RollingHitRate#

Description#

Fraction of strictly-positive samples in the trailing window:

\[ \text{HitRate}[t] = \frac{1}{w}\ \text{count}(r_i > 0,\ i \in \text{window}) \]

Output in [0, 1]. Composes detail::RollingSum over the indicator (r > 0).

NaN handling#

Policy: ignore. A NaN in any input at index t causes the function to skip that step: output at t is NaN and internal state is unchanged. Subsequent finite samples are processed as if step t had not occurred.

Examples#

Usage example#

import numpy as np
import plotly.graph_objects as go
from plotly.subplots import make_subplots
from screamer import RollingHitRate

np.random.seed(0)
ret = np.random.normal(0.0005, 0.01, size=300)
hit = RollingHitRate(window_size=63)(ret)   # fraction of positive returns in the last 63 bars

fig = make_subplots(rows=2, cols=1, shared_xaxes=True,
                    row_heights=[0.55, 0.45], vertical_spacing=0.08)
fig.add_trace(go.Scatter(y=ret, mode="lines", name="returns"), row=1, col=1)
fig.add_trace(go.Scatter(y=hit, mode="lines", name="hit rate",
                         line=dict(color="red")), row=2, col=1)
fig.update_layout(title="Fraction of positive returns over 63 bars (RollingHitRate)",
                  margin=dict(l=20, r=20, t=60, b=20),
                  legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor="right", x=1))
fig.update_yaxes(title_text="returns", row=1, col=1)
fig.update_yaxes(title_text="hit rate", row=2, col=1)
fig.show()