`StochRSI`#

Description#

StochRSI (Chande & Kroll, 1994) applies the Stochastic oscillator formula to an RSI series rather than to price. It is a "rate of change of momentum" indicator: more responsive than RSI alone and useful for spotting RSI turning points.

\[\begin{split} \begin{aligned} \text{RSI}[t] &= \text{RollingRSI}(\text{rsi\_period},\ \text{method}=\text{wilder})(x)[t] \\ \text{raw\_K} &= 100 \cdot \frac{\text{RSI} - \min(\text{RSI},\ \text{stoch\_period})}{\max(\text{RSI},\ \text{stoch\_period}) - \min(\text{RSI},\ \text{stoch\_period})} \\ \%K[t] &= \text{SMA}(\text{raw\_K},\ \text{smooth\_k}) \\ \%D[t] &= \text{SMA}(\%K,\ d) \end{aligned} \end{split}\]

1-input, 2-output (FunctorBase<_, 1, 2>).

Setting it up#

Configuration	Constructor	TA-Lib equivalent
Fast StochRSI (TA-Lib's `STOCHRSI`) -- default	`StochRSI(14, 14, 1, 3)`	`STOCHRSI(close, timeperiod=14, fastk_period=14, fastd_period=3)`
Slow StochRSI -- add smoothing on K	`StochRSI(14, 14, 3, 3)`	not in TA-Lib; common in pandas-ta `stochrsi(..., k=3, ...)`

smooth_k=1 is the identity SMA, collapsing %K = raw_K (the "fast" form). smooth_k >= 2 gives the slow form.

Parameters#

rsi_period (int, default 14): RSI lookback. Uses Wilder smoothing internally.
stoch_period (int, default 14): rolling min/max window applied to the RSI series.
smooth_k (int, default 1): SMA period applied to raw_K.
d (int, default 3): SMA period applied to %K to produce %D.

Warmup: both outputs are NaN until %D is valid, at sample index rsi_period + stoch_period + smooth_k + d - 3 (TA-Lib's convention -- gate both K and D together). For defaults that is index 29.

Range-zero: when the RSI window is flat (max == min), raw_K is undefined; we return 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 StochRSI

np.random.seed(0)
close = 100*np.exp(np.cumsum(np.random.normal(0, 0.01, size=300)))
out = StochRSI(14, 14, 1, 3)(close)
k = out[:, 0]
d = out[:, 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=close, name="close"), row=1, col=1)
fig.add_trace(go.Scatter(y=k, name="%K", line=dict(color="red")), row=2, col=1)
fig.add_trace(go.Scatter(y=d, name="%D", line=dict(color="orange")), row=2, col=1)
fig.update_layout(title="Price with Stochastic RSI (StochRSI)",
                  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="%K / %D (0-100)", row=2, col=1)
fig.show()

Implementation Details#

Composition: an internal RollingRSI(rsi_period, method="wilder"), two detail::MonotonicDeque (for rolling RSI min / max), and two detail::RollingMean (for the smooth_k and d SMAs). Amortised O(1) per step.

Reference#

Matches talib.STOCHRSI(close, timeperiod=N, fastk_period=K, fastd_period=D) bit-exactly (to ~1e-13 in tests/test_third_party_alignment.py).

StochRSI#