RSI / Indicator Divergence

Divergence describes a disagreement between price and a momentum indicator about the strength of a move. When price reaches a new extreme but the indicator's matching extreme is less pronounced, the second push was produced with less net momentum than the first. RSI is the indicator most often used for this reading, but the concept is indicator-agnostic: the same comparison works with MACD, Stochastic, CCI, volume-derived series, or a second correlated instrument.

A divergence is a condition, not a trade. Every implementation adds rules to turn it into an entry, a stop and an exit, so most of the disagreement between published versions concerns not what divergence is, but how it is detected, confirmed and traded.

## How it works

RSI compares average gains to average losses over a lookback window, so it responds to the rate of change rather than to the level of price. A trend that keeps printing new highs on progressively smaller impulses will show a lower RSI peak at the later high, and that gap between price structure and momentum structure is the divergence.

Detecting it requires three decisions: which candles count as swing highs or lows, which two of those swings are compared, and which indicator value is read at each. A swing point is usually only identifiable some bars after it forms, so a divergence becomes actionable later than it appears on a finished chart — a lag that is invisible in review and decisive in testing.

## Main variants

Regular (classic) divergence is the reversal-oriented reading: price makes a higher high while the oscillator makes a lower high, or price makes a lower low while the oscillator makes a higher low.

Hidden divergence inverts the comparison as a continuation reading: in an uptrend, price makes a higher low while the oscillator makes a lower low, suggesting the pullback gave up less ground than momentum implies.

Indicator substitution keeps the logic and changes the sensor. MACD and its histogram, Stochastic, CCI, Awesome Oscillator, OBV or cumulative delta each shift the timing and frequency of the pivots without altering the underlying idea.

Cross-asset divergence, often labelled SMT, applies the comparison to two correlated instruments — an index against a sibling index, or a currency pair against a dollar index — where one takes out a prior extreme and the other does not. Here the second market plays the role of the indicator.

Multi-timeframe versions locate the divergence on a higher timeframe and take the entry on a lower one.

## What differentiates implementations

The swing-detection rule is normally the largest source of variation: fractals, a fixed number of bars either side of a pivot, or a ZigZag threshold each select different highs and lows from the same chart. Next comes the pairing rule — how far apart the two swings may be, and whether intervening swings invalidate the comparison. Versions also differ in whether they read wicks or closes, whether the indicator value is taken at its own peak or at the price pivot's bar, and what must happen before the signal is actionable: a structure break, a trendline break, a close beyond a level, or arrival at a predefined zone. Trend filters, session windows and the stop logic account for most of the rest.

## Common mistakes

The most common is treating divergence as a standalone reversal trigger. Momentum can decelerate for a long stretch while a trend continues, so the condition may repeat several times before price responds, and each repetition costs a trade when there is no filter or invalidation rule. Next is hindsight: on a finished chart the eye pairs the two swings that worked, a selection unavailable in real time. Third is repainting — using a pivot at its own bar index when it could only be confirmed several bars later — which injects future information into every rule built on it. Others include leaving the pairing rule undefined, tuning the RSI period until historical signals look clean, and assuming a cross-asset relationship still holds without checking the correlation regime.

## How to evaluate and backtest a version

Start by codifying the rules until two readers would mark the same signals on the same chart: pivot definition, pairing window, indicator reading, confirmation event, invalidation and exit. A version that cannot be written this way can be illustrated but not tested.

Timestamp each signal at the bar where it became confirmable, not at the pivot bar. Then compare against a baseline: the same entry trigger without the divergence condition. Divergence usually acts as a filter over an existing structure or level entry, so the question is what it adds beyond the trigger itself.

Read results by regime and by instrument rather than in aggregate, since trending and ranging periods treat this condition very differently. Sweep the sensitive parameters — RSI period, pivot lookback, maximum distance between swings — and prefer settings sitting on a broad plateau rather than an isolated peak. Include spread and commission, since many divergence entries are counter-trend with tight stops. Finally, check whether the outcome depends on a handful of trades, and validate on data not used while tuning.

## Versions decoded on this page

Each page below documents one implementation with its own detection, confirmation and risk rules:

- AI Macro Divergence Swing Trading Strategy — BeSomebodyFX - DXY, EURUSD Divergences — Titanes del Trading - Divergence Trading System — StockShodh - SMT Divergence Trading Strategy — Asia Forex Mentor (Ezekiel Chew) - Highs and Lows, Divergence, Multiple Time Frames — Blueberry.

Strategies in this concept (29)

Frequently asked questions

Is divergence a reversal signal or a continuation signal?

It can be read either way depending on how the swings are paired. Regular divergence compares a new price extreme against a weaker indicator extreme and is used as a reversal-oriented reading; hidden divergence compares a pullback that holds better than momentum suggests and is used as a continuation reading. Neither is a signal by itself: both describe a condition, and the surrounding rules decide whether it becomes a trade.

Does it matter whether I use RSI, MACD or Stochastic?

Less than the detection rules do. Each indicator has different smoothing and sensitivity, so the same chart produces pivots at slightly different bars and a different number of signals. What this means in practice is that settings are not transferable between indicators, and a version tested with one oscillator has not been tested with another. The indicator and the pivot rules have to be evaluated together as a single rule set.

Why do divergences look much better on a chart than in a backtest?

Three reasons usually account for the gap. On a finished chart the eye pairs the two swings that worked, a choice not available in real time. Pivots are only confirmable some bars after they form, so live entries happen later and worse than marked ones. And rules that reference a pivot at its own bar index repaint, quietly using information that did not exist yet. Fixing the timestamps alone often changes the result substantially.

What is SMT divergence and how does it relate to indicator divergence?

SMT applies the same comparison, but the second series is another correlated instrument instead of an oscillator: one market takes out a prior high or low and the correlated one fails to. The logic is identical; the extra requirement is that the correlation regime holds during the period being traded or tested, and that both series use synchronised timestamps and comparable session hours. A relationship that has drifted turns the signal into noise.

What should confirm a divergence before an entry is taken?

Most workable implementations require an event that shifts the burden of proof after the divergence is visible: a break of the swing that formed it, a trendline break, a close beyond a defined level, or the condition appearing at a pre-marked zone rather than anywhere on the chart. Just as important is the invalidation rule, which states when the divergence stops being valid without a trade being taken.

Which timeframes and markets does the concept apply to?

Any timeframe with a definable swing structure can be used. Lower timeframes generate more pivots and therefore more signals, more noise and more sensitivity to spread and commission, since divergence entries are often counter-trend with tight stops. Results also do not transfer between instruments: a parameter set that behaves one way on an index future will not necessarily behave the same on a currency pair, so each instrument needs its own test.

All strategy concepts · Explore all strategies