Add backtest pipeline, betting_brain filters, score coherence + social v3

betting_brain.py:
- HARD_MIN_SAMPLES=50 floor for calibrator bypass
- ev_edge < 0 + >= 0.20 hard vetoes
- BTTS muted (grid search found no profitable config)
- Per-market optimal envelopes (MS, OU25)
- Score coherence filter: main_pick must agree with score prediction
- HTFT reversal cross-check for MS picks

feature_builder.py / data_loader.py:
- Real home/away_position from data (was hardcoded 10)
- Cup detection wired into UpsetEngine
- _estimate_league_position with 300-day season filter

New scripts:
- diagnostic_backtest.py: per-bet diagnostic backtest with loss patterns
- optimize_filters.py: grid search per-market optimal thresholds
- analyze_backtest_csv.py: root-cause hypothesis testing on CSV
- compare_backtests.py: side-by-side validation with verdict
- test_score_coherence.py: smoke test for coherence filter (20/20 pass)

Reports:
- diagnostic_backtest_20260525_024437 (50-match smoke)
- diagnostic_backtest_20260525_035649 (1000-match in-sample)
- filter_optimization_patch.json (grid search winners per market)

Social poster v3:
- satori + resvg HTML/CSS rendering pipeline
- Twemoji football/basketball + flag SVGs
- caption SEO: 12 curated hashtags per post
- image SEO: descriptive filenames + .json metadata sidecar
- /health, /preview-png, /run-now endpoints

Docs:
- mds/SESSION_HANDOFF.md: full session state for cross-machine continuity
- mds/SOCIAL_POSTER_SETUP.md: API keys + test commands

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

ai-engine/services/orchestrator/feature_builder.py

@@ -225,20 +225,43 @@ class FeatureBuilderMixin:
         if enrichment_failures:
             print(f"⚠️ Enrichment partial failures for {data.match_id}: {', '.join(enrichment_failures)}")
 
+        # ── Cup game detection (used by upset engine + elo dampening below) ──
+        _league_name_lower = (getattr(data, 'league_name', '') or '').lower()
+        _cup_keywords = ("kupa", "cup", "coupe", "copa", "coppa", "pokal",
+                         "trophy", "shield", "ziraat", "süper kupa", "super cup",
+                         "beker", "taça", "taca")
+        _is_cup_match = any(kw in _league_name_lower for kw in _cup_keywords)
+
+        # ── League size hint: top European leagues 18-20 teams, lower 16-24 ──
+        # We don't have a per-league team count, so fall back to 20 (standard).
+        # When standings infra lands this should pull from seasons table.
+        _league_total_teams = 20
+
         # Upset engine features
         upset_atmosphere, upset_motivation, upset_fatigue = 0.0, 0.0, 0.0
         try:
             upset_engine = get_upset_engine()
+            # Use the real position estimates from data_loader; fall back to mid-
+            # table (10) only when the loader couldn't compute one. Hardcoding 10
+            # for every team made motivation_score collapse to 0 for everyone.
+            _home_pos = getattr(data, 'home_position', None)
+            _away_pos = getattr(data, 'away_position', None)
+            if _home_pos is None or _home_pos <= 0:
+                _home_pos = 10
+            if _away_pos is None or _away_pos <= 0:
+                _away_pos = 10
             upset_feats = upset_engine.get_features(
                 home_team_name=getattr(data, 'home_team_name', '') or '',
                 home_team_id=data.home_team_id,
                 away_team_name=getattr(data, 'away_team_name', '') or '',
                 league_name=getattr(data, 'league_name', '') or '',
-                home_position=10,
-                away_position=10,
+                home_position=_home_pos,
+                away_position=_away_pos,
                 match_date_ms=data.match_date_ms,
+                is_cup_match=_is_cup_match,
                 home_days_rest=int(home_rest),
                 away_days_rest=int(away_rest),
+                total_teams=_league_total_teams,
             )
             upset_atmosphere = upset_feats.get('upset_atmosphere', 0.0)
             upset_motivation = upset_feats.get('upset_motivation', 0.0)
@@ -276,15 +299,10 @@ class FeatureBuilderMixin:
         is_season_start = 1.0 if match_month in (7, 8, 9) else 0.0
         is_season_end = 1.0 if match_month in (5, 6) else 0.0
 
-        # ── Cup game detection: dampen home advantage in feature space ──
-        _league_name = (getattr(data, 'league_name', '') or '').lower()
-        _cup_keywords = ("kupa", "cup", "coupe", "copa", "coppa", "pokal",
-                         "trophy", "shield", "ziraat", "süper kupa", "super cup")
-        _is_cup = any(kw in _league_name for kw in _cup_keywords)
-
         # ── Derived / Interaction features (V27) ──
         # Cup games: home ELO advantage is ~30% weaker (rotation, lower motivation)
-        elo_diff = (home_elo - away_elo) * (0.70 if _is_cup else 1.0)
+        # Uses _is_cup_match computed earlier (before upset engine call).
+        elo_diff = (home_elo - away_elo) * (0.70 if _is_cup_match else 1.0)
         form_elo_diff = home_form_elo_val - away_form_elo_val
         attack_vs_defense_home = data.home_goals_avg - data.away_conceded_avg
         attack_vs_defense_away = data.away_goals_avg - data.home_conceded_avg