Ehlers Stochastic

John Ehlers' Adaptive Stochastic. Pre-filters price through a RoofingFilter (high-pass + SuperSmoother bandpass) to isolate the tradable cycle band, then applies the classic Stochastic %K formula to the filtered output over period bars and re-smooths with a 2-bar SuperSmoother. Output is on the Ehlers [-1, +1] convention rather than the conventional [0, 100].

Quick reference

Item	Value
Family	Ehlers / Cycle (DSP)
Input type	f64
Output type	f64
Output range	[-1, +1]
Default parameters	period is required (Ehlers' typical 20)
Warmup period	period + roofing_warmup (~period + 50)
Interpretation	> 0.5 overbought; < -0.5 oversold; near-zero-lag at cycle band

Formula

filtered_t = RoofingFilter(10, 48).update(x_t)

over `period` bars of filtered values:
    raw_stoch_t = 2 · (filtered_t - min(filtered)) / (max - min) - 1

stoch_t = SuperSmoother-2-bar(raw_stoch)

The 2 · (X − MinX) / (MaxX − MinX) - 1 rescale puts the output in [-1, +1] instead of the classic [0, 100]. See crates/wickra-core/src/indicators/ehlers_stochastic.rs.

Parameters

Name	Type	Default	Constraint	Description
period	usize	none	> 0	Rolling-window length for the inner stochastic.

EhlersStochastic::new returns Error::PeriodZero for period == 0. The inner Roofing Filter uses fixed (10, 48) periods — Ehlers' canonical defaults; not user-tunable.

Inputs / Outputs

Indicator<Input = f64, Output = f64>. Python: EhlersStochastic(period).batch(prices) returns a 1-D np.ndarray with NaN in the warmup prefix. Node: same shape; update(value) returns number | null.

Warmup

warmup_period() accounts for Roofing-Filter warmup (~50 bars) plus the rolling window (period). For period = 20, expect ~70 bars before stable output.

Edge cases

• Constant input. Roofing Filter outputs zero; stochastic denominator is zero; the indicator returns None until variation appears.
• Saturation in trends. Even with Roofing Filter pre-processing, strong trends can leak residual signal that saturates the stochastic at ±1. Combine with a trend filter.
• Reset. reset() clears the Roofing Filter, the rolling window, and the 2-tap smoother.

Examples

Rust

use wickra::{BatchExt, EhlersStochastic, Indicator};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let prices: Vec<f64> = (0..200)
        .map(|i| 100.0 + (f64::from(i) * 0.3).sin() * 5.0)
        .collect();
    let mut es = EhlersStochastic::new(20)?;
    println!("row 100 = {:?}", es.batch(&prices)[100]);
    Ok(())
}

Python

import numpy as np
import wickra as ta

prices = 100 + np.sin(np.linspace(0, 60, 200)) * 5
es = ta.EhlersStochastic(20)
print('warmup:', es.warmup_period())  # ~70
print('row 100:', es.batch(prices)[100])

Node

const wickra = require('wickra');
const es = new wickra.EhlersStochastic(20);
const prices = Array.from({ length: 200 },
  (_, i) => 100 + Math.sin(i * 0.3) * 5);
console.log('row 100:', es.batch(prices)[100]);

Streaming

use wickra::{EhlersStochastic, Indicator};

let mut es = EhlersStochastic::new(20).unwrap();
let price_stream: Vec<f64> = Vec::new(); // your live price feed
for px in price_stream {
    if let Some(v) = es.update(px) {
        if v > 0.5 { /* overbought */ }
        if v < -0.5 { /* oversold */ }
    }
}

Interpretation

• Bandpass-prefiltered Stochastic. The Roofing Filter strips out trend and noise before the stochastic computes; what's left is the genuine cycle component, so overbought / oversold readings are meaningful rather than just "we're trending up".
• Near-zero-lag. The Roofing Filter has minimal phase lag in the cycle band; the stochastic on top adds typical rolling-window lag (period / 2).
• ±0.5 thresholds. Use as standard overbought / oversold; the [-1, +1] scale makes thresholds universal across markets.

Common pitfalls

• Treating it as a regular Stochastic. The output range is [-1, +1], not [0, 100]. Direct threshold comparisons against classical Stochastic levels (80, 20) won't work.
• Short warmup expectations. ~70 bars before stable on period = 20. Backtests on short data will underperform.
• Trend masking. Strong trends do leak into the cycle band; Ehlers Stochastic does not completely eliminate trend bias. Pair with a trend filter.

References

• John F. Ehlers, Cycle Analytics for Traders, Wiley (2013), ch. 7 — Adaptive Stochastic on Roofing Filter output.

See also

• RoofingFilter — the underlying pre-filter.
• Stochastic — classical Stochastic on raw price.
• SuperSmoother — final smoothing stage.
• Indicators-Overview — full taxonomy.