What is divergence in trading?
Divergence is one of the few chart ideas that is genuinely simple to define and genuinely easy to misuse. The definition takes one sentence. The trouble starts because there are four of them, two of which mean the opposite of the other two, and because the pattern is structurally impossible to see at the moment it matters. This page is the lookup: what counts, which one you are looking at, and how old it already is.
What exactly counts as a divergence?
Three things have to be true, and all three are mechanical rather than a matter of judgement.
First, you need two pivots on price. A pivot high is a bar whose high is greater than the highs of some number of bars on each side of it — call that number R. A pivot low is the mirror image. R is a setting you choose; everything else on this page follows from it.
Second, you need the two matching pivots on an indicator, taken at the same bars. The indicator can be RSI, MACD, an oscillator of your own, or volume — the mechanic does not care. What changes with the choice is not whether a divergence exists but what the disagreement is about.
Third, the two pairs have to disagree in direction. Price up, indicator down; or price down, indicator up. If both go the same way there is no divergence, and that is the ordinary case: most of the time an indicator confirms the move that produced it.
Notice what the definition does not contain. There is no clause about what happens next. Divergence is a description of two things that already occurred, stated as a comparison. Every prediction attached to it is something a reader added.
What are the four types?
Two questions produce four answers. Are you looking at highs or at lows? And which line made the bigger extreme, price or the indicator? The second question is the one that gets skipped, and it is the one that flips the meaning.
| Type | Price | Indicator | Reads as |
|---|---|---|---|
| Regular bearish | higher high | lower high | the rally is losing strength |
| Regular bullish | lower low | higher low | the sell-off is losing strength |
| Hidden bearish | lower high | higher high | a pullback inside a downtrend — continuation |
| Hidden bullish | higher low | lower low | a pullback inside an uptrend — continuation |
“Reads as” is the conventional interpretation, not a claim about what price will do. The evidence in every row is identical in kind: one of the two lines made a new extreme and the other did not.
The shortcut worth memorising is this: regular = price made the extreme; hidden = the indicator made the extreme. Regular divergences are read as the current move running out of fuel. Hidden divergences are read as a trend catching its breath. Confuse the two and you do not get a slightly weaker signal, you get the inverted one — a continuation pattern mistaken for a reversal pattern is exactly the error that puts someone short into an uptrend.
How late is it by the time you can see it?
This is the part that no amount of skill or screen time removes, and it can be worked out exactly. It comes from two facts about pivots, both of which follow from the definition in the first section.
Fact one: two pivots of the same kind cannot be closer than R+1 bars apart. Suppose two pivot highs sat only R bars apart. Then the second one falls inside the first one’s right-hand window, so it must be lower than the first. But the first also falls inside the second one’s left-hand window, so it must be lower than the second. Both cannot be true. So the minimum gap is R+1 bars.
Fact two: the second pivot is not a pivot until R more bars have closed. You cannot know a bar is a high until you have seen the bars that failed to exceed it. That is not a limitation of your software; it is what the word means.
Add them. The soonest a divergence can exist and be visible is (R+1) + R = 2R+1 bars after the first pivot. With R set to 5, that is 11 bars — and that is the best case, not the typical one, because real pivots are usually further apart than the minimum.
| Chart | First pivot → second (6 bars) | Second → confirmed (5 bars) | Age when it first exists (11 bars) |
|---|---|---|---|
| 15m | 1 h 30 | 1 h 15 | 2 h 45 |
| 1H | 6 h | 5 h | 11 h |
| 4H | 24 h | 20 h | 44 h |
| 1D | 6 days | 5 days | 11 days |
| 1W | 6 weeks | 5 weeks | 11 weeks |
Our own calculation, worked with R = 5. Reproduce any row as bar duration × 11. Change R and every figure changes with it — the general result is 2R+1 bars.
Read the last column as what it is. On a 4-hour chart, the earliest possible moment you can say “there is a bearish divergence here” is 44 hours after the first high formed. Nothing about that is fixable by watching more closely. It is the price of the pattern being real rather than guessed at, and it is the single strongest argument for treating divergence as context rather than as a trigger.
Why do divergences seem to be everywhere?
Because the number of lines you could draw grows much faster than the number the definition allows. This is worth doing with arithmetic, because it explains an experience almost every new trader has: the feeling that once you learn about divergence you start seeing it on every chart.
Put n pivot highs on screen. The number of pairs of pivots you could connect a line between is n(n−1)/2. But only pivots that sit next to each other form a divergence the definition recognises, and there are only n−1 of those. Everything else skips over an intervening pivot — which means selecting a more convenient starting point, which means the pattern was constructed rather than found.
| Pivots on screen | Pairs you could connect | Neighbouring pairs | Share that skip a pivot |
|---|---|---|---|
| 4 | 6 | 3 | 50.0% |
| 6 | 15 | 5 | 66.7% |
| 8 | 28 | 7 | 75.0% |
| 10 | 45 | 9 | 80.0% |
| 12 | 66 | 11 | 83.3% |
Our own calculation. Pairs = n(n−1)/2; neighbouring pairs = n−1; the share that skip is 1 − 2/n, which rises with every extra pivot you allow yourself to scan.
With 8 pivots in view — a perfectly ordinary screen — 21 of the 28 available lines, 75%, are ones the definition does not sanction. Zoom out to find more pivots and the share gets worse, not better. The abundance is a property of the counting, not evidence about the market.
What are the most common mistakes?
| Mistake | Why it fails | Instead |
|---|---|---|
| Acting on it before the second pivot is confirmed | Until the pivot is confirmed you are predicting a divergence, not reading one. If price makes a higher high instead, the pattern you traded never existed. | Wait for confirmation, and accept that waiting is the cost of the pattern being real. |
| Pairing pivots that are not neighbours | Skipping over an intervening pivot to reach a more flattering one manufactures the signal. The chart did not offer it; you selected it. | Pair each pivot with the one next to it, and let the awkward ones stand. |
| Treating regular and hidden as the same pattern | They carry opposite readings. A hidden bullish divergence read as a regular one turns a continuation signal into a reversal signal. | Name which of the four you are looking at before you draw any conclusion from it. |
| Counting the ones that worked | Divergences that preceded a turn are memorable; the ones that printed and were run over are not. Recall is not a sample. | Mark every one on a year of a single chart, failures included, and count. |
| Using it against a trend that is still intact | A weaker push is still a push in the same direction. Fading it puts you on the opposite side of the prevailing move with no evidence that the move has stopped. | Use it to reduce exposure you already have, not to open a position facing the other way. |
When is this page’s own advice wrong?
The caution here is calibrated for one use: opening a position against a trend because a divergence appeared. For that use the lag and the counting problem are decisive. There are two situations where the same caution is misplaced.
The first is managing a position you already hold. If you are long and a regular bearish divergence confirms, you are not timing an entry — you are deciding whether to tighten a stop or take part of the position off. The 11-bar lag costs you almost nothing, because the decision was not time-critical to begin with. Used this way divergence is a reasonable input, and the worst outcome is that you reduced risk during a move that carried on without you.
The second is a confirmed range. In a market oscillating inside horizontal support and resistance, there is no trend for the pattern to fight, and fading a weakening push at the top of a range is a coherent trade for reasons that have little to do with the divergence itself. The signal is not doing the heavy lifting; the range is. That is a fair use as long as you are honest about which one you are actually trading.
What survives in every case is the modest claim. The second push was weaker than the first. That is true, it is useful, and it is all the pattern ever says. For the longer argument about whether that is enough to trade on, and what RSI readings convert to in terms of buying and selling pressure, see the RSI lesson.
FAQ
Does divergence predict a reversal? No. It reports that the second push was weaker than the first, which is a statement about the leg that has already happened. A trend can weaken and then re-accelerate, printing a textbook divergence at every step while price keeps going. Weakening and reversing are different claims, and only the first one is supported by what you can see.
What is the difference between regular and hidden divergence? Which side made the bigger extreme. In a regular divergence price makes the new extreme and the indicator does not, which reads as the move losing strength. In a hidden divergence the indicator makes the new extreme and price does not, which reads as an ordinary pullback inside a trend that is still intact. They point in opposite directions, so mixing them up inverts your conclusion.
Which indicator is best for divergence — RSI, MACD, or volume? The mechanic is the same on any of them, so the honest answer is that the choice matters less than people expect. What does change is what the disagreement means: an RSI divergence is about the balance of recent up and down moves, while a volume divergence is about participation. Use one, know which question it answers, and do not stack three and count agreement as confirmation.
Can I trade divergence on a 5-minute chart? You can draw one, but the lag does not shrink in proportion to your patience. With a 5-bar pivot setting, the earliest a divergence can exist is 11 bars — 55 minutes on a 5-minute chart — and on fast charts the pivots themselves are mostly noise. The pattern gets easier to find and less informative at the same time.