fantabeto/tests/test_models.py

"""Tests for the model and optimization modules."""

import numpy as np
import pandas as pd
import pytest


class TestGBMEnsemble:
    def test_fit_predict(self):
        from src.models.gbm_model import GBMEnsemble

        np.random.seed(42)
        X = pd.DataFrame(np.random.randn(100, 10))
        y = pd.Series(np.random.randn(100) * 2 + 6.5)  # ~fantavoto range

        model = GBMEnsemble(n_estimators=50, n_bootstrap=20)
        model.fit(X, y)

        preds = model.predict(X)
        assert len(preds) == len(y)
        assert preds.dtype == np.float64

    def test_distribution_prediction(self):
        from src.models.gbm_model import GBMEnsemble

        np.random.seed(42)
        X = pd.DataFrame(np.random.randn(50, 5))
        y = pd.Series(np.random.randn(50) + 6.5)

        model = GBMEnsemble(n_estimators=50, n_bootstrap=30)
        model.fit(X, y)

        mean, std = model.predict_distribution(X)
        assert len(mean) == len(y)
        assert len(std) == len(y)
        assert np.all(std > 0)

    def test_feature_importance(self):
        from src.models.gbm_model import GBMEnsemble

        np.random.seed(42)
        X = pd.DataFrame(np.random.randn(100, 5))
        X.columns = ["f1", "f2", "f3", "f4", "f5"]
        y = pd.Series(np.random.randn(100) + 6.5)

        model = GBMEnsemble(n_estimators=50)
        model.fit(X, y)

        importance = model.feature_importance()
        assert len(importance) == 5


class TestSinhArcsinhDistribution:
    def test_normal_case(self):
        from src.models.distribution_head import (
            SinhArcsinhDistribution, sinh_arcsinh_params,
        )

        raw = np.array([[6.5, 0.5, 0.0, 0.5]])  # loc=6.5, scale~softplus(0.5)
        loc, scale, skew, tail = sinh_arcsinh_params(raw)

        dist = SinhArcsinhDistribution(loc, scale, skew, tail)
        mean = dist.mean()
        assert 5 < mean[0] < 8  # reasonable range

    def test_skewed_case(self):
        from src.models.distribution_head import SinhArcsinhDistribution

        # Attacker with high upside skew
        dist = SinhArcsinhDistribution(
            np.array([7.0]), np.array([1.0]),
            np.array([1.5]), np.array([1.2]),  # positive skew → right tail
        )
        mean = dist.mean()
        assert mean[0] > 7.0  # right-skewed → mean > loc


class TestCardClassifier:
    def test_fit_predict(self):
        from src.models.card_model import CardClassifier

        np.random.seed(42)
        n = 200
        X = pd.DataFrame(np.random.randn(n, 10))
        # ~15% yellow card rate
        y = pd.Series((np.random.rand(n) < 0.15).astype(int))

        model = CardClassifier(card_type="yellow")
        model.fit(X, y)

        probs = model.predict_proba(X)
        assert len(probs) == n
        assert np.all(probs >= 0) and np.all(probs <= 1)


class TestAuctionSolver:
    def test_solve(self):
        from src.optimization.auction_solver import AuctionSolver, AuctionConfig

        config = AuctionConfig(
            total_budget=500, n_gk=3, n_def=8, n_mid=8, n_fwd=6,
        )
        solver = AuctionSolver(config=config)

        # Create a small player pool
        players = []
        roles = ["P"] * 5 + ["D"] * 15 + ["C"] * 15 + ["A"] * 10
        np.random.seed(42)
        for i, role in enumerate(roles):
            players.append({
                "name": f"Player_{i}",
                "team": f"Team_{i % 20}",
                "role": role,
                "projected_points": np.random.uniform(5, 9),
                "market_value": np.random.randint(5, 30),
            })

        solver.add_players(pd.DataFrame(players))
        result = solver.solve()

        assert "selected_players" in result
        assert len(result["selected_players"]) == config.n_gk + config.n_def + config.n_mid + config.n_fwd
        assert result["total_cost"] <= config.total_budget

    def test_grid_auction(self):
        from src.optimization.auction_solver import AuctionSolver, PlayerValuation

        solver = AuctionSolver()
        round_players = [
            PlayerValuation("Player_A", "Inter", "A", 8.5, 30, 60),
            PlayerValuation("Player_B", "Milan", "A", 8.0, 25, 55),
            PlayerValuation("Player_C", "Juventus", "D", 7.0, 15, 35),
        ]

        recs = solver.grid_auction_strategy(round_players)
        assert "Player_A" in recs
        assert recs["Player_A"]["max_bid"] > 0


class TestLineupSolver:
    def test_optimize(self):
        from src.optimization.lineup_solver import LineupSolver, PlayerScore

        np.random.seed(42)
        pool = []
        for i in range(25):
            role = (
                "P" if i == 0 else
                "D" if i < 9 else
                "C" if i < 17 else "A"
            )
            pool.append(PlayerScore(
                name=f"P{i}", role=role, team="T", oppteam="O",
                home=True, fv_mean=np.random.uniform(5.5, 8),
                fv_std=1.0, mv_mean=6.0, mv_std=0.5,
                starter_prob=np.random.uniform(0.5, 1.0),
            ))

        solver = LineupSolver(iters=500)
        result = solver.optimize(pool)

        assert "lineup" in result
        assert len(result["lineup"]) == 11
        assert result["win_probability"] > 0

    def test_validate_lineup(self):
        from src.optimization.lineup_solver import LineupSolver, PlayerScore

        solver = LineupSolver()

        valid = [
            PlayerScore("GK", "P", "T", "O", True, 6.0, 1, 6.0, 0.5),
            *[PlayerScore(f"D{j}", "D", "T", "O", True, 6.0, 1, 6.0, 0.5) for j in range(4)],
            *[PlayerScore(f"M{k}", "C", "T", "O", True, 6.0, 1, 6.0, 0.5) for k in range(4)],
            *[PlayerScore(f"F{l}", "A", "T", "O", True, 6.0, 1, 6.0, 0.5) for l in range(2)],
        ]
        assert solver._validate_lineup(valid)

        # Too few defenders
        invalid = [
            PlayerScore("GK", "P", "T", "O", True, 6.0, 1, 6.0, 0.5),
            *[PlayerScore(f"D{j}", "D", "T", "O", True, 6.0, 1, 6.0, 0.5) for j in range(2)],
            *[PlayerScore(f"M{k}", "C", "T", "O", True, 6.0, 1, 6.0, 0.5) for k in range(5)],
            *[PlayerScore(f"F{l}", "A", "T", "O", True, 6.0, 1, 6.0, 0.5) for l in range(3)],
        ]
        assert not solver._validate_lineup(invalid)

    def test_modificatore(self):
        from src.optimization.lineup_solver import LineupSolver

        solver = LineupSolver()
        # 3 defenders averaging 7.0 → +6 bonus
        bonus = solver._modificatore_bonus([7.0, 7.0, 7.0])
        assert bonus == 6.0

        # Average 6.3 → +1
        bonus = solver._modificatore_bonus([6.5, 6.0, 6.4])
        assert bonus == 1.0

        # Average 5.5 → 0
        bonus = solver._modificatore_bonus([5.5, 5.5, 5.5])
        assert bonus == 0.0


class TestTransferAnalyzer:
    def test_buy_low(self):
        from src.optimization.transfer_analyzer import TransferAnalyzer

        analyzer = TransferAnalyzer()
        df = pd.DataFrame([
            {
                "name": "Underperformer", "team": "TeamA", "role": "A",
                "actual_fv_avg": 4.5, "xg": 0.9, "xa": 0.5,
                "minutes": 900, "market_value": 3,
                "minutes_trend": 1, "historical_fv_avg": 7.0,
            },
            {
                "name": "Overperformer", "team": "TeamB", "role": "A",
                "actual_fv_avg": 9.0, "xg": 0.3, "xa": 0.1,
                "minutes": 500, "market_value": 30,
                "minutes_trend": -1, "historical_fv_avg": 6.5,
            },
        ])

        buy = analyzer.analyze_buy_low(df)
        assert len(buy) >= 0  # May or may not find candidates with this data

    def test_sell_high(self):
        from src.optimization.transfer_analyzer import TransferAnalyzer

        analyzer = TransferAnalyzer()
        df = pd.DataFrame([
            {
                "name": "Overperformer", "team": "TeamB", "role": "A",
                "actual_fv_avg": 9.0, "xg": 0.3, "xa": 0.1,
                "minutes": 500, "historical_fv_avg": 6.5,
            },
        ])

        sell = analyzer.analyze_sell_high(df)
        assert len(sell) > 0
        assert sell.iloc[0]["recommendation"] == "SELL-HIGH"


class TestOpponentModel:
    def test_predict_lineup(self):
        from src.optimization.opponent_model import OpponentModel

        model = OpponentModel()

        # Add some history
        model.add_lineup(1, ["GK1", "D1", "D2", "D3", "D4", "M1", "M2", "M3", "M4", "F1", "F2"], "4-4-2")
        model.add_lineup(2, ["GK1", "D1", "D2", "D3", "D4", "M1", "M2", "M3", "M5", "F1", "F2"], "4-4-2")

        squad = [
            {"name": "GK1", "role": "P", "fv_mean": 6.5, "starter_prob": 0.95},
            *[{"name": f"D{j}", "role": "D", "fv_mean": 6.0, "starter_prob": 0.8} for j in range(8)],
            *[{"name": f"M{k}", "role": "C", "fv_mean": 6.5, "starter_prob": 0.8} for k in range(8)],
            *[{"name": f"F{l}", "role": "A", "fv_mean": 7.0, "starter_prob": 0.8} for l in range(6)],
        ]

        result = model.predict_lineup(squad)
        assert "predicted_lineup" in result
        assert len(result["predicted_lineup"]) == 11
        assert result["predicted_formation"] == "4-4-2"