RSI divergence is a disagreement between price and the relative strength index. In a bullish divergence price makes a lower low while RSI makes a higher low. In a bearish divergence price makes a higher high while RSI makes a lower high.
It looks simple on a finished chart. The hard part is that "a swing" has to be defined, and a swing is only known some bars after it happens.
The rule, written in full
- RSI is the usual 14 period RSI of the close (see the RSI indicator).
- A swing low is a bar whose low is lower than the lows of the 5 bars before it and the 5 bars after it. A swing high is the mirror. The 5 is a setting, and a test that changes it is a different test.
- Bullish divergence: two swing lows in price, the second lower than the first, while RSI at the second is higher than RSI at the first, and the two swings are between 5 and 60 bars apart.
- Bearish divergence: the mirror image on swing highs.
- The signal is dated on the bar where the second swing is confirmed, which is 5 bars after the swing itself. Never on the swing bar: nobody knew it was a swing then.
Worked example (arithmetic, not a chart)
Price makes a swing low of 50.0, then later a swing low of 48.5. RSI at the first low reads 28 and at the second 34. Price is lower, RSI is higher: a bullish divergence. It becomes a signal 5 bars after the second low, when that low is confirmed. The numbers are made up to show the check.
What it can and cannot tell you
- On a chart that is already drawn, divergences are easy to see and look useful. Live, the swing is not confirmed for 5 bars, so part of the move is gone before the signal exists.
- A market in a strong trend can show divergence after divergence while it keeps going.
- Different swing rules give different divergences from the same data. That freedom is why the rule above is fixed first.
How we will test it
We will run exactly the rule above on many markets and years with realistic costs, date every signal on the confirmation bar, keep part of the data unseen until the rule is fixed, and report sample size, spread and the markets where it failed. Last tested: pending re-run. No result is stated here until it is run and dated.
Questions people type
Is RSI divergence reliable? It depends on the swing rule, the market and the costs. We do not state an answer before the test is run.
What is hidden divergence? A variant that compares in the other direction (price makes a higher low while RSI makes a lower low). It needs its own written rule and its own test.
Which timeframe is best? There is no best one. Shorter timeframes produce more signals and more costs.
Related
RSI indicator, MACD indicator, how we test, backtesting, method and limits, position size calculator.
Education only. Nothing here is financial advice or a signal.