An RSI divergence, in which price sets a fresh high but RSI does not, is often described as a reversal signal. In strong trends, divergences can appear many times before any reversal happens, which makes them unreliable on their own.
A method to measure: define divergence precisely, such as a higher price high with a lower RSI high within 20 bars, and record what happens over the next 10 or 20 bars. Count how often price reverses by a meaningful amount compared with continuing.
A hypothetical result might show that divergence alone is close to a coin flip, but divergence at a higher-timeframe level performs better. The point of testing is to find out which combinations, if any, add value.
Automating the detection in code helps avoid subjective marking, which tends to remember the successful examples and forget the failures.
Divergence may be more useful as a warning to tighten stops than as an entry signal.
Have you tested divergence systematically?
RSI divergence: how often does it actually lead to reversals?
-
FXS London Scalper
- Posts: 975
- Joined: Sun Oct 18, 2026 12:25 pm
RSI divergence: how often does it actually lead to reversals?
FXS Official Team | London Session Desk
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: RSI divergence: how often does it actually lead to reversals?
The definition step is the part most people skip, and it's probably why divergence has such a mixed reputation. "Price made a higher high and RSI didn't" sounds precise until you try to code it and realise you have to decide what counts as a high, how many bars apart the two highs can be, and whether the RSI peaks need to line up exactly with the price peaks.
Another thing worth recording in a test like this is what happens to the trend rather than just whether it reverses. A common outcome after a bearish divergence in an uptrend isn't a reversal at all, it's a sideways phase where momentum cools while price holds up. For a trend follower that's still useful information, because it suggests tightening a stop or not adding, even if it's useless as a short signal.
The higher-timeframe level idea makes sense for a structural reason too. Divergence by itself only says momentum is fading. Fading momentum into empty space often just means the trend is resting. Fading momentum into a level where sellers have shown up before gives the slowdown a reason to turn into something more.
It would also be interesting to split the results by session. On EURUSD a divergence that forms late in the New York afternoon, when volume is thinning, might behave very differently from one that forms in the middle of London, simply because there's less participation to drive a reversal.
On the warning-versus-signal point, I'd lean the same way: as a reason to manage an existing position more tightly it seems far more defensible than as an entry trigger on its own.
Another thing worth recording in a test like this is what happens to the trend rather than just whether it reverses. A common outcome after a bearish divergence in an uptrend isn't a reversal at all, it's a sideways phase where momentum cools while price holds up. For a trend follower that's still useful information, because it suggests tightening a stop or not adding, even if it's useless as a short signal.
The higher-timeframe level idea makes sense for a structural reason too. Divergence by itself only says momentum is fading. Fading momentum into empty space often just means the trend is resting. Fading momentum into a level where sellers have shown up before gives the slowdown a reason to turn into something more.
It would also be interesting to split the results by session. On EURUSD a divergence that forms late in the New York afternoon, when volume is thinning, might behave very differently from one that forms in the middle of London, simply because there's less participation to drive a reversal.
On the warning-versus-signal point, I'd lean the same way: as a reason to manage an existing position more tightly it seems far more defensible than as an entry trigger on its own.