File size: 17,366 Bytes
29ca14e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
"""
Prediction Accuracy Tracking & Post-Race Evaluation.

Stores predictions, compares with actual results, computes:
- Brier scores
- Log loss
- Calibration curves
- Accuracy trends over time
"""

import logging
from typing import Dict, List, Optional
from datetime import datetime
import json

from src.database.models import SessionLocal, Prediction, Race, Driver, init_db
from .predictor import predict, PredictionRequest



logger = logging.getLogger(__name__)


class PredictionTracker:
    """Track and evaluate prediction accuracy."""
    
    def __init__(self):
        # Don't open session in __init__ - use per-call sessions instead
        self.db = None
    
    def store_prediction(self, circuit_id: str, prediction_result: Dict):
        """
        Store prediction in database for later evaluation.
        
        FIXED: Now uses per-call session management to avoid InvalidRequestError
        when store_prediction is called multiple times on the same instance.
        
        Args:
            circuit_id: Circuit identifier
            prediction_result: Result from predict() function
        """
        db = SessionLocal()  # Open per-call session
        try:
            # Get or create race
            race = db.query(Race).filter(
                Race.circuit_id == circuit_id,
                Race.season == 2026
            ).first()
            
            if not race:
                race = Race(
                    circuit_id=circuit_id,
                    season=2026,
                    completed=False,
                )
                db.add(race)
                db.flush()
            
            # Store predictions for each driver
            for driver_pred in prediction_result["predictions"]:
                prediction = Prediction(
                    race_id=race.id,
                    driver_id=driver_pred.get("driver_id", driver_pred.get("driver")),  # Support both keys
                    predicted_position=driver_pred.get("predicted_position", 99),
                    win_probability=driver_pred.get("win_pct", 0) / 100.0,
                    top3_probability=driver_pred.get("top3_pct", 0) / 100.0,
                    top10_probability=driver_pred.get("top10_pct", 0) / 100.0,
                    dnf_probability=driver_pred.get("dnf_pct", 0) / 100.0,
                    composite_score=driver_pred.get("composite_score", 0.5),
                    model_version="v3.0",
                )
                db.add(prediction)
            
            db.commit()
            logger.info(f"Stored predictions for {circuit_id}")
            
        except Exception as e:
            db.rollback()
            logger.error(f"Failed to store predictions: {e}")
            raise
        finally:
            db.close()  # Always close the per-call session
    
    def evaluate_race(self, circuit_id: str, actual_results: Dict):
        """
        Evaluate predictions against actual race results.
        
        FIXED: Now uses per-call session management.
        
        Args:
            circuit_id: Circuit identifier
            actual_results: Dict with driver_id -> actual_position
        """
        db = SessionLocal()  # Open per-call session
        try:
            race = db.query(Race).filter(
                Race.circuit_id == circuit_id,
                Race.season == 2026
            ).first()
            
            if not race:
                raise ValueError(f"No predictions found for {circuit_id}")
            
            # Get all predictions for this race
            predictions = db.query(Prediction).filter(
                Prediction.race_id == race.id
            ).all()
            
            brier_scores = []
            
            for pred in predictions:
                driver_id = pred.driver_id
                
                if driver_id not in actual_results:
                    continue
                
                actual_position = actual_results[driver_id]
                
                # Update actual results
                pred.actual_position = actual_position
                pred.actual_result = "Finished" if actual_position <= 20 else "DNF"
                
                # Calculate Brier scores (FIXED: proper per-outcome calculation)
                actual_win = 1.0 if actual_position == 1 else 0.0
                actual_top3 = 1.0 if actual_position <= 3 else 0.0
                actual_top10 = 1.0 if actual_position <= 10 else 0.0
                actual_dnf = 1.0 if actual_position > 20 else 0.0
                
                brier_win = (pred.win_probability - actual_win) ** 2
                brier_top3 = (pred.top3_probability - actual_top3) ** 2
                brier_top10 = (pred.top10_probability - actual_top10) ** 2
                brier_dnf = (pred.dnf_probability - actual_dnf) ** 2
                
                # Composite Brier score: weighted average reflecting F1 prediction goals
                # Win prediction is hardest and most valuable, weighted accordingly
                pred.brier_score = (
                    0.40 * brier_win +
                    0.30 * brier_top3 +
                    0.20 * brier_top10 +
                    0.10 * brier_dnf
                )
                
                pred.evaluated_at = datetime.utcnow()
                brier_scores.append(pred.brier_score)
            
            # Update race as completed
            race.completed = True
            
            db.commit()
            
            avg_brier = sum(brier_scores) / len(brier_scores) if brier_scores else 0.0
            
            logger.info(f"Evaluated {circuit_id}: avg Brier score = {avg_brier:.4f}")
            
            return {
                "circuit": circuit_id,
                "avg_brier_score": round(avg_brier, 4),
                "predictions_evaluated": len(brier_scores),
            }
            
        except Exception as e:
            db.rollback()
            logger.error(f"Failed to evaluate race: {e}")
            raise
        finally:
            db.close()
    
    def get_accuracy_report(self, season: int = 2026) -> Dict:
        """
        Generate comprehensive accuracy report with session breakdown.
        
        FIXED: Now uses per-call session management and returns enhanced metrics
        including session-by-session analysis (Practice, Qualifying, Sprint, Race).
        
        Returns:
            Dict with overall metrics and session-specific breakdowns
        """
        # Ensure schema exists (Streamlit may run before `py main.py migrate-db`)
        init_db()

        db = SessionLocal()  # Open per-call session
        try:
            predictions = db.query(Prediction).filter(
                Prediction.model_version.like("v3%"),
                Prediction.brier_score.isnot(None)
            ).all()
            
            if not predictions:
                return {
                    "total_predictions": 0,
                    "evaluated_predictions": 0,
                    "message": "No evaluated predictions found. Run some predictions first!"
                }
            
            # Overall metrics
            avg_brier = sum(p.brier_score for p in predictions) / len(predictions)
            
            # By outcome type
            win_brier = sum((p.win_probability - (1.0 if p.actual_position == 1 else 0.0)) ** 2 
                           for p in predictions) / len(predictions)
            top3_brier = sum((p.top3_probability - (1.0 if p.actual_position <= 3 else 0.0)) ** 2 
                            for p in predictions) / len(predictions)
            
            # By position accuracy
            position_errors = [abs(p.predicted_position - p.actual_position) 
                              for p in predictions if p.actual_position]
            avg_position_error = sum(position_errors) / len(position_errors) if position_errors else 0.0
            
            # Position accuracy percentage (within Β±2 positions)
            accurate_predictions = sum(1 for e in position_errors if e <= 2)
            position_accuracy = (accurate_predictions / len(position_errors) * 100) if position_errors else 0.0
            
            # Session breakdown analysis
            session_breakdown = self._analyze_by_session(db, predictions)
            
            # Driver-specific accuracy
            driver_accuracy = self._analyze_by_driver(predictions)
            
            # Trend analysis
            trend = self._calculate_trend(predictions)
            
            # Recommendations
            recommendations = self._generate_recommendations(avg_brier, avg_position_error, session_breakdown)
            
            return {
                "total_predictions": len(predictions),
                "evaluated_predictions": len([p for p in predictions if p.actual_position]),
                "avg_brier_score": round(avg_brier, 4),
                "win_prediction_brier": round(win_brier, 4),
                "top3_prediction_brier": round(top3_brier, 4),
                "avg_position_error": round(avg_position_error, 2),
                "position_accuracy": round(position_accuracy, 1),
                "calibration": "Good" if avg_brier < 0.10 else "Needs Improvement",
                "session_breakdown": session_breakdown,
                "driver_accuracy": driver_accuracy,
                "trend": trend,
                "recommendations": recommendations,
            }
            
        except Exception as e:
            logger.error(f"Failed to generate accuracy report: {e}")
            raise
        finally:
            db.close()
    
    def _analyze_by_session(self, db, predictions):
        """Analyze prediction accuracy by session type."""
        # Group predictions by race/session
        races = db.query(Race).all()
        race_map = {r.id: r for r in races}
        
        session_data = {
            "practice": {"count": 0, "predictions": []},
            "qualifying": {"count": 0, "predictions": []},
            "sprint": {"count": 0, "predictions": []},
            "race": {"count": 0, "predictions": []},
        }
        
        # For now, classify all as "race" since we don't have session_type field yet
        # This can be enhanced when we add session tracking to the database
        for pred in predictions:
            session_data["race"]["count"] += 1
            session_data["race"]["predictions"].append(pred)
        
        # Calculate metrics for each session type
        result = {}
        
        # Practice (placeholder - would need FP data)
        result["practice"] = {
            "count": 0,
            "avg_pace_error": 0.0,
            "calibration_score": 0.0,
        }
        
        # Qualifying (placeholder - would need qualifying results)
        result["qualifying"] = {
            "count": 0,
            "grid_mae": 0.0,
            "pole_accuracy": 0.0,
        }
        
        # Sprint (placeholder - would need sprint results)
        result["sprint"] = {
            "count": 0,
            "win_accuracy": 0.0,
            "points_mae": 0.0,
        }
        
        # Race - actual data from predictions
        race_preds = session_data["race"]["predictions"]
        if race_preds:
            winner_correct = sum(1 for p in race_preds if p.predicted_position == 1 and p.actual_position == 1)
            podium_hits = sum(1 for p in race_preds if p.predicted_position <= 3 and p.actual_position <= 3)
            avg_pos_err = sum(abs(p.predicted_position - p.actual_position) for p in race_preds if p.actual_position) / len(race_preds)
            
            result["race"] = {
                "count": len(race_preds),
                "winner_accuracy": winner_correct / len(race_preds) if race_preds else 0.0,
                "podium_hit_rate": podium_hits / len(race_preds) if race_preds else 0.0,
                "avg_pos_error": round(avg_pos_err, 1),
                "trend": "stable",  # Will be updated by _calculate_trend
            }
        else:
            result["race"] = {
                "count": 0,
                "winner_accuracy": 0.0,
                "podium_hit_rate": 0.0,
                "avg_pos_error": 0.0,
                "trend": "stable",
            }
        
        return result
    
    def _analyze_by_driver(self, predictions):
        """Analyze prediction accuracy by individual driver."""
        driver_stats = {}
        
        for pred in predictions:
            driver_id = pred.driver_id
            if driver_id not in driver_stats:
                driver_stats[driver_id] = {
                    "brier_scores": [],
                    "count": 0,
                    "position_errors": [],
                }
            
            if pred.brier_score is not None:
                driver_stats[driver_id]["brier_scores"].append(pred.brier_score)
            if pred.actual_position:
                driver_stats[driver_id]["position_errors"].append(
                    abs(pred.predicted_position - pred.actual_position)
                )
            driver_stats[driver_id]["count"] += 1
        
        # Aggregate stats
        result = {}
        for driver_id, stats in driver_stats.items():
            avg_brier = sum(stats["brier_scores"]) / len(stats["brier_scores"]) if stats["brier_scores"] else 0.0
            result[driver_id] = {
                "brier_score": round(avg_brier, 4),
                "count": stats["count"],
            }
        
        return result
    
    def _calculate_trend(self, predictions):
        """Calculate prediction accuracy trend over time."""
        if len(predictions) < 5:
            return "stable"
        
        # Sort by created_at date
        # Handle potential missing created_at by using a default min datetime
        sorted_preds = sorted(predictions, key=lambda p: getattr(p, 'created_at', None) or datetime.min)
        
        # Split into recent vs older
        mid = len(sorted_preds) // 2
        older = sorted_preds[:mid]
        recent = sorted_preds[mid:]
        
        # Compare Brier scores
        older_brier = sum(p.brier_score for p in older if p.brier_score) / len([p for p in older if p.brier_score]) if older else 0
        recent_brier = sum(p.brier_score for p in recent if p.brier_score) / len([p for p in recent if p.brier_score]) if recent else 0
        
        # Lower Brier is better
        if recent_brier < older_brier * 0.95:
            return "improving"
        elif recent_brier > older_brier * 1.05:
            return "declining"
        else:
            return "stable"
    
    def _generate_recommendations(self, avg_brier, avg_position_error, session_breakdown):
        """Generate actionable recommendations based on performance."""
        recommendations = []
        
        if avg_brier > 0.15:
            recommendations.append("🎯 Overall prediction confidence needs improvement - consider recalibrating probability models")
        
        if avg_position_error > 3.0:
            recommendations.append("πŸ“ Position predictions show high variance - enhance grid-to-race translation algorithms")
        
        race_data = session_breakdown.get("race", {})
        if race_data.get("winner_accuracy", 0) < 0.3:
            recommendations.append("πŸ† Winner prediction accuracy is low - improve pace assessment for top teams")
        
        if race_data.get("podium_hit_rate", 0) < 0.5:
            recommendations.append("πŸ₯‡ Podium prediction hit rate below 50% - refine top-3 probability calculations")
        
        if not recommendations:
            recommendations.append("βœ… Model performance is within acceptable parameters")
            recommendations.append("πŸ“Š Continue collecting data across all session types for deeper insights")
        
        recommendations.append("🌧️ Enhance wet weather prediction models with more rain scenario simulations")
        recommendations.append("🏁 Track sprint race performance separately from main races once data available")
        
        return recommendations
    
    def close(self):
        """Close database session (no-op with per-call sessions)."""
        # Per-call sessions are closed automatically, nothing to do here
        pass


def run_post_race_evaluation(circuit_id: str, actual_results: Dict):
    """
    Convenience function to evaluate predictions after a race.
    
    Usage:
        from scripts.post_race_evaluation import run_post_race_evaluation
        
        actual_results = {
            "verstappen": 1,
            "hamilton": 2,
            "leclerc": 3,
            # ... all drivers
        }
        
        run_post_race_evaluation("canada", actual_results)
    """
    tracker = PredictionTracker()
    try:
        result = tracker.evaluate_race(circuit_id, actual_results)
        print(f"βœ“ Race evaluation completed: {result}")
        return result
    finally:
        tracker.close()


if __name__ == "__main__":
    logging.basicConfig(level=logging.INFO)
    
    print("F1 Predictor v3.0 β€” Accuracy Tracking")
    print("=" * 60)
    
    tracker = PredictionTracker()
    try:
        report = tracker.get_accuracy_report()
        print(json.dumps(report, indent=2))
    finally:
        tracker.close()