feat: ML-optimized accumulation scoring with dashboard toggle

Train GradientBoostedClassifier on 2,601 days of historical data
(2018-2025) to find optimal metric weights for identifying the best
long-term buying opportunities. Uses time-series cross-validation
to prevent look-ahead bias.

Key results:
- pct_above_200w_sma: 50.7% weight (was 11.1% equal)
- drawdown: 14.6%, lth_rp: 10.9%, rhodl: 8.9%
- fear_greed demoted from 11.1% to 5.1%
- nupl/mvrv nearly eliminated (0.7-1.8%)

ML Strong Accumulation bracket: avg +210% 1yr (vs +176% classic)

New files: ml/optimizer.py, config/ml_weights.json
Modified: scoring/engine.py (score_all_ml), backtesting/engine.py
(ml_mode), dashboard/server.py (Classic/ML toggle)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

scoring/engine.py

@@ -428,3 +428,104 @@ def score_all(metrics):
         "scored_count": len(valid_scores),
         "total_count": len(results),
     }
+
+
+# ── ML-Optimized Scoring ──────────────────────────────────────────────
+
+ML_WEIGHTS_PATH = os.path.join(
+    os.path.dirname(os.path.dirname(os.path.abspath(__file__))),
+    "config",
+    "ml_weights.json",
+)
+
+# Maps scoring engine metric keys to ML weight keys
+_ML_KEY_MAP = {
+    "fear_greed": "fear_greed",
+    "puell_multiple": "puell_multiple",
+    "mvrv_zscore": "mvrv_zscore",
+    "drawdown": "drawdown",
+    "price_vs_200w_sma": "pct_above_200w_sma",
+    "reserve_risk": "reserve_risk",
+    "rhodl_ratio": "rhodl_ratio",
+    "nupl": "nupl",
+    "lth_realized_price": "pct_above_lth_rp",
+}
+
+
+def load_ml_weights():
+    """Load ML-optimized weights from config."""
+    try:
+        with open(ML_WEIGHTS_PATH) as f:
+            data = json.load(f)
+        return data.get("weights", {})
+    except Exception:
+        return {}
+
+
+def score_all_ml(metrics):
+    """Score all metrics using ML-optimized weights.
+
+    Same output format as score_all() but uses learned weights
+    instead of equal weighting. Each metric still shows its
+    individual 0-10 score plus the ML weight applied to it.
+    """
+    # Get classic scores first (reuses all individual scoring logic)
+    classic = score_all(metrics)
+    ml_weights = load_ml_weights()
+
+    if not ml_weights:
+        # Fallback to classic if no ML weights available
+        classic["ml_mode"] = False
+        classic["ml_error"] = "ML weights not found — run ml/optimizer.py"
+        return classic
+
+    results = classic["metrics"]
+
+    # Compute ML-weighted composite
+    weighted_sum = 0.0
+    weight_total = 0.0
+
+    for m in results:
+        if m["score"] is None:
+            continue
+        ml_key = _ML_KEY_MAP.get(m["key"])
+        if ml_key is None:
+            # Hash ribbons or unknown metric — use small default weight
+            w = 0.01
+        else:
+            w = ml_weights.get(ml_key, 0.0)
+
+        m["ml_weight"] = round(w, 4)
+        m["ml_contribution"] = round(m["score"] * w * 10, 2)
+        weighted_sum += m["score"] * w
+        weight_total += w
+
+    # Normalize if weights don't sum to 1 (e.g., missing metrics)
+    if weight_total > 0:
+        composite = weighted_sum / weight_total * 10
+    else:
+        composite = 0
+
+    # Assessment text (same thresholds as classic)
+    if composite >= 80:
+        assessment = "EXTREME ACCUMULATION ZONE"
+    elif composite >= 65:
+        assessment = "STRONG ACCUMULATION ZONE"
+    elif composite >= 50:
+        assessment = "MODERATE OPPORTUNITY"
+    elif composite >= 35:
+        assessment = "NEUTRAL"
+    elif composite >= 20:
+        assessment = "CAUTION — OVERHEATED"
+    else:
+        assessment = "EXTREME CAUTION"
+
+    return {
+        "metrics": results,
+        "composite_score": round(composite, 1),
+        "assessment": assessment,
+        "scored_count": classic["scored_count"],
+        "total_count": classic["total_count"],
+        "ml_mode": True,
+        "classic_score": classic["composite_score"],
+    }