Direct answer
“Divergence in RSI” means that the movement in the Relative Strength Index (RSI) does not match the movement in price. Typically, price is making a new swing high while RSI is making a lower high (bearish divergence), or price is making a new swing low while RSI is making a higher low (bullish divergence). The core idea is momentum disagreement: price may be extending, while the RSI’s internal momentum measure is not confirming.
Importantly, divergence is a descriptive pattern of past movement. It is not a guaranteed signal of future direction, and different choices about RSI settings and swing points can change whether “divergence” appears.
Mechanism or definition
RSI is an oscillator that maps recent price changes into a bounded value (commonly 0 to 100). It uses two ingredients:
- The magnitude of recent gains and recent losses (often based on a fixed lookback period such as 14 bars).
- The ratio between average gains and average losses over that lookback.
When people talk about “RSI divergence,” they usually compare swing points. For example, in bearish divergence:
- Price: higher high (HH) between two swing peaks.
- RSI: lower high (LH) over the same general swing-to-swing interval.
A key detail is “over the same general time window.” RSI is derived from a rolling calculation, so the exact lookback length and the bar boundaries affect RSI shape. Also, divergence requires identifying swing highs/lows, which is partly subjective: two analysts can draw different swing points on the same chart.
Evidence or example
Assume a simple setup with RSI calculated on a fixed bar interval (for example, based on 14 periods). Consider two consecutive price swings:
- Swing 1 ends at time T1: price reaches a peak.
- Swing 2 ends at time T2: price reaches a higher peak than at T1.
Now compare RSI values at the peaks:
- If RSI at T2 is lower than RSI at T1, that is consistent with bearish divergence.
However, a common failure mode is “moving the goalposts.” If the analyst changes the RSI lookback period, the RSI peak can shift, and the relative ordering (higher vs lower) can flip. Another failure mode is matching mismatched swing points: if the second RSI “peak” is compared to a different price peak (or to a point that is not a clear swing), divergence may be created by inconsistent measurement.
Even when divergence is visible, it can dissipate quickly because RSI is sensitive to new bars that alter the rolling gains/losses. That means divergence can be identified after-the-fact but may not persist.
Limitations and risks
Several limitations matter in practice:
-
Confirmation limits (it may not “follow through”). Divergence describes disagreement at specific points, but it does not define what happens next or how long it should last. A later price move can erase the earlier momentum disagreement.
-
Calculation and selection sensitivity. RSI settings (lookback length, input source such as close vs another price) and the subjective choice of swing highs/lows can materially change whether divergence is present.
-
Hindsight bias and selection bias. Analysts often remember examples where divergence seemed to “work,” while ignoring the many cases where it did not. This creates an overestimation of reliability based on what is easiest to find on a chart.
-
Market condition dependence. Relationships between price swings and momentum oscillators can vary with volatility regimes, trend strength, and execution frictions. Historical patterns do not automatically establish a future relationship.
Verification or next question
To verify divergence claims independently, you can check the mechanics on your own chart:
- Use the same RSI calculation method and period across comparisons.
- Mark the swing highs/lows consistently (for example, by using a fixed rule such as “local extrema” rather than arbitrary placement).
- Re-check after each new bar: does divergence remain, strengthen, or disappear?
If you want a next step, a useful question is how RSI can be combined with other, independent forms of information (for example, trend context or volatility) to reduce reliance on a single descriptive pattern. Another practical question is how RSI-based analysis can be backtested responsibly, since the same selection and hindsight issues can distort results when you tune parameters on historical data.