A divergence is a disagreement: price makes a new extreme, but the momentum indicator measuring that move refuses to confirm it. It is one of the few indicator signals with a real mechanical story behind it — and one of the easiest to misuse, because it invites you to fight trends.
What the disagreement means
Momentum oscillators — RSI, MACD, stochastic — measure the rate of price change. When price grinds to a new high but RSI prints a lower high than last time, the market is telling you the second push covered new ground with less force: smaller candles, more overlap, weaker follow-through. The engine is still moving the car, but the tachometer is dropping. That does not mean the car stops; it means the acceleration phase is over, and trends die from deceleration first.
Regular divergence — the reversal flavour
- Bearish: price makes a higher high, the oscillator makes a lower high. The uptrend's latest push lacked force.
- Bullish: price makes a lower low, the oscillator makes a higher low. The sell-off is losing energy.
Regular divergence appears at the ends of moves, which is precisely its danger: it fires early and repeatedly. A strong trend can print two, three, four consecutive divergences while continuing to run — each new extreme "diverging" against a momentum peak set in the explosive early phase. Divergence is a condition, not a trigger.
Hidden divergence — the continuation flavour
- Bullish hidden: price makes a higher low (an uptrend pullback), but the oscillator makes a lower low. Momentum washed out deeply while price barely gave ground — the pullback is weak relative to the fear it generated, and the trend is favoured to resume.
- Bearish hidden: the mirror, on a downtrend's bounce.
Hidden divergence is less famous and more useful, because it points with the trend: it is a pullback-quality filter for the structure trades you already want to take.
Spotting them correctly
Most "divergences" people mark are drawing errors. The checks that matter: compare swing to swing — connect two clearly separated peaks on price and the same two moments on the indicator, never a peak to a mid-slope wiggle. Use closed candles; an in-progress bar can erase a divergence in an hour. Keep the two swings reasonably close — momentum comparisons across half a year of bars mean little. And prefer divergences where the oscillator's first extreme sat in stretched territory (RSI above 70 or below 30), because a fade from a genuinely hot reading carries more information than a wobble around the midline.
MACD-histogram divergences behave the same way but fire even earlier, since the histogram is a second-order measure. Earlier is not better; it is just earlier.
The rule that saves the method
Divergence gives you permission to look — price gives you permission to act. The losing pattern is shorting a rising market the moment RSI prints a lower high; the market's answer is usually another leg up and another divergence. The workable pattern is: spot the divergence, then demand a structure event in its direction — a break of the last swing low, a trendline loss, a failed retest — and trade that, with the stop above the extreme that completed the divergence. The oscillator found the weakness; the chart confirms the turn. In that order, divergence trading is early-warning analysis. In the reverse order, it is top-picking with a technical excuse.
One last honesty note: every divergence that "nailed a top" in hindsight is easy to find. Scroll charts and mark the ones that failed too — there are more of them, and knowing the base rate is what separates using a tool from believing in it.