RMI in plain terms
RMI (Relative Momentum Index) is a momentum-based indicator that summarizes how “strong” price movement has been over a recent lookback window. In practice, it is computed from price changes, then transformed into a bounded index value using fixed parameters (for example, a lookback length and a smoothing method). Because those steps are based on prior bars (candles), RMI is inherently a delayed description of what has already happened.
A key point for verification is that RMI is not a guarantee of direction. Even if its formula is correct, its output depends on the parameter settings and on what price series you feed into it (for example, close prices versus another field) and the timeframe you use.
How RMI can break down (mechanism and failure modes)
RMI’s usefulness can fail for several reasons that are consistent with how momentum indicators work:
-
Lookback window and smoothing lag Because RMI summarizes recent changes, it can lag during fast reversals. When price acceleration shifts quickly, the same window that captured “momentum” earlier can start reflecting the opposite behavior, causing the index to whipsaw.
-
Scaling and threshold misuse Many people treat index levels (such as crossing a threshold) as if they have fixed meaning. But the numeric behavior of an oscillator depends on how the underlying values are transformed and on the window length. Two traders using different parameter settings can get different crossings from the same market.
-
Regime dependence Momentum tends to behave differently across trend-like conditions versus range-bound conditions. In strong trends, momentum measures can stay elevated longer; in choppy markets, they can repeatedly rise and fall without follow-through.
-
Non-market frictions affecting outcomes Even if RMI tracks price relationships well on a chart, real outcomes depend on costs (spreads/commissions), execution, and slippage, none of which are captured by the indicator itself. Also, the indicator assumes a consistent data stream; gaps or different data sources can change the computed history.
Evidence or example: what to test independently
With no real-time assumptions, you can still validate limitations using historical data in a controlled way:
- Parameter sensitivity test: compute RMI with at least two lookback lengths and compare how often the indicator “signals” (for example, notable turns) move closer to or farther from actual turning points.
- Timeframe consistency check: repeat the same calculation on different timeframes. Momentum indicators often reflect different dynamics when bar size changes.
- Walk-forward comparison: instead of judging on one period, compare performance characteristics across multiple non-overlapping windows. Historical relationships that appear strong in one sample can weaken later.
When you run these checks, keep the assumptions explicit: the same price field, the same data source, and the same parameter definitions. If a behavior disappears after changing only one of those conditions, that is a practical limitation.
Limitations and risks to consider
The main limitations of RMI are uncertainty, context dependence, and calculation assumptions.
- Historical relationships do not establish future results. A pattern that matched past price changes can fail later because market behavior changes.
- Outputs vary with market conditions, costs, execution, and jurisdiction. RMI itself does not model how trades would actually fill under different liquidity and cost conditions.
- Indicator signals are not standalone evidence. RMI can suggest momentum shifts, but treating it as a single deciding input can amplify false positives, especially in noisy price action.
Verification and next question
To independently verify what RMI can (and cannot) do for your use case, focus on the limitations you can measure: lag, regime dependence, parameter sensitivity, and robustness across different historical windows. A useful next question is: under which market conditions does RMI behave differently for the specific timeframe and parameter setup you plan to use?