Initial commit with 2026 World Cup Quant Platform core modules and CI/CD

platform/backend/app/analytics/player_props_sim.py

@@ -0,0 +1,79 @@
+"""球員道具盤（Props）蒙地卡羅模擬模組。"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+import numpy as np
+
+
+@dataclass(frozen=True)
+class PlayerPropsDistribution:
+    shots: np.ndarray
+    shots_on_target: np.ndarray
+    passes: np.ndarray
+
+
+def simulate_player_stats(
+    player_metrics: dict,
+    opponent_defense_metrics: dict,
+    iterations: int = 10_000,
+) -> PlayerPropsDistribution:
+    """快速模擬球員事件次數分佈。"""
+
+    if iterations <= 0:
+        raise ValueError('iterations 必須大於 0')
+
+    avg_touches = float(player_metrics.get('avg_touches', 45) or 0.0)
+    base_shot_rate = float(player_metrics.get('shots_per_touch', 0.08) or 0.0)
+    base_target_rate = float(player_metrics.get('shot_on_target_rate', 0.35) or 0.0)
+    base_pass_rate = float(player_metrics.get('passes_per_touch', 0.65) or 0.0)
+
+    opp_pressure = float(opponent_defense_metrics.get('pressing_index', 1.0) or 1.0)
+    opp_tackling = float(opponent_defense_metrics.get('marking_index', 1.0) or 1.0)
+
+    adj_touches = max(1.0, avg_touches * max(0.6, 1.0 / max(0.5, opp_pressure)))
+    shot_lambda = adj_touches * base_shot_rate
+    pass_lambda = adj_touches * base_pass_rate
+
+    rng = np.random.default_rng()
+    shots = rng.poisson(lam=shot_lambda, size=iterations)
+    passes = rng.poisson(lam=pass_lambda, size=iterations)
+
+    # 對方壓迫會降低射正率
+    effective_target_rate = max(0.02, base_target_rate / max(opp_tackling, 0.3))
+    shots_on_target = rng.binomial(shots, p=min(effective_target_rate, 0.99), size=iterations)
+
+    return PlayerPropsDistribution(shots=shots.astype(int), shots_on_target=shots_on_target.astype(int), passes=passes.astype(int))
+
+
+def evaluate_prop_bet(
+    simulated_distribution: PlayerPropsDistribution,
+    line: float,
+    odds: float,
+) -> dict[str, float | bool]:
+    """從 10,000 次模擬結果計算超過盤口機率與 EV。"""
+
+    if odds <= 1:
+        raise ValueError('odds 必須大於 1')
+    if line < 0:
+        raise ValueError('line 必須大於等於 0')
+
+    shots = simulated_distribution.shots
+    if shots.size == 0:
+        raise ValueError('distribution 為空')
+
+    probability_over = float((shots > line).mean())
+    from .ev_calculator import calculate_expected_value
+
+    ev = calculate_expected_value(probability_over, odds)
+
+    return {
+        'metric': 'shots',
+        'line': line,
+        'over_probability': round(probability_over, 6),
+        'under_probability': round(1.0 - probability_over, 6),
+        'implied_ev': ev['ev_value'],
+        'ev_percentage': ev['ev_percentage'],
+        'is_value_bet': bool(ev['is_value_bet']),
+    }