One of order flow's most useful warnings comes not from delta agreeing with price but from delta refusing to. When price grinds to a new high while the aggression behind it shrinks, the move is being powered by less and less real buying — a delta divergence. It does not time the turn, but it tells you the fuel is running low.
What delta divergence means
A delta divergence is a disagreement between the direction of price and the direction of aggression. In the ordinary case, price and delta move together: new highs come on strong positive delta, new lows on strong negative delta. A divergence breaks that link — price reaches a new extreme, but the delta accompanying it is weaker than the delta at the prior extreme. A higher high on smaller buying delta says each new push is being achieved with less aggressive demand than the last; a lower low on smaller selling delta says the same about supply. The message is always about waning force behind an extending move, which is why divergence is read as an early warning of exhaustion rather than an immediate reversal signal.
Bar delta divergence versus CVD divergence
Divergence can be read on single-bar delta or on cumulative delta, and the cumulative version is usually cleaner. Bar-by-bar delta is noisy, so comparing the delta of one high bar to another can mislead. Cumulative volume delta — CVD — sums each bar's delta into a running line, smoothing the noise into a trend of aggression you can compare directly against price, exactly as the diagram plots it beneath the candles. When price makes a higher high but the CVD line makes a lower high, aggressive buying is measurably weaker at the second peak; the divergence is visible as two lines pointing different ways. Most traders therefore watch CVD divergence for swing signals and reserve raw bar-delta divergence for fine-grained, single-bar reads.
Bearish divergence: price up, aggression down
The diagram shows the textbook bearish case: price prints a higher high near the end of the move while CVD makes a lower high. Read plainly, the second price peak was reached on less aggressive buying than the first — buyers are being asked to pay up for a level that took less demand to make than the previous one, a sign their conviction is fading. It frequently coincides with absorption, a large passive seller quietly meeting the thinning aggression, or with simple exhaustion as the pool of willing buyers empties. Bearish delta divergence is a warning that an uptrend is running on fumes, common near tops, and it invites caution on longs and preparation for a possible fade rather than an instant short.
Bullish divergence: the mirror image
Bullish delta divergence is the exact inverse and appears near bottoms. Price grinds to a lower low, but CVD makes a higher low — the new price trough was reached on less aggressive selling than the previous one. Sellers are losing their force even as price nudges to fresh lows, hinting that supply is drying up and that the downtrend is losing its engine. As with the bearish case, it often accompanies absorption, this time by a large passive buyer soaking up the last of the selling. It is a signal to ease off shorts and watch for a reversal higher, not a command to buy immediately. The logic is symmetric with the bearish case in every respect: an extending move made on shrinking aggression.
Confirming a divergence before acting
Divergence identifies weakening force but says nothing about timing, and a market can diverge for a long while before it turns — or never turn at all. That makes confirmation essential. Wait for price itself to act: a break of a short-term trendline, a failure to hold the new extreme, an exhaustion or absorption print at the level, or a shift in the CVD line's own slope. Divergence at a meaningful level — prior support or resistance, a value-area edge, a measured target — is far more trustworthy than divergence in open space. The disciplined sequence is to let divergence put you on alert, then require an independent trigger before committing, so that you are trading a confirmed turn rather than a mere warning.
Why divergences fail, and how to survive them
Delta divergences fail often enough that treating them as automatic reversals is a fast way to lose. A strong trend can absorb a divergence and keep going, printing several before one finally matters, so a lone divergence is weak evidence. On fragmented or off-exchange markets the bid/ask tagging behind delta is only estimated, which manufactures divergences that are really data artefacts. And a divergence with no level and no price confirmation is just a curiosity. The survivors treat divergence as one input in a weight-of-evidence approach: strongest on clean futures data, at a level, with price confirmation and ideally an oscillator agreeing. Used that way it is a genuine edge; used as a naked trigger it is a trap.