(--muted);font-size:10px;font-weight:800;letter-spacing:.055em;text-transform:uppercase}select,input{min-height:39px;padding:8px 11px;color:var(--ink);background:#fbfcfe;border:1px solid #cfd8e5;border-radius:9px;font:inherit;text-transform:none;letter-spacing:0}
+button{min-height:39px;padding:8px 14px;border:1px solid #cfd8e5;border-radius:9px;color:#34445a;background:#fff;cursor:pointer;font-weight:750}button:hover,button.active{color:#fff;border-color:var(--blue);background:var(--blue)}
+.status{display:flex;align-items:center;gap:8px;color:var(--muted);font-size:12px;white-space:nowrap}.status:before{content:"";width:8px;height:8px;border-radius:50%;background:#24b47e;box-shadow:0 0 0 4px #24b47e1f}
+.notice{display:flex;justify-content:space-between;align-items:center;gap:18px;margin-top:16px;padding:15px 18px;border-color:#b9ddd7;background:linear-gradient(105deg,#f3fffc,#fff)}.notice-main{display:flex;align-items:center;gap:12px}.notice-icon{display:grid;place-items:center;flex:0 0 38px;height:38px;border-radius:12px;color:#fff;background:var(--teal);font-weight:900}.notice strong{display:block;font-size:13px}.notice small{display:block;margin-top:4px;color:var(--muted);line-height:1.45}.policy{padding:8px 11px;border:1px solid #f0d49c;border-radius:9px;color:#8b5a05;background:#fff8e8;font-size:10px;font-weight:800;text-transform:uppercase;letter-spacing:.04em}
+.kpi-grid{display:grid;grid-template-columns:repeat(4,minmax(155px,1fr));gap:14px;margin:16px 0}.kpi{position:relative;overflow:hidden;padding:17px 18px;border-top:4px solid var(--signal,var(--line))}.kpi span{color:var(--muted);font-size:11px;font-weight:700}.kpi strong{display:block;margin-top:9px;font-size:27px;letter-spacing:-.035em}.kpi small{display:block;margin-top:5px;color:var(--muted);font-size:10px;line-height:1.4}
+.section{margin-top:16px;padding:22px}.section-head{display:flex;justify-content:space-between;align-items:start;gap:18px;margin-bottom:17px}.section-head h2{margin:0;font-size:18px;letter-spacing:-.018em}.section-head p{max-width:780px;margin:5px 0 0;color:var(--muted);font-size:11px;line-height:1.5}
+.legend{display:flex;gap:7px;flex-wrap:wrap;justify-content:flex-end}.legend button{display:inline-flex;align-items:center;gap:7px;min-height:31px;padding:6px 10px;border-radius:999px;font-size:10px}.legend button i{width:16px;height:3px;border-radius:999px;background:var(--series)}.legend button.off{opacity:.43;text-decoration:line-through}
+.chart svg{display:block;width:100%;height:auto;min-height:330px}.empty{display:grid;place-items:center;min-height:210px;padding:36px;color:var(--muted);text-align:center}.empty strong{display:block;color:var(--ink);font-size:15px}.empty span{display:block;max-width:520px;margin-top:7px;font-size:11px;line-height:1.55}
+.progress-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:9px;margin-bottom:15px}.progress-card{padding:12px;border:1px solid var(--line);border-radius:11px;background:#fbfcfe}.progress-head{display:flex;justify-content:space-between;gap:10px;align-items:center}.progress-head strong{font-size:11px}.progress-head span{color:var(--muted);font-size:9px}.progress-track{height:6px;margin:9px 0 7px;overflow:hidden;border-radius:999px;background:#e7ecf3}.progress-track i{display:block;height:100%;border-radius:inherit;background:var(--series)}.progress-card small{color:var(--muted);font-size:9px;line-height:1.4}.interim-note{margin-top:10px;padding:9px 11px;border-left:3px solid var(--blue);border-radius:7px;color:#536176;background:#f2f5ff;font-size:9px;line-height:1.45}
+.horizon-grid{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:13px}.horizon-card{padding:18px;border:1px solid var(--line);border-radius:14px;background:linear-gradient(180deg,#fff,#fbfcfe)}.horizon-title{display:flex;justify-content:space-between;align-items:center}.horizon-title h3{margin:0;font-size:16px}.clock{padding:5px 8px;border-radius:999px;color:#325278;background:#edf3ff;font-size:9px;font-weight:850;text-transform:uppercase}.pair{display:grid;grid-template-columns:1fr 1fr;gap:8px;margin-top:14px}.model-slice{padding:12px;border-left:4px solid var(--series);border-radius:10px;background:#f4f7fa}.model-slice span{display:block;color:var(--muted);font-size:10px}.model-slice strong{display:block;margin-top:6px;font-size:20px}.model-slice small{display:block;margin-top:5px;color:var(--muted);font-size:9px;line-height:1.45}.maturity{margin-top:12px;padding-top:11px;border-top:1px solid var(--line);color:var(--muted);font-size:10px}
+.explorer-controls{display:flex;gap:10px;align-items:end;flex-wrap:wrap}.search{min-width:220px}.holding-summary{display:grid;grid-template-columns:repeat(4,1fr);gap:9px;margin-bottom:12px}.mini{padding:12px;border:1px solid var(--line);border-radius:11px;background:#fbfcfe}.mini span{display:block;color:var(--muted);font-size:9px}.mini strong{display:block;margin-top:5px;font-size:16px}
+.table-wrap{overflow:auto;border:1px solid var(--line);border-radius:12px;max-height:560px}table{width:100%;border-collapse:collapse;background:#fff}th,td{padding:11px 13px;border-bottom:1px solid #edf1f5;font-size:11px;text-align:right;white-space:nowrap}th:first-child,td:first-child{text-align:left}th{position:sticky;top:0;z-index:1;color:#526075;background:#f7f9fc;font-size:9px;letter-spacing:.055em;text-transform:uppercase}tr:last-child td{border-bottom:0}tbody tr:hover{background:#f8faff}.sort{min-height:0;padding:0;border:0;border-radius:0;color:inherit;background:transparent;font-size:inherit;text-transform:inherit}.sort:hover{color:var(--blue);background:transparent}
+.model-grid{display:grid;grid-template-columns:repeat(2,minmax(0,1fr));gap:13px}.model-card{padding:18px;border:1px solid var(--line);border-radius:14px;background:#fff}.model-title{display:flex;justify-content:space-between;gap:12px;align-items:center}.model-title h3{margin:0;font-size:15px}.badge{padding:5px 8px;border-radius:999px;color:#0a7467;background:#e7f7f4;font-size:9px;font-weight:850;text-transform:uppercase}.metrics{display:grid;grid-template-columns:repeat(3,1fr);gap:7px;margin:13px 0}.metric{padding:9px;border-radius:9px;background:#f3f6fa}.metric span{display:block;color:var(--muted);font-size:8px}.metric strong{display:block;margin-top:5px;font-size:13px}.features{display:flex;gap:5px;flex-wrap:wrap;margin-top:10px}.feature{padding:4px 7px;border-radius:999px;color:#536176;background:#eef2f7;font-size:8px}.model-meta{color:var(--muted);font-size:9px;line-height:1.55}.explain-toggle{width:100%;margin-top:13px;color:#3156d3;background:#f5f7ff;border-color:#d9e0fa;font-size:10px}.explain-toggle:hover{color:#fff}.explain-panel{margin-top:10px;padding:12px;border:1px solid var(--line);border-radius:10px;background:#fbfcfe}.explain-panel[hidden]{display:none}.shap-row{display:grid;grid-template-columns:minmax(95px,1.3fr) 2fr 58px;gap:8px;align-items:center;margin:7px 0}.shap-row span{overflow:hidden;color:#526075;font-size:8px;text-overflow:ellipsis;white-space:nowrap}.shap-bar{height:8px;overflow:hidden;border-radius:999px;background:#e7ecf3}.shap-bar i{display:block;height:100%;border-radius:inherit;background:linear-gradient(90deg,#3156d3,#0f9f8f)}.shap-row strong{text-align:right;font-size:8px}.shap-empty{color:var(--muted);font-size:9px;line-height:1.55}
+.health-grid{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:9px}.health{padding:14px;border:1px solid var(--line);border-radius:11px;background:#fbfcfe}.health-head{display:flex;justify-content:space-between;align-items:center}.health-head span{font-size:10px;font-weight:800}.dot{width:8px;height:8px;border-radius:50%;background:#24b47e}.dot.warn{background:var(--amber)}.health strong{display:block;margin-top:9px;font-size:17px}.health small{display:block;margin-top:5px;color:var(--muted);font-size:9px;line-height:1.45}.warning-list{display:grid;gap:7px;margin-top:12px}.warning{padding:10px 12px;border-left:4px solid var(--amber);border-radius:8px;color:#665126;background:#fff8e8;font-size:10px;line-height:1.45}
+.pos{color:#078071}.neg{color:var(--red)}.muted{color:var(--muted)}
+@media(max-width:1050px){.horizon-grid{grid-template-columns:1fr}.model-grid{grid-template-columns:1fr}.health-grid,.holding-summary{grid-template-columns:repeat(2,1fr)}.hero-row{align-items:start;flex-direction:column}.hero-badge{min-width:0}}
+@media(max-width:700px){main{width:calc(100% - 20px)}header{padding-left:20px;padding-right:20px}.toolbar,.notice,.section-head{align-items:stretch;flex-direction:column}.kpi-grid,.health-grid,.holding-summary,.pair,.metrics,.progress-grid{grid-template-columns:1fr}.search{min-width:0;width:100%}.hero-badge{display:none}.page-switcher{display:grid}.page-link{min-width:0}.shap-row{grid-template-columns:92px 1fr 48px}}
+</style></head><body>
+<header><div class="hero-row"><div><div class="eyebrow">AI Portfolio Research</div><h1>Horizon Comparison</h1><p class="hero-copy">Six independently tuned portfolios tracked on their natural 5, 10, and 20-session clocks. Interim marks, matured outcomes, and validation evidence remain explicitly separated.</p><nav class="page-switcher" aria-label="Monitor pages"><a class="page-link" href="../live_forward/dashboard.html"><strong>Live Forward Monitor</strong><small>Daily evidence and shared stock-level view</small></a><a class="page-link active" href="dashboard.html" aria-current="page"><strong>Horizon Comparison</strong><small>5 / 10 / 20-session experiment</small></a></nav></div><div class="hero-badge"><span>Evidence status</span><strong id="heroStatus">Paper-live · collecting</strong><span id="heroDate"></span></div></div></header>
+<main>
+<div class="toolbar panel"><div class="toolbar-group"><label>Decision date<select id="decisionSelect"></select></label><label>Family<select id="familyFilter"><option value="all">All families</option><option value="ridge">Ridge</option><option value="xgb_ranker">XGBoost</option></select></label><label>Horizon<select id="horizonFilter"><option value="all">All horizons</option><option value="5">5 sessions</option><option value="10">10 sessions</option><option value="20">20 sessions</option></select></label></div><div class="toolbar-group"><button id="refreshData">Refresh data</button><div class="status" id="dataStatus">Frozen evidence loaded</div></div></div>
+<section class="notice panel"><div class="notice-main"><div class="notice-icon">→</div><div><strong id="noticeTitle"></strong><small id="noticeCopy"></small></div></div><div class="policy" id="policyBadge"></div></section>
+<div class="kpi-grid" id="kpis"></div>
+<section class="section panel"><div class="section-head"><div><h2>Prospective strategy evidence</h2><p>Partial mark-to-market outcomes appear as soon as an entered portfolio has a new observable open. They remain interim until the complete 5, 10, or 20-session window matures.</p></div><div class="legend" id="strategyLegend"></div></div><div class="progress-grid" id="strategyProgress"></div><div class="chart" id="strategyChart"></div><div class="interim-note">The decision session is not a holding session. A July 29 close signal enters at the July 30 open, so July 29 must remain 0/5 rather than borrowing a pre-entry return. Interim marks answer “what has happened so far?” and are never mixed with validation estimates.</div></section>
+<section class="section panel"><div class="section-head"><div><h2>Horizon comparison</h2><p>Paired Ridge and XGBoost development evidence with each arm’s own purge, feature subset, refit interval, and maturity date.</p></div></div><div class="horizon-grid" id="horizonGrid"></div></section>
+<section class="section panel"><div class="section-head"><div><h2>Selected model specifications</h2><p>The winning development configuration for each visible arm. Open a model’s explainability panel to inspect genuine SHAP importance when that frozen vintage captured it.</p></div></div><div class="model-grid" id="modelGrid"></div></section>
+<section class="section panel"><div class="section-head"><div><h2>Pipeline and risk controls</h2><p>Freshness, universe, NLP, tactical routing, survival guardrails, and artifact integrity for the selected decision date.</p></div></div><div class="health-grid" id="healthGrid"></div><div class="warning-list" id="warnings"></div></section>
+<section class="section panel"><div class="section-head"><div><h2>Matured forecast history</h2><p>Prospective outcomes appear only after their full label interval matures. Pending forecasts are not scored as zero.</p></div></div><div id="maturedTable"></div></section>
+</main>
+<script id="dashboard-data" type="application/json">__DASHBOARD_PAYLOAD__</script>
+<script>
+const data=JSON.parse(document.getElementById('dashboard-data').textContent),decisionSelect=document.getElementById('decisionSelect'),familyFilter=document.getElementById('familyFilter'),horizonFilter=document.getElementById('horizonFilter'),STATE_KEY='live-forward-horizon-dashboard-v1';
+const COLORS={ridge:'#3156d3',xgb_ranker:'#0f9f8f',ridge_h5:'#7290eb',ridge_h10:'#4f70dc',ridge_h20:'#3156d3',xgb_ranker_h5:'#70c8bd',xgb_ranker_h10:'#3eb2a4',xgb_ranker_h20:'#0f9f8f',spy:'#d89614'},names={ridge:'Ridge',xgb_ranker:'XGBoost'},fmtPct=v=>v==null||!Number.isFinite(Number(v))?'n/a':(Number(v)*100).toFixed(2)+'%',fmtNum=(v,d=3)=>v==null||!Number.isFinite(Number(v))?'n/a':Number(v).toFixed(d),human=v=>String(v??'').replaceAll('_',' ').replace(/\b\w/g,c=>c.toUpperCase()),tone=v=>Number(v)>=0?'pos':'neg',armColor=a=>COLORS[`${a.family}_h${a.horizon_sessions}`]||COLORS[a.family],armLabel=a=>`${names[a.family]||human(a.family)} · ${a.horizon_sessions} sessions`;
+let saved={};try{saved=JSON.parse(localStorage.getItem(STATE_KEY)||'{}')}catch{}let hiddenSeries=new Set(saved.hiddenSeries||[]);
+function save(){try{localStorage.setItem(STATE_KEY,JSON.stringify({decisionDate:decisionSelect.value,family:familyFilter.value,horizon:horizonFilter.value,hiddenSeries:[...hiddenSeries]}))}catch{}}
+function option(value,label){const node=document.createElement('option');node.value=value;node.textContent=label;return node}
+const dates=[...new Set(data.arms.map(a=>a.decision_date))].sort().reverse();dates.forEach(d=>decisionSelect.appendChild(option(d,d)));decisionSelect.value=dates.includes(saved.decisionDate)?saved.decisionDate:dates[0]||'';familyFilter.value=['ridge','xgb_ranker'].includes(saved.family)?saved.family:'all';horizonFilter.value=['5','10','20'].includes(saved.horizon)?saved.horizon:'all';
+function decisionArms(){return data.arms.filter(a=>a.decision_date===decisionSelect.value)}
+function visibleArms(){return decisionArms().filter(a=>(familyFilter.value==='all'||a.family===familyFilter.value)&&(horizonFilter.value==='all'||String(a.horizon_sessions)===horizonFilter.value))}
+function table(rows,columns,sortState){if(!rows.length)return'<div class="empty"><div><strong>No matching records</strong><span>Adjust the filters or wait for the required outcome window.</span></div></div>';const head=columns.map(([key,label])=>`<th><button class="sort" data-sort="${key}">${label}${sortState?.key===key?(sortState.direction==='asc'?' ↑':' ↓'):''}</button></th>`).join(''),body=rows.map(row=>`<tr>${columns.map(([key])=>{const value=row[key],shown=key.includes('return')||key==='weight'||key==='sentiment_coverage'?fmtPct(value):key==='score'?fmtNum(value,5):value??'—';return`<td class="${key.includes('return')&&value!=null?tone(value):''}">${shown}</td>`}).join('')}</tr>`).join('');return`<div class="table-wrap"><table><thead><tr>${head}</tr></thead><tbody>${body}</tbody></table></div>`}
+function renderNotice(){const arms=decisionArms(),pending=arms.filter(a=>a.status==='pending').length,next=[...new Set(arms.map(a=>a.expected_entry_date).filter(Boolean))].sort()[0],policy=data.rebalance_policy||{};document.getElementById('noticeTitle').textContent=`${pending} frozen forecasts ready for the next session`;document.getElementById('noticeCopy').textContent=`Decision ${decisionSelect.value} · expected entry ${next||'pending'} · partial marks update session by session; final labels wait for full maturity.`;document.getElementById('policyBadge').textContent=`${human(policy.status||'under audit')} · buffer ${policy.buffer_enabled?'on':'off'}`;document.getElementById('heroDate').textContent=`Latest decision · ${dates[0]||'n/a'}`}
+function renderKpis(){const arms=visibleArms(),matured=data.matured.filter(x=>!decisionSelect.value||x.decision_date<=decisionSelect.value),audit=data.input_audit||{},universe=data.universe||{},items=[['Active arms',arms.length,'#3156d3',`${decisionArms().length} total for this decision`],['Matured forecasts',matured.length,'#7559c7',matured.length?'Prospective outcomes available':'First outcomes still pending'],['Eligible universe',universe.final_names,'#0f9f8f',`${universe.current_liquid_names??0} liquid names`],['Sentiment coverage',arms[0]?.sentiment_coverage,'#d89614',`${audit.processed_news_rows??0} recent articles processed`]];document.getElementById('kpis').innerHTML=items.map(([label,value,color,detail],i)=>`<article class="kpi panel" style="--signal:${color}"><span>${label}</span><strong>${i===3?fmtPct(value):value??'n/a'}</strong><small>${detail}</small></article>`).join('');document.getElementById('dataStatus').textContent=`${dates.length} decision date${dates.length===1?'':'s'} · ${data.arms.length} frozen forecasts`}
+function strategyGroups(){const by=new Map,spyByDate=new Map;data.strategy_returns.forEach(row=>{const id=`${row.family}_h${row.horizon_sessions}`;if(!by.has(id))by.set(id,[]);by.get(id).push(row);if(Number.isFinite(Number(row.spy_return))&&!spyByDate.has(row.return_date))spyByDate.set(row.return_date,{return_date:row.return_date,gross_return:Number(row.spy_return)})});if(spyByDate.size)by.set('spy',[...spyByDate.values()]);return [...by.entries()].map(([id,rows])=>({id,rows:rows.sort((a,b)=>String(a.return_date).localeCompare(String(b.return_date)))}))}
+function renderStrategy(){const arms=visibleArms(),visibleIds=new Set([...arms.map(a=>`${a.family}_h${a.horizon_sessions}`),'spy']),groups=strategyGroups().filter(g=>visibleIds.has(g.id)),legend=document.getElementById('strategyLegend');document.getElementById('strategyProgress').innerHTML=arms.map(a=>{const rows=data.strategy_returns.filter(r=>r.family===a.family&&Number(r.horizon_sessions)===a.horizon_sessions&&r.rebalance_decision_date===a.decision_date).sort((x,y)=>String(x.return_date).localeCompare(String(y.return_date))),observed=rows.length,cumulative=rows.reduce((value,row)=>value*(1+Number(row.gross_return)),1)-1,spyCumulative=rows.reduce((value,row)=>Number.isFinite(Number(row.spy_return))?value*(1+Number(row.spy_return)):value,1)-1,asOf=data.experiment.decision_date,status=observed?`${fmtPct(cumulative)} interim · ${fmtPct(cumulative-spyCumulative)} vs SPY`:(asOf<a.expected_entry_date?`Awaiting ${a.expected_entry_date} open`:'Entered · first mark pending');return`<article class="progress-card"><div class="progress-head"><strong>${armLabel(a)}</strong><span>${observed}/${a.horizon_sessions} sessions held</span></div><div class="progress-track"><i style="width:${Math.min(100,observed/a.horizon_sessions*100)}%;--series:${armColor(a)}"></i></div><small>${status} · complete label ${a.expected_maturity_date}</small></article>`}).join('');legend.innerHTML=groups.map(g=>`<button class="${hiddenSeries.has(g.id)?'off':'active'}" data-series="${g.id}" style="--series:${COLORS[g.id]||'#7559c7'}"><i></i>${g.id==='spy'?'SPY':human(g.id.replace('_h',' · h'))}</button>`).join('');legend.querySelectorAll('[data-series]').forEach(b=>b.addEventListener('click',()=>{hiddenSeries.has(b.dataset.series)?hiddenSeries.delete(b.dataset.series):hiddenSeries.add(b.dataset.series);renderStrategy();save()}));const visible=groups.filter(g=>!hiddenSeries.has(g.id));if(!visible.length){const entries=arms.map(a=>a.expected_entry_date).filter(Boolean).sort();document.getElementById('strategyChart').innerHTML=`<div class="empty"><div><strong>No observable post-entry mark yet</strong><span>The ${decisionSelect.value} close decision first becomes invested at the ${entries[0]||'next configured'} open. The chart begins with the first post-entry open-to-open return and does not wait for full maturity.</span></div></div>`;return}const lines=visible.map(g=>{let cumulative=1;return{id:g.id,points:g.rows.map(r=>({date:r.return_date,value:(cumulative*=1+Number(r.gross_return))-1}))}}),stamps=[...new Set(lines.flatMap(l=>l.points.map(p=>p.date)))].sort(),vals=lines.flatMap(l=>l.points.map(p=>p.value)),min=Math.min(0,...vals),max=Math.max(0,...vals),pad=Math.max((max-min)*.14,.001),w=1180,h=360,left=66,right=22,top=18,bottom=48,pw=w-left-right,ph=h-top-bottom,x=s=>left+stamps.indexOf(s)/Math.max(1,stamps.length-1)*pw,y=v=>top+(1-(v-(min-pad))/((max+pad)-(min-pad)||1))*ph,ticks=[0,.25,.5,.75,1].map(f=>max+pad-f*((max+pad)-(min-pad))),grid=ticks.map(v=>`<line x1="${left}" x2="${left+pw}" y1="${y(v)}" y2="${y(v)}" stroke="#e6ebf2"/><text x="${left-9}" y="${y(v)+4}" text-anchor="end" fill="#69768a" font-size="10">${fmtPct(v)}</text>`).join(''),paths=lines.map(line=>`<path d="${line.points.map((p,i)=>`${i?'L':'M'}${x(p.date)},${y(p.value)}`).join(' ')}" fill="none" stroke="${COLORS[line.id]}" stroke-width="${line.id==='spy'?2:3}" stroke-dasharray="${line.id==='spy'?'6 5':''}" stroke-linecap="round"/>`).join('');document.getElementById('strategyChart').innerHTML=`<svg viewBox="0 0 ${w} ${h}">${grid}<line x1="${left}" x2="${left+pw}" y1="${y(0)}" y2="${y(0)}" stroke="#9aa7b8" stroke-dasharray="5 5"/>${paths}</svg>`}
+function renderHorizons(){const all=decisionArms(),cards=[5,10,20].map(h=>{const pair=all.filter(a=>a.horizon_sessions===h).sort((a,b)=>a.family.localeCompare(b.family)),slices=pair.map(a=>`<div class="model-slice" style="--series:${armColor(a)}"><span>${names[a.family]}</span><strong class="${tone(a.validation?.top40_raw_excess)}">${fmtPct(a.validation?.top40_raw_excess)}</strong><small>Validation top-40 excess<br>Rank IC ${fmtNum(a.validation?.rank_ic)} · ${a.feature_count} features</small></div>`).join(''),maturity=pair[0]?.expected_maturity_date||'pending';return`<article class="horizon-card"><div class="horizon-title"><h3>${h}-session horizon</h3><span class="clock">Refit every ${h}</span></div><div class="pair">${slices}</div><div class="maturity">Expected first maturity · <strong>${maturity}</strong></div></article>`});document.getElementById('horizonGrid').innerHTML=cards.join('')}
+function renderModels(){const node=document.getElementById('modelGrid');node.innerHTML=visibleArms().map(a=>{const v=a.validation||{},features=a.features.slice(0,12).map(f=>`<span class="feature">${human(f)}</span>`).join(''),extra=Math.max(0,a.features.length-12),shap=a.shap_summary||[],max=Math.max(...shap.map(x=>Number(x.mean_abs_shap)||0),0),chart=shap.length?`<div class="model-meta">${human(shap[0].method)} · ${shap[0].explanation_rows} current-cohort rows · mean absolute SHAP</div>${shap.slice(0,12).map(x=>`<div class="shap-row"><span title="${human(x.feature)}">${human(x.feature)}</span><div class="shap-bar"><i style="width:${max?Number(x.mean_abs_shap)/max*100:0}%"></i></div><strong>${fmtNum(x.mean_abs_shap,4)}</strong></div>`).join('')}`:`<div class="shap-empty"><strong>SHAP was not captured for this frozen vintage.</strong><br>The July 29 artifacts remain immutable. Genuine ${a.family==='ridge'?'linear SHAP':'TreeSHAP'} summaries will be saved with each new forecast; coefficients or XGBoost gain are not substituted or mislabeled as SHAP.</div>`;return`<article class="model-card"><div class="model-title"><h3>${armLabel(a)}</h3><span class="badge">${a.refitted_today?'Refit today':'Frozen model'}</span></div><div class="model-meta">Training ${a.training_start||'n/a'} to ${a.training_end||'n/a'} · ${Number(a.training_rows||0).toLocaleString()} rows · ${v.completed_trials||'n/a'} eligible trials</div><div class="metrics"><div class="metric"><span>Top-40 excess</span><strong class="${tone(v.top40_raw_excess)}">${fmtPct(v.top40_raw_excess)}</strong></div><div class="metric"><span>Rank IC</span><strong>${fmtNum(v.rank_ic)}</strong></div><div class="metric"><span>Positive IC</span><strong>${fmtPct(v.positive_ic_rate)}</strong></div></div><div class="model-meta">Selected ${a.feature_count} features · ${human(v.selection_metric||'n/a')}</div><div class="features">${features}${extra?`<span class="feature">+${extra} more</span>`:''}</div><button class="explain-toggle" type="button" data-explain="${a.id}" aria-expanded="false">Show SHAP feature importance</button><div class="explain-panel" data-explain-panel="${a.id}" hidden>${chart}</div></article>`}).join('')||'<div class="empty"><div><strong>No visible model specifications</strong><span>Change the family or horizon filter.</span></div></div>';node.querySelectorAll('[data-explain]').forEach(button=>button.addEventListener('click',()=>{const panel=node.querySelector(`[data-explain-panel="${button.dataset.explain}"]`),opening=panel.hidden;panel.hidden=!opening;button.setAttribute('aria-expanded',String(opening));button.textContent=opening?'Hide SHAP feature importance':'Show SHAP feature importance'}))}
+function renderHealth(){const audit=data.input_audit||{},universe=data.universe||{},survival=data.survival_validation||{},tactical=data.tactical_orders||[],cards=[['Market data',audit.market_end===data.experiment.decision_date?'Current':'Review',`${audit.downloaded_stock_tickers??0}/${audit.requested_stock_tickers??0} histories · through ${audit.market_end||'n/a'}`,audit.market_end===data.experiment.decision_date],['News + NLP',audit.finbert_fallback_permitted===false?'Strict FinBERT':'Review',`${audit.processed_news_rows??0} articles · ${audit.nlp_feature_tickers??0} tickers`,audit.finbert_fallback_permitted===false],['Universe',`${universe.final_names??0} eligible`,`Hard exclusions ${universe.hard_exclusions??0} · ML exclusions ${universe.ml_exclusions??0}`,Number(universe.final_names)>=40],['Tactical routing',`${tactical.length} orders`,tactical.length?'Review routed orders':'No event passed every frozen gate',true],['Survival ML',survival.activated?'Active':'Shadow only',survival.activated?'Activation gates passed':human(survival.reason||'not activated'),!survival.activated],['Artifacts',`${decisionArms().length} forecasts`,`${decisionArms().filter(a=>a.status==='pending').length} pending · immutable manifests`,decisionArms().length===6],['Sentiment freshness',audit.latest_news_age_hours!=null?`${Number(audit.latest_news_age_hours).toFixed(2)} hours`:'n/a',`Latest publication ${audit.latest_news_publication||'n/a'}`,Number(audit.latest_news_age_hours)<=72],['Failures',`${(data.runs.at(-1)?.model_failures)||0} model failures`,`Latest cycle ${data.runs.at(-1)?.status||'n/a'}`,!(data.runs.at(-1)?.model_failures)]],node=document.getElementById('healthGrid');node.innerHTML=cards.map(([label,value,detail,ok])=>`<article class="health"><div class="health-head"><span>${label}</span><i class="dot ${ok?'':'warn'}"></i></div><strong>${value}</strong><small>${detail}</small></article>`).join('');const warnings=[];if(!survival.activated)warnings.push('Survival ML remains shadow-only; no stock is excluded by an unvalidated threshold.');if(data.rebalance_policy?.status==='under_audit')warnings.push('Full top-40 refresh remains under audit; the rank/score buffer has not been activated.');if(!data.matured.length)warnings.push('No v2 forecast has matured. Today’s successful run is operational evidence, not realized alpha.');document.getElementById('warnings').innerHTML=warnings.map(x=>`<div class="warning">${x}</div>`).join('')}
+function renderMatured(){const rows=data.matured.filter(x=>(familyFilter.value==='all'||x.family===familyFilter.value)&&(horizonFilter.value==='all'||String(x.horizon_sessions)===horizonFilter.value));document.getElementById('maturedTable').innerHTML=table(rows,[['decision_date','Decision'],['family','Model'],['horizon_sessions','Horizon'],['selected_return','Selected return'],['active_return_vs_universe','Excess vs universe'],['active_return_vs_spy','Excess vs SPY']],null)}
+function renderAll(){renderNotice();renderKpis();renderStrategy();renderHorizons();renderModels();renderHealth();renderMatured();save()}
+[decisionSelect,familyFilter,horizonFilter].forEach(node=>node.addEventListener('change',renderAll));document.getElementById('refreshData').addEventListener('click',()=>{save();location.reload()});renderAll();
+</script></body></html>"""
diff --git a/src/experiments/public_dashboard_validation.py b/src/experiments/public_dashboard_validation.py
index fe9264d..1f08834 100644
--- a/src/experiments/public_dashboard_validation.py
+++ b/src/experiments/public_dashboard_validation.py
@@ -18,6 +18,7 @@ def validate_public_dashboard_bundle(output_dir: str | Path) -> dict[str, Any]:
     root = Path(output_dir)
     required = [
         root / "index.html",
+        root / "horizons.html",
         root / "assets/dashboard.css",
         root / "assets/dashboard.js",
         root / "assets/loader.js",
@@ -86,6 +87,11 @@ def validate_public_dashboard_bundle(output_dir: str | Path) -> dict[str, Any]:
     manifest = json.loads((root / "data/manifest.json").read_text(encoding="utf-8"))
     if manifest.get("dashboard_sha256") != dashboard_sha256(dashboard_bytes):
         raise ValueError("Public dashboard hash does not match its deployment manifest")
+    horizon_bytes = (root / "horizons.html").read_bytes()
+    if manifest.get("horizon_dashboard_sha256") != dashboard_sha256(horizon_bytes):
+        raise ValueError(
+            "Public horizon dashboard hash does not match its deployment manifest"
+        )
     return {
         "output_dir": str(root),
         "snapshots": len(snapshots),
@@ -94,6 +100,10 @@ def validate_public_dashboard_bundle(output_dir: str | Path) -> dict[str, Any]:
         "policies": len(cohort_monitor.get("policies", [])),
         "dashboard_bytes": len(dashboard_bytes),
         "dashboard_sha256": manifest["dashboard_sha256"],
+        "horizon_forecast_arms": int(
+            manifest.get("horizon_forecast_arms", 0)
+        ),
+        "horizon_dashboard_bytes": len(horizon_bytes),
     }
 
 
diff --git a/tests/test_live_dashboard.py b/tests/test_live_dashboard.py
index 5c7cdcb..56d2514 100644
--- a/tests/test_live_dashboard.py
+++ b/tests/test_live_dashboard.py
@@ -35,6 +35,9 @@ def test_dashboard_uses_date_ranges_and_overlays_models() -> None:
     assert "SPY · always shown" in html
     assert "Keep one stock list" in html
     assert "Rebalance horizon" in html
+    assert 'class="page-link active" href="dashboard.html"' in html
+    assert "../live_forward_horizon_matched_v2/dashboard.html" in html
+    assert 'id="stock-level"' in html
 
 
 def test_dashboard_draws_visible_session_anchors_and_selected_model_cards() -> None:
diff --git a/tests/test_live_dashboard_publish.py b/tests/test_live_dashboard_publish.py
index 357425c..73d3a53 100644
--- a/tests/test_live_dashboard_publish.py
+++ b/tests/test_live_dashboard_publish.py
@@ -150,10 +150,15 @@ def test_publish_site_splits_assets_and_passes_validation(
     result = publish_live_dashboard(tmp_path / "private", site)
 
     index = (site / "index.html").read_text(encoding="utf-8")
+    horizons = (site / "horizons.html").read_text(encoding="utf-8")
     assert '<link rel="stylesheet" href="./assets/dashboard.css">' in index
     assert '<script src="./assets/loader.js" defer></script>' in index
+    assert 'href="./horizons.html"' in index
+    assert 'href="./index.html"' in horizons
+    assert "Horizon Comparison" in horizons
     assert "ridge:2026-07-22" not in index
     assert result["snapshots"] == 1
+    assert result["horizon_forecast_arms"] == 0
     assert validate_public_dashboard_bundle(site)["evaluations"] == 1
 
 
diff --git a/tests/test_live_forward_20d.py b/tests/test_live_forward_20d.py
index 37498f5..319c79c 100644
--- a/tests/test_live_forward_20d.py
+++ b/tests/test_live_forward_20d.py
@@ -12,6 +12,7 @@
     _refit_sessions_for_horizon,
     build_exit_labels_fast,
     build_liquid_eligibility,
+    build_shap_feature_summary,
     build_tactical_sentiment_orders,
     build_timestamp_safe_nlp_features,
     choose_high_recall_threshold,
@@ -27,6 +28,38 @@
 from src.utils.schemas import TextPost
 
 
+def test_ridge_shap_summary_uses_current_cohort_background() -> None:
+    dates = pd.DatetimeIndex(["2026-07-29"])
+    index = pd.MultiIndex.from_product(
+        [dates, ["A", "B", "C"]],
+        names=["date", "ticker"],
+    )
+    current = pd.DataFrame(
+        {
+            "momentum": [-1.0, 0.0, 1.0],
+            "sentiment": [0.5, -0.5, 1.5],
+        },
+        index=index,
+    )
+    target = pd.Series([-0.4, 0.0, 0.7], index=index)
+    model = RidgeRankModel(alpha=1.0).fit(current, target)
+
+    summary = build_shap_feature_summary(
+        current,
+        model,
+        ["momentum", "sentiment"],
+        family="ridge",
+    )
+
+    assert set(summary["feature"]) == {"momentum", "sentiment"}
+    assert summary["mean_abs_shap"].gt(0.0).all()
+    assert summary["method"].eq(
+        "linear_shap_current_cohort_background"
+    ).all()
+    assert summary["explanation_rows"].eq(3).all()
+    assert summary["importance_rank"].tolist() == [1, 2]
+
+
 def test_fast_exit_labels_match_reference_implementation() -> None:
     dates = pd.date_range("2023-01-02", periods=8, freq="B")
     index = pd.MultiIndex.from_product(
@@ -495,6 +528,10 @@ def test_end_to_end_cycle_freezes_and_matures_forecast(
     assert first["forecasts"][0]["strategic_rebalance"]
     assert first["forecasts"][0]["strategy_rebalance"]
     assert first["forecasts"][0]["refit_interval_sessions"] == 5
+    assert first["forecasts"][0]["explainability"]["status"] == "available"
+    assert (
+        first_manifest.parent / "shap_summary.csv"
+    ).exists()
     assert replay["forecasts"][0]["reused_existing_snapshot"]
     assert first_manifest.read_bytes() == frozen_bytes
     assert second["forecasts"][0]["strategic_rebalance"]
@@ -509,4 +546,6 @@ def test_end_to_end_cycle_freezes_and_matures_forecast(
     assert third["strategy_daily_return_rows"] >= 1
     assert (tmp_path / "matured_forecasts.csv").exists()
     assert (tmp_path / "strategy_daily_returns.csv").exists()
+    strategy_returns = pd.read_csv(tmp_path / "strategy_daily_returns.csv")
+    assert strategy_returns["spy_return"].notna().all()
     assert (tmp_path / "dashboard.html").exists()
diff --git a/tests/test_live_forward_20d_dashboard.py b/tests/test_live_forward_20d_dashboard.py
new file mode 100644
index 0000000..b864649
--- /dev/null
+++ b/tests/test_live_forward_20d_dashboard.py
@@ -0,0 +1,175 @@
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+import pandas as pd
+
+from src.experiments.live_forward_20d_dashboard import (
+    build_horizon_dashboard_payload,
+    horizon_dashboard_html,
+)
+
+
+def _write_json(path: Path, payload: dict[str, object]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(payload), encoding="utf-8")
+
+
+def test_dashboard_payload_collects_public_arm_evidence(tmp_path: Path) -> None:
+    root = tmp_path / "reports" / "live_forward_horizon_matched_v2"
+    forecast_root = root / "forecasts" / "2026-07-29" / "ridge_h5"
+    _write_json(
+        forecast_root / "manifest.json",
+        {
+            "family": "ridge",
+            "horizon_sessions": 5,
+            "decision_date": "2026-07-29",
+            "model_refit_date": "2026-07-29",
+            "status": "pending",
+            "expected_entry_date": "2026-07-30",
+            "expected_maturity_date": "2026-08-06",
+            "refit_interval_sessions": 5,
+            "refitted_today": True,
+            "strategy_rebalance": True,
+            "strategic_rebalance": True,
+            "eligible_names": 250,
+            "selected_names": 2,
+            "sentiment_coverage": 1.0,
+            "feature_count": 2,
+            "features": ["momentum_63", "nlp_sentiment_3d"],
+            "explainability": {
+                "status": "available",
+                "method": "tree_shap",
+                "rows": 250,
+                "artifact": "C:/private/shap_summary.csv",
+            },
+        },
+    )
+    pd.DataFrame(
+        [
+            {
+                "ticker": "AAA",
+                "sector": "Technology",
+                "rank": 1,
+                "score": 0.03,
+                "weight": 0.5,
+                "private_column": "do-not-publish",
+            }
+        ]
+    ).to_csv(forecast_root / "selected_holdings.csv", index=False)
+
+    selection_path = tmp_path / "selection.csv"
+    pd.DataFrame(
+        [
+            {
+                "row_type": "aggregate",
+                "selection_eligible": True,
+                "top40_raw_excess": 0.012,
+                "rank_ic": 0.04,
+                "positive_ic_rate": 0.6,
+                "resolved_feature_count": 2,
+                "trial": 7,
+            }
+        ]
+    ).to_csv(selection_path, index=False)
+    _write_json(
+        root
+        / "model_versions"
+        / "ridge"
+        / "h5"
+        / "2026-07-29"
+        / "manifest.json",
+        {
+            "training_start": "2023-01-01",
+            "training_end": "2026-07-21",
+            "training_rows": 1000,
+            "model_config": {"alpha": 10.0},
+            "model_selection": {
+                "report": str(selection_path),
+                "selection_metric": "top40_raw_excess",
+            },
+        },
+    )
+    _write_json(
+        root / "forecasts" / "2026-07-29" / "run_summary.json",
+        {
+            "decision_date": "2026-07-29",
+            "status": "live_forward_20d_cycle_complete",
+            "matured_forecasts": 0,
+            "strategy_daily_return_rows": 0,
+            "tactical_orders": 0,
+            "model_failures": [],
+        },
+    )
+
+    summary = {
+        "experiment_version": "live_forward_horizon_matched_v2",
+        "decision_date": "2026-07-29",
+        "primary_horizon_sessions": 20,
+        "comparison_design": "horizon_matched",
+        "invested_weight": 1.0,
+        "rebalance_policy": {"status": "under_audit", "buffer_enabled": False},
+        "refit_sessions_by_horizon": {"5": 5, "10": 10, "20": 20},
+        "input_audit": {"market_end": "2026-07-29"},
+        "universe": {"final_names": 250},
+        "survival_validation": {"activated": False},
+    }
+    payload = build_horizon_dashboard_payload(root, summary, pd.DataFrame())
+
+    assert payload["schema_version"] == 1
+    assert len(payload["arms"]) == 1
+    arm = payload["arms"][0]
+    assert arm["selected"] == [
+        {
+            "ticker": "AAA",
+            "sector": "Technology",
+            "rank": 1,
+            "score": 0.03,
+            "weight": 0.5,
+        }
+    ]
+    assert arm["validation"]["top40_raw_excess"] == 0.012
+    assert arm["validation"]["completed_trials"] == 1
+    assert arm["explainability"] == {
+        "status": "available",
+        "method": "tree_shap",
+        "rows": 250,
+    }
+    assert "private_column" not in json.dumps(payload)
+    assert "C:/private" not in json.dumps(payload)
+    assert "model_versions" not in json.dumps(payload)
+
+
+def test_dashboard_html_exposes_v2_information_architecture() -> None:
+    html = horizon_dashboard_html(
+        {
+            "experiment": {"decision_date": "2026-07-29"},
+            "rebalance_policy": {},
+            "input_audit": {},
+            "universe": {},
+            "survival_validation": {},
+            "arms": [],
+            "matured": [],
+            "strategy_returns": [],
+            "runs": [],
+            "tactical_orders": [],
+        }
+    )
+
+    assert "Horizon-Matched Live Forward Monitor" in html
+    assert 'id="decisionSelect"' in html
+    assert 'id="horizonGrid"' in html
+    assert "Live Forward Monitor" in html
+    assert "../live_forward/dashboard.html" in html
+    assert 'id="strategyProgress"' in html
+    assert "The decision session is not a holding session" in html
+    assert "sessions held" in html
+    assert 'id="modelGrid"' in html
+    assert 'id="healthGrid"' in html
+    assert 'id="maturedTable"' in html
+    assert "never mixed with validation estimates" in html
+    assert "Show SHAP feature importance" in html
+    assert "live-forward-horizon-dashboard-v1" in html
+    assert "width=device-width,initial-scale=1" in html
+    assert "__DASHBOARD_PAYLOAD__" not in html