Merge pull request #13 from ricequant/add_correlation

1.0.4: 增加相关系数指标 & add testcase & update README.MD
ricequant · Oct 28, 2022 · b5c55f7 · b5c55f7
2 parents 8cace38 + b98b062
commit b5c55f7
Show file tree

Hide file tree

Showing 4 changed files with 29 additions and 2 deletions.
diff --git a/README.MD b/README.MD
@@ -70,5 +70,6 @@ Out[4]:
  'excess_annual_volatility': 0.14574262245479186,
  'excess_max_drawdown': 0.050301584825935344,
  'var': 0.025161857442751474,
- 'win_rate': 0.3}
+ 'win_rate': 0.3,
+ 'correlation': 0.7136492542281812}
 ```
diff --git a/rqrisk/risk.py b/rqrisk/risk.py
@@ -207,5 +207,10 @@ def win_rate(self):
     def excess_win_rate(self):
         return len(self._portfolio[self._portfolio > self._benchmark]) / self.period_count
 
+    @indicator_property(min_period_count=2)
+    def correlation(self):
+        # 相关系数算法参考 https://numpy.org/doc/stable/reference/generated/numpy.corrcoef.html?highlight=corr#numpy.corrcoef
+        return np.corrcoef(self._portfolio, self._benchmark)[0][1]
+
     def all(self):
         return {k: getattr(self, k) for k, v in self.__class__.__dict__.items() if isinstance(v, IndicatorProperty)}
diff --git a/setup.cfg b/setup.cfg
@@ -5,7 +5,7 @@
 
 [metadata]
 name = rqrisk
-version = 1.0.3
+version = 1.0.4
 
 [versioneer]
 VCS = git

diff --git a/test.py b/test.py
@@ -73,6 +73,11 @@
     index=pd.date_range('2000-1-30', periods=9, freq='W')
 )
 
+volatile_monthly_benchmark = pd.Series(
+    np.array([1, 2, -5, 3, 10, -3, -1, 4, 1]) / 100,
+    index=pd.date_range('2000-1-30', periods=9, freq='M')
+)
+
 
 def _r(returns, benchmark_returns, risk_free_rate, period=DAILY):
     if benchmark_returns is None:
@@ -262,3 +267,19 @@ def _assert(returns, benchmark, desired_win_rate):
     _assert(positive_returns, zero_benchmark, 1)
     _assert(negative_returns, zero_benchmark, 0)
 
+
+def test_correlation():
+    """ 测试相关系数 """
+    def _assert(returns, benchmark, period):
+        r = _r(returns, benchmark, 0, period)
+        assert_almost_equal(r.correlation, returns.corr(benchmark))
+
+    _assert(one_return, one_benchmark, DAILY)                       # np.nan
+    _assert(positive_returns, simple_benchmark, DAILY)              # np.nan
+    _assert(positive_returns, volatile_benchmark, DAILY)            # 0.05773502691896258
+    _assert(negative_returns, simple_benchmark, DAILY)              # np.nan
+    _assert(negative_returns, volatile_benchmark, DAILY)            # -0.32984900530449723
+    _assert(weekly_returns, simple_weekly_benchamrk, WEEKLY)        # np.nan
+    _assert(weekly_returns, volatile_weekly_benchmark, WEEKLY)      # -0.3411258215912419
+    _assert(monthly_returns, simple_monthly_benchamrk, MONTHLY)     # np.nan
+    _assert(monthly_returns, volatile_monthly_benchmark, MONTHLY)   # -0.3411258215912419