numerai-tools 0.4.0.dev2__tar.gz → 0.4.2.dev1__tar.gz

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: numerai_tools
-Version: 0.4.0.dev2
+Version: 0.4.2.dev1
 Summary: A collection of open-source tools to help interact with Numerai, model data, and automate submissions.
 Home-page: https://github.com/numerai/numerai-tools
 Maintainer: Numerai

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/numerai_tools/scoring.py

@@ -1,6 +1,5 @@
 from typing import List, Tuple, Union, Optional, TypeVar
 
-import torch
 import numpy as np
 import pandas as pd  # type: ignore
 from scipy import stats  # type: ignore
@@ -582,10 +581,11 @@ def meta_portfolio_contribution(
     targets: pd.Series,
 ) -> pd.Series:
     """Calculates the "meta portfolio" score:
-        - rank, normalize, and power the signal
-        - convert signal into neutralized weights
+        - rank, normalize, and power each signal
+        - convert each signal into neutralized weights
+        - generate the stake-weighted portfolio
+        - calculate the gradient of the portfolio w.r.t. the stakes
         - multiplying the weights by the targets
-
     Arguments:
         predictions: pd.DataFrame - the predictions to evaluate
         stakes: pd.Series - the stakes to use as weights
@@ -599,22 +599,27 @@ def meta_portfolio_contribution(
     predictions, neutralizers, sample_weights, targets = filter_sort_index_many(
         [predictions, neutralizers, sample_weights, targets]
     )
-
     stake_weights = weight_normalize(stakes.fillna(0))
     assert np.isclose(stake_weights.sum(), 1), "Stakes must sum to 1"
-
     weights = tie_kept_rank__gaussianize__pow_1_5(predictions).apply(
         lambda s_prime: generate_neutralized_weights(
             s_prime, neutralizers, sample_weights
         )
     )
-
-    w = torch.tensor(weights[stakes.index].values)
-    s = torch.tensor(stake_weights.values, requires_grad=True)
-    t = torch.tensor(targets.values)
+    w = weights[stakes.index].values
+    s = stake_weights.values
+    t = targets.values
     swp = w @ s
     swp = swp - swp.mean()
-    swp = swp / swp.abs().sum()
-    alpha = swp @ t
-    mpc = torch.autograd.grad(alpha, s)[0].numpy()
+    swp_abs_sum = np.sum(np.abs(swp))
+    swp_sign = np.sign(swp)
+    alpha_unnormalized_swp_grad = (
+        1
+        / np.power(swp_abs_sum, 2)
+        * (swp_abs_sum * t - swp_sign * np.dot(swp, t)).reshape(-1, 1)
+    )
+    zero_mean_jac_vec_prod = (
+        alpha_unnormalized_swp_grad - alpha_unnormalized_swp_grad.mean()
+    )
+    mpc = (w.T @ zero_mean_jac_vec_prod).squeeze()
     return pd.Series(mpc, index=stakes.index)

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/numerai_tools/signals.py

@@ -19,7 +19,8 @@ def churn(
 
     For 2 given series with overlapping indices, churn is 1 - Spearman Correlation.
     If top_bottom is provided, the churn is calculated as the average of the % of
-    tickers that stay in the top and bottom predictions.
+    tickers that stay in the top and bottom predictions. This is only relevant when
+    the series are rank signals and not portfolio weights.
 
     Arguments:
         s1: pd.Series - the first series to compare
@@ -43,3 +44,29 @@ def churn(
     assert s1.std() > 0, "s1 must have non-zero standard deviation"
     assert s2.std() > 0, "s2 must have non-zero standard deviation"
     return 1 - spearman_correlation(s1, s2)
+
+
+def turnover(
+    s1: pd.Series,
+    s2: pd.Series,
+):
+    """Calculate the turnover between two series. Turnover is the total change in weights between
+    the two series divided by 2.
+
+    For 2 given series with overlapping indices, join the series on index, fill nans with zeroes
+    and calculate turnover as the absolute total difference between the two series divided by 2.
+    This is only relevant when the series are portfolio weights and not rank signals.
+
+    Arguments:
+        s1: pd.Series - the first series to compare
+        s2: pd.Series - the second series to compare
+        top_bottom: Optional[int] - the number of top and bottom predictions to use
+                                    when calculating the correlation. Results in
+                                    2*top_bottom predictions.
+
+    Returns:
+        float - the turnover between the two series
+    """
+    s1, s2 = filter_sort_index(s1, s2)
+    turnover = (s1 - s2).abs().sum() / 2
+    return turnover

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/numerai_tools.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: numerai-tools
-Version: 0.4.0.dev2
+Version: 0.4.2.dev1
 Summary: A collection of open-source tools to help interact with Numerai, model data, and automate submissions.
 Home-page: https://github.com/numerai/numerai-tools
 Maintainer: Numerai

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/numerai_tools.egg-info/requires.txt

@@ -2,4 +2,3 @@ pandas<3.0.0,>=2.2.2
 numpy<3.0.0,>=2.0.0
 scipy<2.0.0,>=1.13.0
 scikit-learn<2.0.0,>=1.5.0
-torch

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup
 from setuptools import find_packages
 
-VERSION = "0.4.0.dev2"
+VERSION = "0.4.2.dev1"
 
 
 def load(path):
@@ -43,6 +43,5 @@ if __name__ == "__main__":
             "numpy>=2.0.0,<3.0.0",
             "scipy>=1.13.0,<2.0.0",
             "scikit-learn>=1.5.0,<2.0.0",
-            "torch",
         ],
     )

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/tests/test_scoring.py

@@ -318,8 +318,8 @@ class TestScoring(unittest.TestCase):
                 [5, 1],
             ]
         )
-        v = pd.Series([3, 2, 1, 2, 3]).T
-        t = pd.Series([1, 2, 3, 2, 1]).T
+        v = pd.Series([1, 0.5, 1, 0.5, 1]).T
+        t = pd.Series([1, 0, 1, 0, 1]).T
         score = alpha(s, N, v, t)
         np.testing.assert_allclose(score, 0.0, atol=1e-14, rtol=1e-14)
 

{numerai_tools-0.4.0.dev2 → numerai_tools-0.4.2.dev1}/tests/test_signals.py

@@ -3,7 +3,7 @@ import unittest
 import numpy as np
 import pandas as pd  # type: ignore
 
-from numerai_tools.signals import churn
+from numerai_tools.signals import churn, turnover
 
 
 class TestSignals(unittest.TestCase):
@@ -39,6 +39,13 @@ class TestSignals(unittest.TestCase):
         tmp = churn(self.up, self.constant, top_bottom=2)
         assert np.isclose(tmp, 0), tmp
 
+    def test_turnover(self):
+        assert np.isclose(turnover(self.up, self.up), 0)
+        assert np.isclose(turnover(self.up, self.up_down), 3)
+        assert np.isclose(turnover(self.up, self.oscillate), 4.5)
+        assert np.isclose(turnover(self.up, self.down), 6)
+        assert np.isclose(turnover(self.up, self.constant), 3.5)
+
 
 if __name__ == "__main__":
     unittest.main()