Major refactor: Fantabeto 26/27 — modular package, GBM ensemble, MILP/MCTS optimization

Phase 1: Data Engineering - Refactored notebooks into src/{scraper,features,models,optimization,bot} - FBref scraper with proxy rotation + Playwright Cloudflare bypass - Fantacalcio.it integrated scraper (authenticated API + HTML fallback) - api-football RapidAPI client for supplementary xG/xA/injuries - RAG news pipeline: Gazzetta, Sky Sport, Di Marzio → injury/suspension/tactical extraction - 26/27 season config: teams, scoring rules, name mappings, news sources Phase 2: SOTA ML Architecture - GBM Ensemble (LightGBM + CatBoost + XGBoost) with stacked blending - Bootstrap ensemble for uncertainty quantification - SinhArcsinh distribution head (ported from original TF Probability) - Card classifiers (yellow/red), penalty model, goal probability (Poisson) - Temporal GNN for player interaction modeling (crosses→goals, passes→assists) - Optuna hyperparameter tuning with time-series CV Phase 3: Operations Research - Auction solver: MILP knapsack with PuLP (budget + role constraints) - Grid Auction (Asta a Griglia): Minimax game theory bidding strategy - Weekly lineup optimizer: MCTS maximizing win probability vs opponent - Modificatore Difesa integration + captain selection - Transfer market analyzer: buy-low/sell-high via xG regression to mean - Opponent behavior modeling from historical lineage patterns Phase 4: Agentic Workflow - Telegram bot: auto-briefing (Friday + Sunday morning) - Tactical briefing generator with start/sit recommendations - GitHub Actions CI/CD: scheduled pipeline (scrape → predict → notify) Infrastructure: - 31 pytest unit tests (features, models, scraper, optimization) - requirements.txt (lightgbm, catboost, xgboost, optuna, pulp, playwright, langchain) - Makefile with install/test/lint/scrape/train/bot targets - Jupyter notebook: 26_27_strategy.ipynb demonstrating auction + matchday 1 mockup - Completely rewritten README.md with architecture diagram
2026-08-11 13:16:07 +08:00
parent 6d9167596c
commit 3b065775f5
44 changed files with 5754 additions and 57 deletions
@@ -0,0 +1,155 @@
+"""Transfer market analysis ('Svincolati' / free agent pool).
+
+Identifies buy-low and sell-high targets using advanced metrics:
+- Regression to the mean: compares actual vs expected output
+- xG/xA vs actual goals/assists divergence
+- Minutes trending up/down
+- Market value arbitrage
+"""
+
+import logging
+from typing import Optional
+
+import numpy as np
+import pandas as pd
+
+logger = logging.getLogger(__name__)
+
+
+class TransferAnalyzer:
+    """Analyzes the Svincolati (free agent) market for arbitrage opportunities."""
+
+    def __init__(self, regression_factor: float = 0.3):
+        self.regression_factor = regression_factor
+
+    def compute_expected_output(
+        self, xg: float, xa: float, historical_mean: float
+    ) -> float:
+        """Compute regressed expected output using xG/xA.
+
+        Shrinks toward the player's historical mean (regression to the mean).
+        """
+        raw_expected = xg * 3.0 + xa * 1.0  # convert to Fantavoto scale
+        regressed = (
+            self.regression_factor * historical_mean +
+            (1 - self.regression_factor) * raw_expected
+        )
+        return regressed
+
+    def analyze_buy_low(
+        self, players_df: pd.DataFrame, min_minutes: int = 180
+    ) -> pd.DataFrame:
+        """Identify buy-low candidates.
+
+        Criteria:
+        - xG/xA significantly exceed actual output
+        - Minutes trending up
+        - Low market value relative to projection
+
+        Args:
+            players_df: DataFrame with columns:
+                [name, team, role, actual_fv_avg, xg, xa, minutes, market_value,
+                minutes_trend, historical_fv_avg]
+
+        Returns:
+            DataFrame of buy-low candidates ranked by opportunity.
+        """
+        df = players_df.copy()
+        df = df[df["minutes"] >= min_minutes]
+
+        if "xg" not in df.columns or "xa" not in df.columns:
+            logger.warning("xG/xA data missing; using basic analysis")
+            return pd.DataFrame()
+
+        # Expected fantavoto from xG/xA
+        df["expected_fv"] = df.apply(
+            lambda r: self.compute_expected_output(
+                r.get("xg", 0), r.get("xa", 0), r.get("historical_fv_avg", 6.0)
+            ),
+            axis=1,
+        )
+
+        # Divergence: expected minus actual
+        df["fv_divergence"] = df["expected_fv"] - df.get("actual_fv_avg", 6.0)
+
+        df["buy_low_score"] = (
+            df["fv_divergence"] * 2.0 +  # underperformance signal
+            df.get("minutes_trend", 0) * 0.5 +  # trending up
+            (1.0 / (df.get("market_value", 1) + 1)) * 10  # cheap
+        )
+
+        buy_low = df[df["buy_low_score"] > 0].sort_values("buy_low_score", ascending=False)
+
+        result = buy_low[[
+            "name", "team", "role", "actual_fv_avg", "expected_fv",
+            "fv_divergence", "buy_low_score", "market_value",
+        ]].copy()
+
+        result["recommendation"] = "BUY-LOW"
+        result["confidence"] = pd.cut(
+            result["buy_low_score"],
+            bins=[-np.inf, 1, 3, 5, np.inf],
+            labels=["Low", "Medium", "High", "Very High"],
+        )
+
+        logger.info(
+            f"Found {len(result)} buy-low candidates "
+            f"(avg divergence: {result['fv_divergence'].mean():.2f})"
+        )
+        return result
+
+    def analyze_sell_high(
+        self, players_df: pd.DataFrame, min_minutes: int = 180
+    ) -> pd.DataFrame:
+        """Identify sell-high candidates.
+
+        Criteria:
+        - Actual output exceeds xG/xA by large margin
+        - Minutes trending down
+        - High market value vs projection
+        """
+        df = players_df.copy()
+        df = df[df["minutes"] >= min_minutes]
+
+        if "xg" not in df.columns:
+            return pd.DataFrame()
+
+        df["expected_fv"] = df.apply(
+            lambda r: self.compute_expected_output(
+                r.get("xg", 0), r.get("xa", 0), r.get("historical_fv_avg", 6.0)
+            ),
+            axis=1,
+        )
+
+        # Overperformance
+        df["fv_divergence"] = df.get("actual_fv_avg", 6.0) - df["expected_fv"]
+
+        df["sell_high_score"] = (
+            df["fv_divergence"] * 3.0 +  # overperformance signal
+            (df.get("minutes_trend", 0) * -0.5 if "minutes_trend" in df.columns else 0)
+        )
+
+        sell_high = df[df["sell_high_score"] > 1].sort_values("sell_high_score", ascending=False)
+
+        result = sell_high[[
+            "name", "team", "role", "actual_fv_avg", "expected_fv",
+            "fv_divergence", "sell_high_score",
+        ]].copy()
+
+        result["recommendation"] = "SELL-HIGH"
+        logger.info(f"Found {len(result)} sell-high candidates")
+        return result
+
+    def full_transfer_report(self, players_df: pd.DataFrame) -> dict:
+        """Generate complete transfer market report."""
+        buy = self.analyze_buy_low(players_df)
+        sell = self.analyze_sell_high(players_df)
+
+        return {
+            "buy_low": buy,
+            "sell_high": sell,
+            "summary": (
+                f"Buy-low targets: {len(buy)} players identified. "
+                f"Sell-high targets: {len(sell)} players identified."
+            ),
+        }