RMI

Relative Momentum Index — Wilder's RSI generalised to a multi-bar momentum lookback, smoother and slower to flip than the RSI in a trend.

Quick reference

Field	Value
Family	Momentum Oscillators
Input type	f64 (single price)
Output type	f64
Output range	[0, 100]; neutral 50 in a flat market
Default parameters	period and momentum are required
Warmup period (warmup_period())	momentum + period
Interpretation	RSI-style overbought/oversold over a longer change window.

Formula

change_t = close_t - close_{t-momentum}
gain     = max(change, 0),  loss = max(-change, 0)
avg_gain, avg_loss = Wilder-smoothed over `period`
RMI      = 100 * avg_gain / (avg_gain + avg_loss)

Roger Altman's RMI (1993) keeps Wilder's smoothing but measures the price change over momentum bars instead of one. A longer momentum window makes the oscillator smoother and stickier — it holds overbought/oversold readings longer in a sustained trend — while momentum = 1 collapses the RMI exactly to the Rsi. A flat market (no gains and no losses) returns the neutral 50.

Source: crates/wickra-core/src/indicators/rmi.rs.

Parameters

Name	Type	Default	Valid range	Source	Description
period	usize	none	>= 1	rmi.rs:62	Wilder smoothing period. 0 errors with Error::PeriodZero.
momentum	usize	none	>= 1	rmi.rs:62	Change lookback. momentum = 1 is the RSI. 0 errors.

(Python class wickra.RMI(period, momentum); Node new ta.RMI(period, momentum).)

Inputs / Outputs

From crates/wickra-core/src/indicators/rmi.rs:

use wickra::{Indicator, Rmi};
// Rmi: Input = f64, Output = f64
const _: fn(&mut Rmi, f64) -> Option<f64> = <Rmi as Indicator>::update;

Python returns float | None (streaming) / array.array('d') (batch, NaN for warmup). Node returns number | null / Array<number> with NaN.

Warmup

warmup_period() returns momentum + period: momentum inputs fill the lookback window, then period momentum-changes seed Wilder's averages. Pinned by accessors_and_metadata (warmup_period() == 19 for Rmi::new(14, 5)) and warmup_then_emits (an Rmi::new(2, 3) returns None for the first four inputs and emits on the fifth).

Edge cases

• momentum = 1 equals RSI. momentum_one_equals_rsi runs a standalone Wilder RSI alongside and asserts identical values.
• Pure uptrend → 100. Every momentum-spaced change is positive, so avg_loss = 0; pinned by pure_uptrend_is_one_hundred.
• Flat market → 50. flat_market_is_neutral pins a constant series to 50.
• NaN / infinity inputs. Ignored, returning the last value — pinned by ignores_non_finite_input.
• Reset. reset_clears_state. Streaming/batch: batch_equals_streaming.

Examples

Rust

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

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut rmi = Rmi::new(3, 2)?;
    let prices: Vec<f64> = (1..=10).map(f64::from).collect();
    let out: Vec<Option<f64>> = rmi.batch(&prices);
    println!("warmup_period = {}", rmi.warmup_period());
    println!("{:?}", out);
    Ok(())
}

Output:

warmup_period = 5
[None, None, None, None, Some(100.0), Some(100.0), Some(100.0), Some(100.0), Some(100.0), Some(100.0)]

On the steady uptrend every 2-bar change is positive, so avg_loss = 0 and the RMI saturates at 100 from the first emission (input momentum + period = 5).

Python

import numpy as np
import wickra as ta

rmi = ta.RMI(3, 2)
out = rmi.batch(np.arange(1.0, 11.0))
print("warmup_period =", rmi.warmup_period())
print(out)

Output:

warmup_period = 5
[ nan  nan  nan  nan 100. 100. 100. 100. 100. 100.]

Node

const ta = require('wickra');
const rmi = new ta.RMI(3, 2);
const prices = Array.from({ length: 10 }, (_, i) => i + 1);
console.log(rmi.batch(prices));
console.log('warmupPeriod:', rmi.warmupPeriod());

Output:

[
  NaN, NaN, NaN, NaN, 100,
  100, 100, 100, 100, 100
]
warmupPeriod: 5

Streaming

use wickra::{Indicator, Rmi};

let mut rmi = Rmi::new(14, 5)?;
let mut last = None;
for i in 0..80 {
    last = rmi.update(100.0 + (f64::from(i) * 0.2).sin() * 5.0);
}
println!("{last:?}");
# Ok::<(), Box<dyn std::error::Error>>(())

Interpretation

The RMI is an RSI you can tune for persistence. By widening the change window (momentum), you make the oscillator hold its readings through the shallow pullbacks that would whip a standard RSI in and out of overbought territory — useful for trend-following entries where you want the oscillator to stay "hot" while the trend runs.

Typical uses:

1. Trend-tolerant overbought/oversold. Use the usual 70/30 (or 80/20) bands; a higher momentum keeps the RMI pinned in strong trends rather than oscillating.
2. Momentum confirmation. RMI holding above 50 confirms bullish momentum; crosses of 50 mark momentum shifts.
3. Divergence. Same divergence reading as the RSI, but less noisy.

Common pitfalls

• Confusing it with DMI/ADX. "RMI" is the Relative Momentum Index, not the Directional Movement Index; it is an RSI variant, not a trend-strength gauge.
• Over-long momentum. Too large a momentum makes the RMI sluggish and late; keep it a fraction of the smoothing period for most uses.

References

Roger Altman, "Relative Momentum Index: Modifying the RSI", Technical Analysis of Stocks & Commodities, 1993.

See also

• Indicator-Rsi — the RMI at momentum = 1.
• Indicator-ConnorsRsi — another RSI extension.
• Indicators-Overview — the full taxonomy.