RollingRank#

Description#

Where does the current value sit within the trailing window?

\[ \text{rank}[t] = (\text{\#values} < y_t) + 1 + \tfrac{1}{2}(\text{\#ties} - 1) \]

Pandas's "average" tie-breaking rule (mean rank among tied values). Returns a 1-based rank in [1, w].

1→1. Circular window buffer + per-step counting sweep; O(W) per step. Bit-exact (0.0) to pandas.Series.rolling(w).rank().

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 RollingRank

np.random.seed(0)
price = 100 * np.exp(np.cumsum(np.random.normal(0.0005, 0.02, size=300)))
rank = RollingRank(window_size=50)(price)   # 1-based rank of the latest price in the last 50 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=price, mode="lines", name="price"), row=1, col=1)
fig.add_trace(go.Scatter(y=rank, mode="lines", name="rank",
                         line=dict(color="red")), row=2, col=1)
fig.update_layout(title="Rank of latest price in a 50-bar window (RollingRank)",
                  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="rank", row=2, col=1)
fig.show()