spforge 0.8.8__py3-none-any.whl → 0.8.19__py3-none-any.whl

{spforge-0.8.8.dist-info → spforge-0.8.19.dist-info}/RECORD

@@ -14,7 +14,7 @@ examples/nba/data/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU
 examples/nba/data/game_player_subsample.parquet,sha256=ODJxHC-mUYbJ7r-ScUFtPU7hrFuxLUbbDSobmpCkw0w,279161
 examples/nba/data/utils.py,sha256=41hxLQ1d6ZgBEcHa5MI0-fG5KbsRi07cclMPQZM95ek,509
 spforge/__init__.py,sha256=8vZhy7XUpzqWkVKpXqwqOLDkQlNytRhyf4qjwObfXgU,468
-spforge/autopipeline.py,sha256=ZUwv6Q6O8cD0u5TiSqG6lhW0j16RlSb160AzuOeL2R8,23186
+spforge/autopipeline.py,sha256=rZ6FhJxcgNLvtr3hTVkEiW4BiorgXxADThfMuQ42orE,29866
 spforge/base_feature_generator.py,sha256=RbD00N6oLCQQcEb_VF5wbwZztl-X8k9B0Wlaj9Os1iU,668
 spforge/data_structures.py,sha256=k82v5r79vl0_FAVvsxVF9Nbzb5FoHqVrlHZlEXGc5gQ,7298
 spforge/features_generator_pipeline.py,sha256=n8vzZKqXNFcFRDWZhllnkhAh5NFXdOD3FEIOpHcay8E,8208
@@ -30,7 +30,7 @@ spforge/estimator/__init__.py,sha256=zIJ4u7WGPOALPx8kVBppBOqklI4lQPl9QBWT8JjjFoY
 spforge/estimator/_conditional_estimator.py,sha256=JSHpOg5lv3kRv_VzSZ0fKbwCO2dJv9XpyLs9lS81psU,4904
 spforge/estimator/_frequency_bucketing_classifier.py,sha256=d7wDpOCoKWf-WoXtzwahjtmAozkFdKE3-pzs477WMYc,6055
 spforge/estimator/_granularity_estimator.py,sha256=pUNmtpDFoOVbS9mHfO-zvidPIKJgWts0y2VnhJ8VWww,3829
-spforge/estimator/_group_by_estimator.py,sha256=aXuDvRWvvgK4SEI_DMYscvathmPb6nkMxnqKgG8HC0Y,2769
+spforge/estimator/_group_by_estimator.py,sha256=o-xv_PJJyWBaKv5Eo4EPbOvb9i0CuebZnX4GtEFp_Js,3120
 spforge/estimator/_ordinal_classifier.py,sha256=j_dfVHeX-6eZgPwwsYbkbP6bPrKH2a5S-N8vfP5hneA,1993
 spforge/estimator/_sklearn_enhancer_estimator.py,sha256=DZ-UlmeazXPd6uEnlbVv79syZ5FPa64voUyKArtjjUs,4664
 spforge/feature_generator/__init__.py,sha256=wfLfUkC_lLOCpy7NgDytK-l3HUAuhikuQXdKCgSGbuA,556
@@ -43,36 +43,36 @@ spforge/feature_generator/_rolling_mean_binary.py,sha256=lmODy-o9Dd9pb8IlA7g4UyA
 spforge/feature_generator/_rolling_mean_days.py,sha256=EZQmFmYVQB-JjZV5k8bOWnaTxNpPDCZAjdfdhiiG4r4,8415
 spforge/feature_generator/_rolling_window.py,sha256=HT8LezsRIPNAlMEoP9oTPW2bKFu55ZSRnQZGST7fncw,8836
 spforge/feature_generator/_utils.py,sha256=KDn33ia1OYJTK8THFpvc_uRiH_Bl3fImGqqbfzs0YA4,9654
-spforge/hyperparameter_tuning/__init__.py,sha256=N2sKG4SvG41hlsFT2kx_DQYMmXsQr-8031Tu_rxlxyY,1015
-spforge/hyperparameter_tuning/_default_search_spaces.py,sha256=Sm5IrHAW0-vRC8jqCPX0pDi_C-W3L_MoEKGA8bx1Zbc,7546
-spforge/hyperparameter_tuning/_tuner.py,sha256=uovhGqhe8-fdhi79aErUmE2h5NCycFQEIRv5WCjpC7E,16732
+spforge/hyperparameter_tuning/__init__.py,sha256=Vcl8rVlJ7M708iPgqe4XxpZWgJKGux0Y5HgMCymRsHg,1099
+spforge/hyperparameter_tuning/_default_search_spaces.py,sha256=SjwXLpvYIu_JY8uPRHeL5Kgp1aa0slWDz8qsKDaohWQ,8020
+spforge/hyperparameter_tuning/_tuner.py,sha256=M79q3saM6r0UZJsRUUgfdDr-3Qii-F2-wuSAZLFtZDo,19246
 spforge/performance_transformers/__init__.py,sha256=U6d7_kltbUMLYCGBk4QAFVPJTxXD3etD9qUftV-O3q4,422
-spforge/performance_transformers/_performance_manager.py,sha256=KwAga6dGhNkXi-MDW6LPjwk6VZwCcjo5L--jnk9aio8,9706
+spforge/performance_transformers/_performance_manager.py,sha256=WmjmlMEnq7y75MiI_s9Y-9eMXIyhPTUKrwsXRtgYp0k,9620
 spforge/performance_transformers/_performances_transformers.py,sha256=0lxuWjAfWBRXRgQsNJHjw3P-nlTtHBu4_bOVdoy7hq4,15536
 spforge/ratings/__init__.py,sha256=OZVH2Lo6END3n1X8qi4QcyAPlThIwAYwVKCiIuOQSQU,576
 spforge/ratings/_base.py,sha256=dRMkIGj5-2zKddygaEA4g16WCyXon7v8Xa1ymm7IuoM,14335
-spforge/ratings/_player_rating.py,sha256=MyqsyLSY6d7_bxDSnF8eWOyXpSCADWGdepdFSGM4cHw,51365
+spforge/ratings/_player_rating.py,sha256=JSTXdaRw_b8ZoZxgmMnZrYG7gPg8GKawqalLd16SK1M,56066
 spforge/ratings/_team_rating.py,sha256=T0kFiv3ykYSrVGGsVRa8ZxLB0WMnagxqdFDzl9yZ_9g,24813
 spforge/ratings/enums.py,sha256=s7z_RcZS6Nlgfa_6tasO8_IABZJwywexe7sep9DJBgo,1739
 spforge/ratings/league_identifier.py,sha256=_KDUKOwoNU6RNFKE5jju4eYFGVNGBdJsv5mhNvMakfc,6019
 spforge/ratings/league_start_rating_optimizer.py,sha256=Q4Vo3QT-r55qP4aD9WftsTB00UOSRvxM1khlyuAGWNM,8582
 spforge/ratings/player_performance_predictor.py,sha256=cMxzQuk0nF1MsT_M32g-3mxVdAEbZ-S7TUjEPYdo3Yg,8361
-spforge/ratings/start_rating_generator.py,sha256=_7hIJ9KRVCwsCoY1GIzY8cuOdHR8RH_BCMeMwQG3E04,6776
+spforge/ratings/start_rating_generator.py,sha256=eSasa5Oe9n4IoTGjFCYyFQAGrJtzrBW-Qor97lmaYuM,6776
 spforge/ratings/team_performance_predictor.py,sha256=ThQOmYQUqKBB46ONYHOMM2arXFH8AkyKpAZzs80SjHA,7217
-spforge/ratings/team_start_rating_generator.py,sha256=ZJe84sTvE4Yep3d4wKJMMJn2Q4PhcCwkO7Wyd5nsYUA,5110
-spforge/ratings/utils.py,sha256=qms5J5SD-FyXDR2G8giDMbu_AoLgI135pjW4nghxROg,3940
+spforge/ratings/team_start_rating_generator.py,sha256=vK-_m8KwcHopchch_lKNHSGLiiNm5q9Lenm0d1cP_po,5110
+spforge/ratings/utils.py,sha256=_zFemqz2jJkH8rn2EZpDt8N6FELUmYp9qCnPzRtOIGU,4497
 spforge/scorer/__init__.py,sha256=wj8PCvYIl6742Xwmt86c3oy6iqE8Ss-OpwHud6kd9IY,256
-spforge/scorer/_score.py,sha256=TR0T9nJj0aeVgGfOE0fZmXlO66CELulYwxhi7ZAxhvY,56184
+spforge/scorer/_score.py,sha256=kNuqiK3F5mUEAVD7KjWYY7E_AkRrspR362QBm_jyElg,57623
 spforge/transformers/__init__.py,sha256=IPCsMcsgBqG52d0ttATLCY4HvFCQZddExlLt74U-zuI,390
 spforge/transformers/_base.py,sha256=-smr_McQF9bYxM5-Agx6h7Xv_fhZzPfpAdQV-qK18bs,1134
 spforge/transformers/_net_over_predicted.py,sha256=5dC8pvA1DNO0yXPSgJSMGU8zAHi-maUELm7FqFQVo-U,2321
 spforge/transformers/_operator.py,sha256=jOH7wdMBLg6R2hlH_FU6eA0gjs-Q0vFimTo7fXgKpjI,2964
-spforge/transformers/_other_transformer.py,sha256=xLfaFIhkFsigAoitB4x3F8An2j9ymdjQy5VrsTvJlrA,3152
+spforge/transformers/_other_transformer.py,sha256=w2a7Wnki3vJe4GAkSa4kealw0GILIo6nE_9-3M10owA,4646
 spforge/transformers/_predictor.py,sha256=2sE6gfVrilXzPVcBurSrtqHw33v2ljygQcEYXt9LhZc,3119
 spforge/transformers/_simple_transformer.py,sha256=zGUFNQYMeoDSa2CoQejQNiNmKCBN5amWTvyOchiUHj0,5660
 spforge/transformers/_team_ratio_predictor.py,sha256=g8_bR53Yyv0iNCtol1O9bgJSeZcIco_AfbQuUxQJkeY,6884
-spforge-0.8.8.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-tests/test_autopipeline.py,sha256=WXHeqBdjQD6xaXVkzvS8ocz0WVP9R7lN0PiHJ2iD8nA,16911
+spforge-0.8.19.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+tests/test_autopipeline.py,sha256=7cNAn-nmGolfyfk3THh9IKcHZfRA-pLYC_xAyMg-No4,26863
 tests/test_autopipeline_context.py,sha256=IuRUY4IA6uMObvbl2pXSaXO2_tl3qX6wEbTZY0dkTMI,1240
 tests/test_feature_generator_pipeline.py,sha256=CK0zVL8PfTncy3RmG9i-YpgwjOIV7yJhV7Q44tbetI8,19020
 tests/cross_validator/test_cross_validator.py,sha256=itCGhNY8-NbDbKbhxHW20wiLuRst7-Rixpmi3FSKQtA,17474
@@ -81,7 +81,7 @@ tests/end_to_end/test_estimator_hyperparameter_tuning.py,sha256=fZCJ9rrED2vT68B9
 tests/end_to_end/test_league_start_rating_optimizer.py,sha256=Mmct2ixp4c6L7PGym8wZc7E-Csozryt1g4_o6OCc1uI,3141
 tests/end_to_end/test_lol_player_kills.py,sha256=RJSYUbPrZ-RzSxGggj03yN0JKYeTB1JghVGYFMYia3Y,11891
 tests/end_to_end/test_nba_player_points.py,sha256=kyzjo7QIcvpteps29Wix6IS_eJG9d1gHLeWtIHpkWMs,9066
-tests/end_to_end/test_nba_player_ratings_hyperparameter_tuning.py,sha256=LXRkI_6Ho2kzJVbNAM17QFhx_MP9WdDJXCO9dWgJGNA,6491
+tests/end_to_end/test_nba_player_ratings_hyperparameter_tuning.py,sha256=0lI4Xtg3V-zmo6prgzdNG80yy7JjvFVO-J_OU0pljyc,6346
 tests/end_to_end/test_nba_prediction_consistency.py,sha256=o3DckJasx_I1ed6MhMYZUo2WSDvQ_p3HtJa9DCWTIYU,9857
 tests/estimator/test_sklearn_estimator.py,sha256=tVfOP9Wx-tV1b6DcHbGxQHZQzNPA0Iobq8jTcUrk59U,48668
 tests/feature_generator/test_lag.py,sha256=5Ffrv0V9cwkbkzRMPBe3_c_YNW-W2al-XH_acQIvdeg,19531
@@ -91,13 +91,15 @@ tests/feature_generator/test_rolling_mean_binary.py,sha256=KuIavJ37Pt8icAb50B23l
 tests/feature_generator/test_rolling_mean_days.py,sha256=EyOvdJDnmgPfe13uQBOkwo7fAteBQx-tnyuGM4ng2T8,18884
 tests/feature_generator/test_rolling_window.py,sha256=YBJo36OK3ILYeXrH06ylXqviUcCaGYaVQaK5RJzwM7Y,23239
 tests/hyperparameter_tuning/test_estimator_tuner.py,sha256=iewME41d6LR2aQ0OtohGFtN_ocJUwTeqvs6L0QDmfG4,4413
-tests/hyperparameter_tuning/test_rating_tuner.py,sha256=PyCFP3KPc4Iy9E_X9stCVxra14uMgC1tuRwuQ30rO_o,13195
-tests/performance_transformers/test_performance_manager.py,sha256=bfC5GiBuzHw-mLmKeEzBUUPuKm0ayax2bsF1j88W8L0,10120
+tests/hyperparameter_tuning/test_rating_tuner.py,sha256=usjC2ioO_yWRjjNAlRTyMVYheOrCi0kKocmHQHdTmpM,18699
+tests/performance_transformers/test_performance_manager.py,sha256=gjuuV_hb27kCo_kUecPKG3Cbot2Gqis1W3kw2A4ovS4,10690
 tests/performance_transformers/test_performances_transformers.py,sha256=A-tGiCx7kXrj1cVj03Bc7prOeZ1_Ryz8YFx9uj3eK6w,11064
-tests/ratings/test_player_rating_generator.py,sha256=FGH3Tq0uFoSlkS_XMldsUKhsovBRBvzH9EbqjKvg2O0,59601
+tests/ratings/test_player_rating_generator.py,sha256=SKLaBQBsHYslc2Nia2AxZ8A9Cy16MbZAWjLyOjvcMnA,64094
+tests/ratings/test_player_rating_no_mutation.py,sha256=GzO3Hl__5K68DS3uRLefwnbcTJOvBM7cZqww4M21UZM,8493
 tests/ratings/test_ratings_property.py,sha256=ckyfGILXa4tfQvsgyXEzBDNr2DUmHwFRV13N60w66iE,6561
 tests/ratings/test_team_rating_generator.py,sha256=cDnf1zHiYC7pkgydE3MYr8wSTJIq-bPfSqhIRI_4Tic,95357
-tests/scorer/test_score.py,sha256=_Vd6tKpy_1GeOxU7Omxci4CFf7PvRGMefEI0gv2gV6A,74688
+tests/ratings/test_utils_scaled_weights.py,sha256=iHxe6ZDUB_I2B6HT0xTGqXBkl7gRlqVV0e_7Lwun5po,4988
+tests/scorer/test_score.py,sha256=rw3xJs6xqWVpalVMUQz557m2JYGR7PmhrsjfTex0b0c,79121
 tests/scorer/test_score_aggregation_granularity.py,sha256=h-hyFOLzwp-92hYVU7CwvlRJ8jhB4DzXCtqgI-zcoqM,13677
 tests/transformers/test_estimator_transformer_context.py,sha256=5GOHbuWCWBMFwwOTJOuD4oNDsv-qDR0OxNZYGGuMdag,1819
 tests/transformers/test_net_over_predicted.py,sha256=vh7O1iRRPf4vcW9aLhOMAOyatfM5ZnLsQBKNAYsR3SU,3363
@@ -105,7 +107,7 @@ tests/transformers/test_other_transformer.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRk
 tests/transformers/test_predictor_transformer.py,sha256=N1aBYLjN3ldpYZLwjih_gTFYSMitrZu-PNK78W6RHaQ,6877
 tests/transformers/test_simple_transformer.py,sha256=wWR0qjLb_uS4HXrJgGdiqugOY1X7kwd1_OPS02IT2b8,4676
 tests/transformers/test_team_ratio_predictor.py,sha256=fOUP_JvNJi-3kom3ZOs1EdG0I6Z8hpLpYKNHu1eWtOw,8562
-spforge-0.8.8.dist-info/METADATA,sha256=fO2JHqnnqOrjkWZ1Zh4rgYg58bi4YzxhSa8I72wqDs4,20047
-spforge-0.8.8.dist-info/WHEEL,sha256=qELbo2s1Yzl39ZmrAibXA2jjPLUYfnVhUNTlyF1rq0Y,92
-spforge-0.8.8.dist-info/top_level.txt,sha256=6UW2M5a7WKOeaAi900qQmRKNj5-HZzE8-eUD9Y9LTq0,23
-spforge-0.8.8.dist-info/RECORD,,
+spforge-0.8.19.dist-info/METADATA,sha256=4q1uKNTzmI9bwRwMJQaM0N6SAaC1RDembf_Gfbm2-mw,20048
+spforge-0.8.19.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+spforge-0.8.19.dist-info/top_level.txt,sha256=6UW2M5a7WKOeaAi900qQmRKNj5-HZzE8-eUD9Y9LTq0,23
+spforge-0.8.19.dist-info/RECORD,,

{spforge-0.8.8.dist-info → spforge-0.8.19.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.10.1)
+Generator: setuptools (80.10.2)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

tests/end_to_end/test_nba_player_ratings_hyperparameter_tuning.py

@@ -96,12 +96,8 @@ def test_nba_player_ratings_hyperparameter_tuning__workflow_completes(
         "confidence_value_denom",
         "confidence_max_sum",
         "use_off_def_split",
-        "performance_predictor",
-        "start_team_weight",
         "start_league_quantile",
         "start_min_count_for_percentiles",
-        "start_min_match_count_team_rating",
-        "start_team_rating_subtract",
     }
     assert set(result.best_params.keys()) == expected_params
 

tests/hyperparameter_tuning/test_rating_tuner.py

@@ -454,3 +454,160 @@ def test_param_spec__categorical_requires_choices():
 
     with pytest.raises(ValueError, match="requires choices"):
         spec.suggest(trial, "test_param")
+
+
+def test_param_ranges__overrides_bounds(
+    player_rating_generator, cross_validator, scorer, sample_player_df_pd
+):
+    """Test that param_ranges overrides low/high bounds while preserving param_type."""
+    tuner = RatingHyperparameterTuner(
+        rating_generator=player_rating_generator,
+        cross_validator=cross_validator,
+        scorer=scorer,
+        direction="minimize",
+        param_ranges={
+            "confidence_weight": (0.2, 0.3),
+        },
+        n_trials=3,
+        show_progress_bar=False,
+    )
+
+    result = tuner.optimize(sample_player_df_pd)
+
+    assert "confidence_weight" in result.best_params
+    assert 0.2 <= result.best_params["confidence_weight"] <= 0.3
+
+
+def test_exclude_params__removes_from_search(
+    player_rating_generator, cross_validator, scorer, sample_player_df_pd
+):
+    """Test that exclude_params removes parameters from search space."""
+    tuner = RatingHyperparameterTuner(
+        rating_generator=player_rating_generator,
+        cross_validator=cross_validator,
+        scorer=scorer,
+        direction="minimize",
+        exclude_params=["use_off_def_split", "confidence_weight"],
+        n_trials=3,
+        show_progress_bar=False,
+    )
+
+    result = tuner.optimize(sample_player_df_pd)
+
+    assert "use_off_def_split" not in result.best_params
+    assert "confidence_weight" not in result.best_params
+    assert "rating_change_multiplier_offense" in result.best_params
+
+
+def test_fixed_params__applies_values_without_tuning(
+    player_rating_generator, cross_validator, scorer, sample_player_df_pd
+):
+    """Test that fixed_params sets values without including in search space."""
+    tuner = RatingHyperparameterTuner(
+        rating_generator=player_rating_generator,
+        cross_validator=cross_validator,
+        scorer=scorer,
+        direction="minimize",
+        fixed_params={"use_off_def_split": False},
+        n_trials=3,
+        show_progress_bar=False,
+    )
+
+    result = tuner.optimize(sample_player_df_pd)
+
+    assert "use_off_def_split" not in result.best_params
+
+
+def test_param_ranges__unknown_param_raises_error(
+    player_rating_generator, cross_validator, scorer, sample_player_df_pd
+):
+    """Test that param_ranges with unknown param raises ValueError."""
+    tuner = RatingHyperparameterTuner(
+        rating_generator=player_rating_generator,
+        cross_validator=cross_validator,
+        scorer=scorer,
+        direction="minimize",
+        param_ranges={"nonexistent_param": (0.0, 1.0)},
+        n_trials=3,
+        show_progress_bar=False,
+    )
+
+    with pytest.raises(ValueError, match="unknown parameter"):
+        tuner.optimize(sample_player_df_pd)
+
+
+def test_param_ranges__non_numeric_param_raises_error(
+    player_rating_generator, cross_validator, scorer, sample_player_df_pd
+):
+    """Test that param_ranges on non-float/int param raises ValueError."""
+    tuner = RatingHyperparameterTuner(
+        rating_generator=player_rating_generator,
+        cross_validator=cross_validator,
+        scorer=scorer,
+        direction="minimize",
+        param_ranges={"use_off_def_split": (0, 1)},
+        n_trials=3,
+        show_progress_bar=False,
+    )
+
+    with pytest.raises(ValueError, match="can only override float/int"):
+        tuner.optimize(sample_player_df_pd)
+
+
+def test_combined_api__param_ranges_exclude_fixed(
+    player_rating_generator, cross_validator, scorer, sample_player_df_pd
+):
+    """Test using param_ranges, exclude_params, and fixed_params together."""
+    tuner = RatingHyperparameterTuner(
+        rating_generator=player_rating_generator,
+        cross_validator=cross_validator,
+        scorer=scorer,
+        direction="minimize",
+        param_ranges={
+            "confidence_weight": (0.2, 1.0),
+            "rating_change_multiplier_offense": (10.0, 150.0),
+        },
+        exclude_params=["start_league_quantile"],
+        fixed_params={"use_off_def_split": False},
+        n_trials=3,
+        show_progress_bar=False,
+    )
+
+    result = tuner.optimize(sample_player_df_pd)
+
+    assert 0.2 <= result.best_params["confidence_weight"] <= 1.0
+    assert 10.0 <= result.best_params["rating_change_multiplier_offense"] <= 150.0
+    assert "start_league_quantile" not in result.best_params
+    assert "use_off_def_split" not in result.best_params
+
+
+def test_default_search_space__excludes_performance_predictor_and_team_start(
+    player_rating_generator,
+):
+    """Test that performance_predictor and team start params are not in default search space."""
+    from spforge.hyperparameter_tuning._default_search_spaces import (
+        get_default_search_space,
+    )
+
+    defaults = get_default_search_space(player_rating_generator)
+
+    assert "performance_predictor" not in defaults
+    assert "start_team_rating_subtract" not in defaults
+    assert "start_team_weight" not in defaults
+    assert "start_min_match_count_team_rating" not in defaults
+
+
+def test_full_player_rating_search_space__includes_all_params():
+    """Test that full search space includes performance_predictor and team start params."""
+    from spforge.hyperparameter_tuning._default_search_spaces import (
+        get_full_player_rating_search_space,
+    )
+
+    full = get_full_player_rating_search_space()
+
+    assert "performance_predictor" in full
+    assert "start_team_rating_subtract" in full
+    assert "start_team_weight" in full
+    assert "start_min_match_count_team_rating" in full
+    assert "rating_change_multiplier_offense" in full
+    assert "confidence_weight" in full

tests/performance_transformers/test_performance_manager.py

@@ -56,6 +56,21 @@ def test_performance_weights_manager_basic_flow(sample_data):
     assert output_df["weighted_performance"].iloc[0] == pytest.approx(0.6)
 
 
+def test_performance_weights_manager_keeps_mean_when_weights_not_normalized():
+    df = pd.DataFrame(
+        {
+            "feat_a": [0.0, 1.0, 2.0, 3.0],
+            "feat_b": [3.0, 2.0, 1.0, 0.0],
+        }
+    )
+    weights = [ColumnWeight(name="feat_a", weight=0.9), ColumnWeight(name="feat_b", weight=0.5)]
+
+    manager = PerformanceWeightsManager(weights=weights, transformer_names=["min_max"], prefix="")
+    output_df = nw.from_native(manager.fit_transform(df)).to_pandas()
+
+    assert output_df["weighted_performance"].mean() == pytest.approx(0.5, abs=1e-6)
+
+
 def test_lower_is_better_logic():
     df = pd.DataFrame({"feat_a": [1.0, 0.0]})
     weights = [ColumnWeight(name="feat_a", weight=1.0, lower_is_better=True)]

tests/ratings/test_player_rating_generator.py

@@ -551,6 +551,63 @@ def test_fit_transform_scales_participation_weight_by_fit_quantile(base_cn):
     assert p1_change / p2_change == pytest.approx(expected_ratio, rel=1e-6)
 
 
+def test_fit_transform_auto_scales_participation_weight_when_out_of_bounds(base_cn):
+    """Automatically enable scaling when participation weights exceed [0, 1]."""
+    df = pl.DataFrame(
+        {
+            "pid": ["P1", "P2", "O1", "O2"],
+            "tid": ["T1", "T1", "T2", "T2"],
+            "mid": ["M1", "M1", "M1", "M1"],
+            "dt": ["2024-01-01"] * 4,
+            "perf": [0.9, 0.9, 0.1, 0.1],
+            "pw": [10.0, 20.0, 10.0, 10.0],
+        }
+    )
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=base_cn,
+        auto_scale_performance=True,
+        start_harcoded_start_rating=1000.0,
+    )
+    gen.fit_transform(df)
+
+    start_rating = 1000.0
+    p1_change = gen._player_off_ratings["P1"].rating_value - start_rating
+    p2_change = gen._player_off_ratings["P2"].rating_value - start_rating
+
+    q = df["pw"].quantile(0.99, "linear")
+    expected_ratio = min(1.0, 10.0 / q) / min(1.0, 20.0 / q)
+
+    assert gen.scale_participation_weights is True
+    assert p1_change / p2_change == pytest.approx(expected_ratio, rel=1e-6)
+
+
+def test_fit_transform_auto_scale_logs_warning_when_out_of_bounds(base_cn, caplog):
+    """Auto-scaling should emit a warning when participation weights exceed [0, 1]."""
+    df = pl.DataFrame(
+        {
+            "pid": ["P1", "P2", "O1", "O2"],
+            "tid": ["T1", "T1", "T2", "T2"],
+            "mid": ["M1", "M1", "M1", "M1"],
+            "dt": ["2024-01-01"] * 4,
+            "perf": [0.9, 0.9, 0.1, 0.1],
+            "pw": [10.0, 20.0, 10.0, 10.0],
+        }
+    )
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=base_cn,
+        auto_scale_performance=True,
+        start_harcoded_start_rating=1000.0,
+    )
+    with caplog.at_level("WARNING"):
+        gen.fit_transform(df)
+
+    assert any(
+        "Auto-scaling participation weights" in record.message for record in caplog.records
+    )
+
+
 def test_future_transform_scales_projected_participation_weight_by_fit_quantile():
     """Future projected participation weights should scale with fit quantile and be clipped."""
     cn = ColumnNames(
@@ -1689,3 +1746,73 @@ def test_fit_transform__player_rating_difference_from_team_projected_feature(bas
     for row in result.iter_rows(named=True):
         expected = row[player_col] - row[team_col]
         assert row[diff_col] == pytest.approx(expected, rel=1e-9)
+
+
+def test_fit_transform__start_league_quantile_uses_existing_player_ratings(base_cn):
+    """
+    Bug reproduction: start_league_quantile should use percentile of existing player
+    ratings for new players, but update_players_to_leagues is never called so
+    _league_player_ratings stays empty and all new players get default rating.
+
+    Expected: New player P_NEW should start at 5th percentile of existing ratings (~920)
+    Actual: New player starts at default 1000 because _league_player_ratings is empty
+    """
+    import numpy as np
+
+    num_existing_players = 60
+    player_ids = [f"P{i}" for i in range(num_existing_players)]
+    team_ids = [f"T{i % 2 + 1}" for i in range(num_existing_players)]
+
+    df1 = pl.DataFrame(
+        {
+            "pid": player_ids,
+            "tid": team_ids,
+            "mid": ["M1"] * num_existing_players,
+            "dt": ["2024-01-01"] * num_existing_players,
+            "perf": [0.3 + (i % 10) * 0.07 for i in range(num_existing_players)],
+            "pw": [1.0] * num_existing_players,
+        }
+    )
+
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=base_cn,
+        auto_scale_performance=True,
+        start_league_quantile=0.05,
+        start_min_count_for_percentiles=50,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    gen.fit_transform(df1)
+
+    existing_ratings = [
+        gen._player_off_ratings[pid].rating_value for pid in player_ids
+    ]
+    expected_quantile_rating = np.percentile(existing_ratings, 5)
+
+    srg = gen.start_rating_generator
+    assert len(srg._league_player_ratings.get(None, [])) >= 50, (
+        f"Expected _league_player_ratings to have >=50 entries but got "
+        f"{len(srg._league_player_ratings.get(None, []))}. "
+        "update_players_to_leagues is never called."
+    )
+
+    df2 = pl.DataFrame(
+        {
+            "pid": ["P_NEW", "P0"],
+            "tid": ["T1", "T2"],
+            "mid": ["M2", "M2"],
+            "dt": ["2024-01-02", "2024-01-02"],
+            "pw": [1.0, 1.0],
+        }
+    )
+    result = gen.future_transform(df2)
+
+    new_player_start_rating = result.filter(pl.col("pid") == "P_NEW")[
+        "player_off_rating_perf"
+    ][0]
+
+    assert new_player_start_rating == pytest.approx(expected_quantile_rating, rel=0.1), (
+        f"New player should start at 5th percentile ({expected_quantile_rating:.1f}) "
+        f"but got {new_player_start_rating:.1f}. "
+        "start_league_quantile has no effect because update_players_to_leagues is never called."
+    )

tests/ratings/test_player_rating_no_mutation.py

@@ -0,0 +1,214 @@
+"""Tests to ensure PlayerRatingGenerator does not mutate input columns."""
+
+import polars as pl
+import pytest
+
+from spforge import ColumnNames
+from spforge.ratings import PlayerRatingGenerator, RatingKnownFeatures
+
+
+@pytest.fixture
+def cn_with_projected():
+    """ColumnNames with both participation_weight and projected_participation_weight."""
+    return ColumnNames(
+        player_id="pid",
+        team_id="tid",
+        match_id="mid",
+        start_date="dt",
+        update_match_id="mid",
+        participation_weight="minutes",
+        projected_participation_weight="minutes_prediction",
+    )
+
+
+@pytest.fixture
+def fit_df():
+    """Training data with minutes > 1 (will trigger auto-scaling)."""
+    return pl.DataFrame(
+        {
+            "pid": ["P1", "P2", "P3", "P4"],
+            "tid": ["T1", "T1", "T2", "T2"],
+            "mid": ["M1", "M1", "M1", "M1"],
+            "dt": ["2024-01-01"] * 4,
+            "perf": [0.6, 0.4, 0.7, 0.3],
+            "minutes": [30.0, 25.0, 32.0, 28.0],
+            "minutes_prediction": [28.0, 24.0, 30.0, 26.0],
+        }
+    )
+
+
+@pytest.fixture
+def future_df():
+    """Future prediction data with minutes > 1 (will trigger auto-scaling)."""
+    return pl.DataFrame(
+        {
+            "pid": ["P1", "P2", "P3", "P4"],
+            "tid": ["T1", "T1", "T2", "T2"],
+            "mid": ["M2", "M2", "M2", "M2"],
+            "dt": ["2024-01-02"] * 4,
+            "minutes": [30.0, 25.0, 32.0, 28.0],
+            "minutes_prediction": [28.0, 24.0, 30.0, 26.0],
+        }
+    )
+
+
+def test_fit_transform_does_not_mutate_participation_weight(cn_with_projected, fit_df):
+    """fit_transform should not modify the participation_weight column values."""
+    # Join result with original to compare values by player_id
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    result = gen.fit_transform(fit_df)
+
+    # Check that each player's minutes value is preserved
+    original_by_player = dict(zip(fit_df["pid"].to_list(), fit_df["minutes"].to_list()))
+    result_by_player = dict(zip(result["pid"].to_list(), result["minutes"].to_list()))
+
+    for pid, original_val in original_by_player.items():
+        result_val = result_by_player[pid]
+        assert result_val == original_val, (
+            f"participation_weight for player {pid} was mutated. "
+            f"Expected {original_val}, got {result_val}"
+        )
+
+
+def test_fit_transform_does_not_mutate_projected_participation_weight(cn_with_projected, fit_df):
+    """fit_transform should not modify the projected_participation_weight column values."""
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    result = gen.fit_transform(fit_df)
+
+    # Check that each player's minutes_prediction value is preserved
+    original_by_player = dict(zip(fit_df["pid"].to_list(), fit_df["minutes_prediction"].to_list()))
+    result_by_player = dict(zip(result["pid"].to_list(), result["minutes_prediction"].to_list()))
+
+    for pid, original_val in original_by_player.items():
+        result_val = result_by_player[pid]
+        assert result_val == original_val, (
+            f"projected_participation_weight for player {pid} was mutated. "
+            f"Expected {original_val}, got {result_val}"
+        )
+
+
+def test_transform_does_not_mutate_participation_weight(cn_with_projected, fit_df, future_df):
+    """transform should not modify the participation_weight column values."""
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    gen.fit_transform(fit_df)
+
+    result = gen.transform(future_df)
+
+    # Check that each player's minutes value is preserved
+    original_by_player = dict(zip(future_df["pid"].to_list(), future_df["minutes"].to_list()))
+    result_by_player = dict(zip(result["pid"].to_list(), result["minutes"].to_list()))
+
+    for pid, original_val in original_by_player.items():
+        result_val = result_by_player[pid]
+        assert result_val == original_val, (
+            f"participation_weight for player {pid} was mutated during transform. "
+            f"Expected {original_val}, got {result_val}"
+        )
+
+
+def test_transform_does_not_mutate_projected_participation_weight(cn_with_projected, fit_df, future_df):
+    """transform should not modify the projected_participation_weight column values."""
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    gen.fit_transform(fit_df)
+
+    result = gen.transform(future_df)
+
+    # Check that each player's minutes_prediction value is preserved
+    original_by_player = dict(zip(future_df["pid"].to_list(), future_df["minutes_prediction"].to_list()))
+    result_by_player = dict(zip(result["pid"].to_list(), result["minutes_prediction"].to_list()))
+
+    for pid, original_val in original_by_player.items():
+        result_val = result_by_player[pid]
+        assert result_val == original_val, (
+            f"projected_participation_weight for player {pid} was mutated during transform. "
+            f"Expected {original_val}, got {result_val}"
+        )
+
+
+def test_future_transform_does_not_mutate_participation_weight(cn_with_projected, fit_df, future_df):
+    """future_transform should not modify the participation_weight column values."""
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    gen.fit_transform(fit_df)
+
+    original_minutes = future_df["minutes"].to_list()
+    result = gen.future_transform(future_df)
+
+    # The minutes column should have the same values as before
+    result_minutes = result["minutes"].to_list()
+    assert result_minutes == original_minutes, (
+        f"participation_weight column was mutated during future_transform. "
+        f"Expected {original_minutes}, got {result_minutes}"
+    )
+
+
+def test_future_transform_does_not_mutate_projected_participation_weight(cn_with_projected, fit_df, future_df):
+    """future_transform should not modify the projected_participation_weight column values."""
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    gen.fit_transform(fit_df)
+
+    original_minutes_pred = future_df["minutes_prediction"].to_list()
+    result = gen.future_transform(future_df)
+
+    # The minutes_prediction column should have the same values as before
+    result_minutes_pred = result["minutes_prediction"].to_list()
+    assert result_minutes_pred == original_minutes_pred, (
+        f"projected_participation_weight column was mutated during future_transform. "
+        f"Expected {original_minutes_pred}, got {result_minutes_pred}"
+    )
+
+
+def test_multiple_transforms_do_not_compound_scaling(cn_with_projected, fit_df, future_df):
+    """Multiple transform calls should not compound the scaling effect."""
+    gen = PlayerRatingGenerator(
+        performance_column="perf",
+        column_names=cn_with_projected,
+        auto_scale_performance=True,
+        features_out=[RatingKnownFeatures.PLAYER_OFF_RATING],
+    )
+    gen.fit_transform(fit_df)
+
+    # Call transform multiple times
+    result1 = gen.transform(future_df)
+    result2 = gen.transform(result1)
+    result3 = gen.transform(result2)
+
+    # After 3 transforms, each player's values should still be the same as original
+    original_by_player = dict(zip(future_df["pid"].to_list(), future_df["minutes_prediction"].to_list()))
+    final_by_player = dict(zip(result3["pid"].to_list(), result3["minutes_prediction"].to_list()))
+
+    for pid, original_val in original_by_player.items():
+        final_val = final_by_player[pid]
+        assert final_val == original_val, (
+            f"Multiple transforms compounded the scaling for player {pid}. "
+            f"Expected {original_val}, got {final_val}"
+        )