ENH: add ability to get daily multi-period forward returns rather than only cumulative

Author: twiecki

alphalens/tests/test_utils.py

@@ -81,6 +81,23 @@ def test_compute_forward_returns(self):
 
         assert_frame_equal(fp, expected)
 
+    def test_compute_forward_returns_non_cum(self):
+        dr = date_range(start='2015-1-1', end='2015-1-3')
+        prices = DataFrame(index=dr, columns=['A', 'B'],
+                           data=[[1, 1], [1, 2], [2, 1]])
+        factor = prices.stack()
+
+        fp = compute_forward_returns(factor, prices, periods=[1, 2],
+                                     cumulative_returns=False)
+
+        ix = MultiIndex.from_product([dr, ['A', 'B']],
+                                     names=['date', 'asset'])
+        expected = DataFrame(index=ix, columns=['1D', '2D'])
+        expected['1D'] = [0., 1., 1., -0.5, nan, nan]
+        expected['2D'] = [1., -0.5, nan, nan, nan, nan]
+
+        assert_frame_equal(fp, expected)
+
     @parameterized.expand([(factor_data, 4, None, False, False,
                             [1, 2, 3, 4, 4, 3, 2, 1]),
                            (factor_data, 2, None, False, False,

alphalens/utils.py

@@ -215,7 +215,8 @@ def infer_trading_calendar(factor_idx, prices_idx):
 def compute_forward_returns(factor,
                             prices,
                             periods=(1, 5, 10),
-                            filter_zscore=None):
+                            filter_zscore=None,
+                            cumulative_returns=True):
     """
     Finds the N period forward returns (as percent change) for each asset
     provided.
@@ -240,6 +241,10 @@ def compute_forward_returns(factor,
         Sets forward returns greater than X standard deviations
         from the the mean to nan. Set it to 'None' to avoid filtering.
         Caution: this outlier filtering incorporates lookahead bias.
+    cumulative_returns : bool, optional
+        If True, forward returns columns will contain cumulative returns.
+        Setting this to False is useful if you want to analyze how predictive
+        a factor is for a single forward day.
 
     Returns
     -------
@@ -278,8 +283,13 @@ def compute_forward_returns(factor,
     column_list = []
 
     for period in sorted(periods):
+        if cumulative_returns:
+            returns = prices.pct_change(period)
+        else:
+            returns = prices.pct_change()
+
         forward_returns = \
-            prices.pct_change(period).shift(-period).reindex(factor_dateindex)
+            returns.shift(-period).reindex(factor_dateindex)
 
         if filter_zscore is not None:
             mask = abs(
@@ -631,7 +641,8 @@ def get_clean_factor_and_forward_returns(factor,
                                          filter_zscore=20,
                                          groupby_labels=None,
                                          max_loss=0.35,
-                                         zero_aware=False):
+                                         zero_aware=False,
+                                         cumulative_returns=True):
     """
     Formats the factor data, pricing data, and group mappings into a DataFrame
     that contains aligned MultiIndex indices of timestamp and asset. The
@@ -741,6 +752,10 @@ def get_clean_factor_and_forward_returns(factor,
         If True, compute quantile buckets separately for positive and negative
         signal values. This is useful if your signal is centered and zero is
         the separation between long and short signals, respectively.
+    cumulative_returns : bool, optional
+        If True, forward returns columns will contain cumulative returns.
+        Setting this to False is useful if you want to analyze how predictive
+        a factor is for a single forward day.
 
     Returns
     -------
@@ -774,7 +789,8 @@ def get_clean_factor_and_forward_returns(factor,
     """
 
     forward_returns = compute_forward_returns(factor, prices, periods,
-                                              filter_zscore)
+                                              filter_zscore,
+                                              cumulative_returns)
 
     factor_data = get_clean_factor(factor, forward_returns, groupby=groupby,
                                    groupby_labels=groupby_labels,