fix(Labelling): didn't forward shift forward returns previously, introduced major lookahead bias (#249)

* fix(Labelling): didn't forward shift forward returns previously, introduced major lookahead bias

* fix(Baseline): use the same config

config/presets.py

@@ -23,16 +23,16 @@ def get_default_config() -> RawConfig:
         assets=["fivemin_crypto"],
         target_asset="BTCUSDT",
         other_assets=[],
-        exogenous_data=[],
+        exogenous_data=["daily_glassnode"],
         load_non_target_asset=True,
-        own_features=["level_1"],
+        own_features=["level_2"],
         other_features=["z_score"],
-        exogenous_features=[],
+        exogenous_features=["z_score"],
         directional_models=classification_models,
         meta_models=meta_models,
         event_filter="cusum_vol",
         event_filter_multiplier=3.5,
-        remove_overlapping_events=True,
+        remove_overlapping_events=False,
         labeling="two_class",
         forecasting_horizon=10,
         transaction_costs=0.002,
@@ -74,7 +74,7 @@ def get_minimal_config() -> RawConfig:
         event_filter_multiplier=3.5,
         remove_overlapping_events=False,
         labeling="two_class",
-        forecasting_horizon=10,
+        forecasting_horizon=1,
         transaction_costs=0.002,
         save_models=True,
         ensembling_method="voting_soft",

labeling/labellers/utils.py

@@ -8,9 +8,7 @@ import numpy as np
 def create_forward_returns(series: pd.Series, period: int) -> ForwardReturnSeries:
     assert period > 0
     indexer = pd.api.indexers.FixedForwardWindowIndexer(window_size=period)
-
-    forward_returns = series.rolling(window=indexer).sum()
-    return forward_returns
+    return series.rolling(window=indexer).sum().shift(-1)
 
 
 def purge_overlapping_events(events: EventsDataFrame) -> EventsDataFrame:

run_baseline.py

@@ -1,6 +1,3 @@
-from re import S
-from typing import Optional
-
 from sklearn.model_selection import train_test_split
 
 from config.types import Config, RawConfig
@@ -65,5 +62,5 @@ def run_sklearn_training(config: Config):
 
 if __name__ == "__main__":
     run_sklearn_pipeline(
-        raw_config=get_minimal_config(),
+        raw_config=get_default_config(),
     )