RecursiveMedian
EMA of a 5-bar median filter for smooth tracking with minimal jitter.
Visual Example

Synthetic ideal per library logic. Generated 2026-07-01 IST via docs/generate_all_previews.py (reproducible; maps to core Next<T> implementation).
Description
EMA of a 5-bar median filter for smooth tracking with minimal jitter.
Use to filter out extreme outliers and noise while maintaining trend sensitivity. Excellent as a baseline for other oscillators.
Part of QuantWave's Ehlers digital signal processing suite. Designed for low-lag cycle and trend work — pair with Roofing Filter or SuperSmoother on noisy inputs.
Standard filters like SMA or EMA are distorted by price spikes. The recursive median filter uses the median to reject outliers and an EMA to provide smoothness, offering a cleaner trend representation than standard moving averages.
Typical applications:
- Use for cycle timing in mean-reverting regimes
- Gate with Hurst exponent or ADX before taking cycle signals
- Allow
12+ bars warm-up for filter state to stabilise - Chain with Roofing Filter when input is noisy
QuantWave implements this via the universal Next<T> trait — bit-identical across Rust streaming, Python streaming, and Polars .ta() batch plugins.
Formula / Specification
Implementation (quantwave-core/src/indicators/recursive_median.rs):
[ \alpha = \frac{\cos(360/P) + \sin(360/P) - 1}{\cos(360/P)} ] [ RM_t = \alpha \cdot \text{Median}(Price, 5)t + (1 - \alpha) \cdot RM ]
Gold-standard parity vectors: quantwave-core/tests/gold_standard/recursive_median.json.
Parameters
| Parameter | Default | Description |
|---|---|---|
lp_period |
12 | Low-pass smoothing period |
Usage Examples
Streaming (Rust)
use quantwave_core::indicators::RECURSIVE_MEDIAN;
use quantwave_core::traits::Next;
let mut ind = RECURSIVE_MEDIAN::new(12);
for price in &prices {
let value = ind.next(price);
}
Streaming (Python)
from quantwave import RECURSIVE_MEDIAN
ind = RECURSIVE_MEDIAN(12)
for price in prices:
value = ind.next(price)
Polars Batch (Python)
import polars as pl
import quantwave as qw
def apply_recursivemedian(series: pl.Series) -> pl.Series:
ind = qw.RECURSIVE_MEDIAN(12)
return pl.Series([ind.next(float(v)) for v in series.to_list()])
df = (
pl.read_csv('ohlcv.csv')
.lazy()
.with_columns(
pl.col("close").map_batches(apply_recursivemedian, return_dtype=pl.Float64).alias("recursivemedian")
)
.collect()
)
All surfaces are bit-identical via the single Next<T> implementation and proptests.
Edge Cases & Limitations
- Recursive DSP filters require a warm-up period; first N bars may be unstable or raw-pass-through.
- Designed for cyclic/mean-reverting regimes; trending markets can produce lag or drift.
- Parameter
period(or equivalent) controls cutoff — too small adds noise, too large adds lag. - Prefer chaining with other Ehlers tools (Roofing Filter, SuperSmoother) on noisy inputs.
- Validated via proptests against gold-standard vectors where available.
- No look-ahead bias; suitable for live streaming and batch feature pipelines.
Boundary Behavior
| Condition | Behavior |
|---|---|
| Warm-up | Leading bars return NaN until warmup_bars is satisfied. |
| period > len | When period exceeds series length, output is all NaN. |
| NaN inputs | NaN in input propagates to output (NaN out). |
| Invalid params | Non-positive period or missing required params raise ValueError. |
| Empty data | Empty input returns an empty result series. |
Related Indicators & See Also
Sources & References
Primary Source: https://www.traders.com/Documentation/FEEDbk_docs/2018/03/TradersTips.html
Implementation: quantwave-core/src/indicators/recursive_median.rs (RECURSIVE_MEDIAN / RECURSIVE_MEDIAN_METADATA).
Parity: quantwave-core/tests/gold_standard/recursive_median.json
Provenance: Standards bulk upgrade 2026-07-01 IST — see docs/DOCUMENTATION_STANDARDS.md.