numerai-tools 0.6.1.dev0__tar.gz → 0.6.1.dev1__tar.gz

{numerai_tools-0.6.1.dev0 → numerai_tools-0.6.1.dev1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: numerai-tools
-Version: 0.6.1.dev0
+Version: 0.6.1.dev1
 Summary: A collection of open-source tools to help interact with Numerai, model data, and automate submissions.
 License: MIT
 License-File: LICENSE
@@ -43,21 +43,3 @@ pip install numerai-tools
 - The `submissions.py` module provides helper functions to ensure your submissions are valid and formatted correctly. Use this in your automated prediction pipelines to ensure uploads don't fail.
 
 - The `signals.py` module provides code specific to Numerai Signals such as churn and turnover. Use this to ensure your Signals submissions are properly formatted.
-
-## Configurable neutralized weights
-
-`generate_neutralized_weights`, `alpha`, and `meta_portfolio_contribution` accept
-a `power` argument applied after ranking and gaussianization. The default is `1`.
-Pass `power=1.5` when reproducing historical Alpha or MPC definitions:
-
-```python
-from numerai_tools.scoring import generate_neutralized_weights
-
-weights = generate_neutralized_weights(
-    predictions,
-    neutralizers,
-    sample_weights,
-    power=1.5,
-)
-```
-

{numerai_tools-0.6.1.dev0 → numerai_tools-0.6.1.dev1}/README.md

@@ -12,21 +12,4 @@ pip install numerai-tools
 
 - The `submissions.py` module provides helper functions to ensure your submissions are valid and formatted correctly. Use this in your automated prediction pipelines to ensure uploads don't fail.
 
-- The `signals.py` module provides code specific to Numerai Signals such as churn and turnover. Use this to ensure your Signals submissions are properly formatted.
-
-## Configurable neutralized weights
-
-`generate_neutralized_weights`, `alpha`, and `meta_portfolio_contribution` accept
-a `power` argument applied after ranking and gaussianization. The default is `1`.
-Pass `power=1.5` when reproducing historical Alpha or MPC definitions:
-
-```python
-from numerai_tools.scoring import generate_neutralized_weights
-
-weights = generate_neutralized_weights(
-    predictions,
-    neutralizers,
-    sample_weights,
-    power=1.5,
-)
-```
+- The `signals.py` module provides code specific to Numerai Signals such as churn and turnover. Use this to ensure your Signals submissions are properly formatted.

{numerai_tools-0.6.1.dev0 → numerai_tools-0.6.1.dev1}/numerai_tools/scoring.py

@@ -161,7 +161,7 @@ def min_max_normalize(s: pd.Series) -> pd.Series:
 
 def variance_normalize(df: pd.DataFrame) -> pd.DataFrame:
     """Scale a df such that all columns have std == 1."""
-    return cast(pd.DataFrame, df / np.std(df, axis=0))
+    return df / np.std(df, axis=0)
 
 
 def weight_normalize(s: S1) -> S1:
@@ -446,14 +446,7 @@ def tie_kept_rank__gaussianize__pow_1_5(df: pd.DataFrame) -> pd.DataFrame:
     Returns:
         pd.DataFrame - the resulting data after applying the 3 functions
     """
-    return _tie_kept_rank__gaussianize__power(df, 1.5)
-
-
-def _tie_kept_rank__gaussianize__power(
-    df: pd.DataFrame, power_value: float
-) -> pd.DataFrame:
-    """Tie-kept rank, gaussianize, then apply a configurable power."""
-    return power(gaussian(tie_kept_rank(df)), power_value)
+    return power(gaussian(tie_kept_rank(df)), 1.5)
 
 
 def tie_kept_rank__gaussianize__neutralize__variance_normalize(
@@ -575,21 +568,11 @@ def generate_neutralized_weights(
     neutralizers: pd.DataFrame,
     sample_weights: pd.Series,
     center_and_normalize: bool = False,
-    power: float = 1,
 ) -> pd.DataFrame:
-    """Convert predictions into neutralized, sample-weighted portfolio weights.
-
-    Arguments:
-        predictions: prediction columns to transform
-        neutralizers: factors to neutralize the transformed predictions against
-        sample_weights: per-row weights applied during and after neutralization
-        center_and_normalize: center and normalize the resulting portfolio weights
-        power: exponent applied after ranking and gaussianization; defaults to 1
-    """
     assert not predictions.isna().any().any(), "Predictions contain NaNs"
     assert not neutralizers.isna().any().any(), "Normalization factors contain NaNs"
     assert not sample_weights.isna().any(), "Weights contain NaNs"
-    ranked_predictions = _tie_kept_rank__gaussianize__power(predictions, power)
+    ranked_predictions = tie_kept_rank__gaussianize__pow_1_5(predictions)
     ranked_predictions, neutralizers, sample_weights = filter_sort_index_many(
         [ranked_predictions, neutralizers, sample_weights]
     )
@@ -609,7 +592,6 @@ def alpha(
     neutralizers: pd.DataFrame,
     sample_weights: pd.Series,
     targets: pd.Series,
-    power: float = 1,
 ) -> pd.Series:
     """Calculates the "alpha" score:
         - rank, normalize, and power the signal
@@ -621,13 +603,10 @@ def alpha(
         neutralizers: pd.DataFrame - the neutralization columns
         sample_weights: pd.Series - the universe sampling weights
         targets: pd.Series - the live targets to evaluate against
-        power: float - exponent applied after ranking and gaussianization
     """
     targets = center(targets)
     predictions, targets = filter_sort_index(predictions, targets)
-    weights = generate_neutralized_weights(
-        predictions, neutralizers, sample_weights, power=power
-    )
+    weights = generate_neutralized_weights(predictions, neutralizers, sample_weights)
     alpha_scores = weights.apply(lambda w: w @ targets) / len(targets)
     return alpha_scores
 
@@ -638,7 +617,6 @@ def meta_portfolio_contribution(
     neutralizers: pd.DataFrame,
     sample_weights: pd.Series,
     targets: pd.Series,
-    power: float = 1,
 ) -> pd.Series:
     """Calculates the "meta portfolio" gradient w.r.t. stakes:
     - rank, normalize, and power each signal
@@ -654,7 +632,6 @@ def meta_portfolio_contribution(
         neutralizers: pd.DataFrame - the neutralization columns
         sample_weights: pd.Series - the universe sampling weights
         targets: pd.Series - the live targets to evaluate against
-        power: float - exponent applied after ranking and gaussianization
     """
     # Align predictions and targets on the same index / universe
     predictions, targets = filter_sort_index(predictions, targets)
@@ -667,9 +644,7 @@ def meta_portfolio_contribution(
     assert np.isclose(stake_weights.sum(), 1), "Stakes must sum to 1"
 
     # Generate neutralized weights W(predictions, neutralizers, sample_weights)
-    weights = generate_neutralized_weights(
-        predictions, neutralizers, sample_weights, power=power
-    )
+    weights = generate_neutralized_weights(predictions, neutralizers, sample_weights)
 
     # Extract aligned matrices/vectors
     w = cast(np.ndarray, weights[stakes.index].values)  # W ∈ R^{N×K}

{numerai_tools-0.6.1.dev0 → numerai_tools-0.6.1.dev1}/numerai_tools/signals.py

@@ -81,7 +81,6 @@ def calculate_max_churn_and_turnover(
     prev_subs: dict[str, pd.Series],
     prev_neutralizers: dict[str, pd.DataFrame],
     prev_sample_weights: dict[str, pd.Series],
-    power: float = 1.5,
 ) -> Tuple[float, float]:
     """Calculate the maximum churn and turnover of the current submission with respect to previous submissions.
     This function iterates over previous submissions and calculates churn and turnover for each submission
@@ -125,10 +124,6 @@ def calculate_max_churn_and_turnover(
               Series is indexed on the same type of tickers/ids as the current submission.
               We expect each of these to cover the full universe of their respective eras.
 
-        power: float
-            - exponent applied after ranking and gaussianization. Defaults to 1.5
-              to preserve the legacy Signals turnover definition.
-
     Returns:
         prev_week_max_churn -- the maximum churn from previous submissions
         prev_week_max_turnover -- the maximum turnover from previous submissions
@@ -157,7 +152,6 @@ def calculate_max_churn_and_turnover(
         curr_neutralizer,
         curr_sample_weight,
         center_and_normalize=True,
-        power=power,
     )[curr_sub.name]
     for datestamp in prev_subs:
         prev_sub = prev_subs[datestamp]
@@ -187,7 +181,6 @@ def calculate_max_churn_and_turnover(
             prev_neutralizer,
             prev_sample_weight,
             center_and_normalize=True,
-            power=power,
         )[prev_sub.name]
         try:
             churn_val = abs(churn(curr_sub, prev_sub))

{numerai_tools-0.6.1.dev0 → numerai_tools-0.6.1.dev1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "numerai-tools"
-version = "0.6.1.dev0"
+version = "0.6.1.dev1"
 description = "A collection of open-source tools to help interact with Numerai, model data, and automate submissions."
 authors = [
     {name = "Numerai Engineering",email = "engineering@numer.ai"}