sharpe for constant series (#428)

* sharpe for constant series

* fmt

* address log warnings (#429)

cvx/simulator/portfolio.py

@@ -355,7 +355,7 @@ def snapshot(
             table.loc[label_strategy, "end"] = self.nav.index[-1].strftime("%Y-%m-%d")
             table.loc[label_strategy, "# assets"] = len(self.assets)
 
-            s = sharpe(self.nav.pct_change().dropna())
+            s = sharpe(self.nav.ffill().pct_change(fill_method=None).dropna())
             table.loc[label_strategy, "Sharpe ratio"] = f"{s:.2f}"
 
             if benchmark is not None:
@@ -372,7 +372,7 @@ def snapshot(
                 table_bench.loc[label_benchmark, "# assets"] = ""
                 table_bench.loc[
                     label_benchmark, "Sharpe ratio"
-                ] = f"{sharpe(benchmark.pct_change().dropna()):.2f}"
+                ] = f"{sharpe(benchmark.ffill().pct_change(fill_method=None).dropna()):.2f}"
 
                 table = pd.concat([table, table_bench], axis=0)
 

cvx/simulator/utils/metric.py

@@ -1,5 +1,4 @@
-from math import sqrt
-
+import numpy as np
 import pandas as pd
 
 
@@ -8,12 +7,16 @@ def _periods(ts):
     compute the number of periods in a time series
     """
     series = pd.Series(data=ts.index)
-    return 365 * 24 * 60 * 60 / (series.diff().mean().total_seconds())
+    return 365 * 24 * 60 * 60 / (series.diff().dropna().mean().total_seconds())
 
 
 def sharpe(ts, n=None):
     """
     compute the sharpe ratio of a time series
     """
-    n = n or _periods(ts)
-    return ts.mean() / ts.std() * sqrt(n)
+    std = ts.std()
+    if std > 0:
+        n = n or _periods(ts)
+        return (ts.mean() / std) * np.sqrt(n)
+    else:
+        return np.inf