RollingPercentile

Description

Percentile position of the current value within the trailing window:

\[ \text{percentile}[t] = \text{rank}[t] / w \]

with the same "average" tie rule as RollingRank. Returns values in [1/w, 1]. Bit-exact (0.0) to pandas.Series.rolling(w).rank(pct=True).

1→1. O(W) per step.

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 RollingPercentile

np.random.seed(0)
price = 100 * np.exp(np.cumsum(np.random.normal(0.0005, 0.02, size=300)))
pct = RollingPercentile(window_size=50)(price)   # where the latest price sits in the last 50 bars, 0 to 1

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=price, mode="lines", name="price"), row=1, col=1)
fig.add_trace(go.Scatter(y=pct, mode="lines", name="percentile",
                         line=dict(color="red")), row=2, col=1)
fig.update_layout(title="Percentile of latest price in a 50-bar window (RollingPercentile)",
                  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="price", row=1, col=1)
fig.update_yaxes(title_text="percentile", row=2, col=1)
fig.show()