File size: 11,048 Bytes
d6ea71e |
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 |
import json
from pathlib import Path
import numpy as np
import pandas as pd
import pytest
import socceraction.spadl as spadl
import socceraction.xthreat as xt
from pandera.typing import DataFrame, Series
from pytest_mock import MockerFixture
from sklearn.exceptions import NotFittedError
from socceraction.spadl import SPADLSchema
from socceraction.spadl.config import field_length, field_width
class TestGridCount:
"""Tests for counting the number of actions occuring in each grid cell.
Grid cells ares represented by 2D pitch coordinates. The (0,0) coordinate
corresponds to the bottom left corner of the pitch. The 2D coordinates are
mapped to a flat index. For a 2x2 grid, these flat indices are:
0 1
2 3
"""
N = 2
M = 2
def test_get_cell_indexes(self) -> None:
"""It should map pitch coordinates to a 2D cell index."""
x = Series[float]([0, field_length / 2 - 1, field_length])
y = Series[float]([0, field_width / 2 + 1, field_width])
xi, yi = xt._get_cell_indexes(x, y, self.N, self.M)
pd.testing.assert_series_equal(xi, pd.Series([0, 0, 1]))
pd.testing.assert_series_equal(yi, pd.Series([0, 1, 1]))
def test_get_cell_indexes_out_of_bounds(self) -> None:
"""It should map out-of-bounds coordinates to the nearest cell index."""
x = Series[float]([-10, field_length + 10])
y = Series[float]([-10, field_width + 10])
xi, yi = xt._get_cell_indexes(x, y, self.N, self.M)
pd.testing.assert_series_equal(xi, pd.Series([0, 1]))
pd.testing.assert_series_equal(yi, pd.Series([0, 1]))
def test_get_flat_indexes(self) -> None:
"""It should map pitch coordinates to a flat index."""
x = Series[float]([0, field_length / 2 - 1, field_length / 2 + 1, field_length])
y = Series[float]([0, field_width / 2 + 1, field_width / 2 - 1, field_width])
idx = xt._get_flat_indexes(x, y, self.N, self.M)
pd.testing.assert_series_equal(idx, pd.Series([2, 0, 3, 1]))
def test_count(self) -> None:
"""It should return the number of occurences in each grid cell."""
x = Series[float]([0, field_length / 2 - 1, field_length, field_length + 10])
y = Series[float]([0, field_width / 2 + 1, field_width, field_width + 10])
cnt = xt._count(x, y, self.N, self.M)
np.testing.assert_array_equal(cnt, [[1, 2], [1, 0]])
class TestModelPersistency:
def test_save_model(self, tmp_path: Path) -> None:
"""It should save a trained xT grid to a JSON file."""
p = tmp_path / "xt_model.json"
model = xt.ExpectedThreat()
model.xT = np.ones((model.w, model.l))
model.save_model(str(p))
assert p.read_text() == json.dumps(model.xT.tolist())
def test_save_model_not_fitted(self, tmp_path: Path) -> None:
"""It should raise an exception when saving an unfitted model."""
p = tmp_path / "xt_model.json"
model = xt.ExpectedThreat()
with pytest.raises(NotFittedError):
model.save_model(str(p))
model.xT = np.zeros((model.w, model.l))
with pytest.raises(NotFittedError):
model.save_model(str(p))
def test_save_model_file_exists(self, tmp_path: Path) -> None:
"""It should raise an exception when the file exists."""
p = tmp_path / "xt_model.json"
p.write_text("create file")
model = xt.ExpectedThreat()
model.xT = np.ones((model.w, model.l))
with pytest.raises(ValueError):
model.save_model(str(p), overwrite=False)
model.save_model(str(p), overwrite=True)
def test_load_model(self, tmp_path: Path) -> None:
"""It should load a saved xT grid from a JSON file."""
# xT grid
gridv = [[0.1, 0.2], [0.1, 0.0]]
# write to file
p = tmp_path / "xt_model.json"
p.write_text(json.dumps(gridv))
# load model
model = xt.load_model(str(p))
# verify
assert model.w == 2
assert model.l == 2
np.testing.assert_array_equal(model.xT, gridv)
def test_get_move_actions(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should filter passes, dribbles and crosses."""
move_actions = xt.get_move_actions(spadl_actions)
assert move_actions.type_id.isin(
[
spadl.config.actiontypes.index("pass"),
spadl.config.actiontypes.index("dribble"),
spadl.config.actiontypes.index("cross"),
]
).all()
def test_get_successful_move_actions(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should filter successful passes, dribbles and crosses."""
move_actions = xt.get_successful_move_actions(spadl_actions)
assert move_actions.type_id.isin(
[
spadl.config.actiontypes.index("pass"),
spadl.config.actiontypes.index("dribble"),
spadl.config.actiontypes.index("cross"),
]
).all()
assert (move_actions.result_id == spadl.config.results.index("success")).all()
def test_action_prob(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should return the proportion of shots and moves for each cell."""
shot_prob, move_prob = xt.action_prob(spadl_actions, 10, 5)
assert shot_prob.shape == (5, 10)
assert move_prob.shape == (5, 10)
assert np.any(shot_prob > 0)
assert np.any(move_prob > 0)
assert np.all(((move_prob + shot_prob) == 1) | ((move_prob + shot_prob) == 0))
def test_scoring_prob(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should return the proportion of successful shots for each cell."""
shots = spadl_actions.type_id == spadl.config.actiontypes.index("shot")
goals = shots & (spadl_actions.result_id == spadl.config.results.index("success"))
scoring_prob = xt.scoring_prob(spadl_actions, 1, 1)
assert scoring_prob.shape == (1, 1)
assert sum(goals) / sum(shots) == scoring_prob[0]
def test_move_transition_matrix() -> None:
"""It should return the move transition matrix."""
pass_id = spadl.config.actiontypes.index("pass")
success_id = spadl.config.results.index("success")
spadl_actions = DataFrame[SPADLSchema](
[
{
"game_id": 1,
"original_event_id": "a",
"action_id": 1,
"period_id": 1,
"time_seconds": 1.0,
"team_id": 1,
"player_id": 1,
"start_x": 10.0,
"end_x": 10.0,
"start_y": 10.0,
"end_y": 10.0,
"bodypart_id": 1,
"type_id": pass_id,
"result_id": success_id,
},
{
"game_id": 1,
"original_event_id": "a",
"action_id": 2,
"period_id": 1,
"time_seconds": 1.2,
"team_id": 1,
"player_id": 1,
"start_x": 10.0,
"end_x": 10.0,
"start_y": 10.0,
"end_y": 10.0,
"bodypart_id": 1,
"type_id": pass_id,
"result_id": success_id,
},
]
)
move_mat = xt.move_transition_matrix(spadl_actions, 2, 2)
assert np.sum(move_mat) == 1
assert move_mat.shape == (4, 4)
# (10, 10) is mapped to flat index 2 in a 2x2 grid
assert move_mat[2, 2] == 1
def test_xt_model_init() -> None:
"""It should initialize all instance variables."""
xTModel = xt.ExpectedThreat(l=8, w=6, eps=1e-3)
assert xTModel.l == 8
assert xTModel.w == 6
assert xTModel.eps == 1e-3
assert np.sum(xTModel.xT) == 0
assert xTModel.scoring_prob_matrix is None
assert xTModel.scoring_prob_matrix is None
assert xTModel.shot_prob_matrix is None
assert xTModel.move_prob_matrix is None
assert xTModel.transition_matrix is None
assert len(xTModel.heatmaps) == 0
def test_xt_model_fit(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should update all instance variables."""
xTModel = xt.ExpectedThreat()
xTModel.fit(spadl_actions)
assert xTModel.scoring_prob_matrix is not None
assert xTModel.shot_prob_matrix is not None
assert xTModel.move_prob_matrix is not None
assert xTModel.transition_matrix is not None
assert len(xTModel.heatmaps) > 0
assert np.sum(xTModel.xT) > 0
def test_xt_model_rate_not_fitted(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should raise a NotFittedError."""
xTModel = xt.ExpectedThreat()
with pytest.raises(NotFittedError):
xTModel.rate(spadl_actions)
def test_xt_model_rate(spadl_actions: DataFrame[SPADLSchema]) -> None:
"""It should rate all successful move actions and assign all other actions NaN."""
xTModel = xt.ExpectedThreat()
xTModel.fit(spadl_actions)
successful_move_actions_idx = xt.get_successful_move_actions(spadl_actions).index
ratings = xTModel.rate(spadl_actions)
assert ratings.shape == (len(spadl_actions),)
assert np.all(~np.isnan(ratings[successful_move_actions_idx]))
assert np.all(np.isnan(np.delete(ratings, successful_move_actions_idx)))
def test_interpolate_xt_grid_no_scipy(mocker: MockerFixture) -> None:
"""It should raise an ImportError if scipy is not installed."""
mocker.patch.object(xt, "interp2d", None)
xTModel = xt.ExpectedThreat()
with pytest.raises(ImportError, match="Interpolation requires scipy to be installed."):
xTModel.interpolator()
@pytest.fixture(scope="session")
def xt_model(sb_worldcup_data: pd.HDFStore) -> xt.ExpectedThreat:
"""Test the xT framework on the StatsBomb World Cup data."""
# 1. Load a set of actions to train the model on
df_games = sb_worldcup_data["games"].set_index("game_id")
# 2. Convert direction of play
actions_ltr = pd.concat(
[
spadl.play_left_to_right(
sb_worldcup_data[f"actions/game_{game_id}"], game.home_team_id
)
for game_id, game in df_games.iterrows()
]
).pipe(DataFrame[SPADLSchema])
# 3. Train xT model
xTModel = xt.ExpectedThreat(l=16, w=12)
xTModel.fit(actions_ltr)
return xTModel
@pytest.mark.e2e
def test_predict(sb_worldcup_data: pd.HDFStore, xt_model: xt.ExpectedThreat) -> None:
games = sb_worldcup_data["games"]
game = games.iloc[-1]
actions = sb_worldcup_data[f"actions/game_{game.game_id}"]
ratings = xt_model.rate(actions)
assert ratings.dtype is np.dtype(np.float64)
assert len(ratings) == len(actions)
@pytest.mark.e2e
def test_predict_with_interpolation(
sb_worldcup_data: pd.HDFStore, xt_model: xt.ExpectedThreat
) -> None:
games = sb_worldcup_data["games"]
game = games.iloc[-1]
actions = sb_worldcup_data[f"actions/game_{game.game_id}"]
ratings = xt_model.rate(actions, use_interpolation=True)
assert ratings.dtype is np.dtype(np.float64)
assert len(ratings) == len(actions)
|