Skip to content

Delta and CVD: what they measure and why a divergence is not a signal

By Author of ORDER FLOW — Market Mechanics

Delta is the simplest subtraction in order flow and the one that gets misread most. Nearly every misreading starts in the same place: taking it as a measure of direction.

The divergence between price and CVD, usually presented as a buy or sell signal, supports opposite readings. To follow the argument you only need to know what order flow records.

The subtraction

Delta is the volume executed at the ask minus the volume executed at the bid. It can be calculated per price level, per candle or over any period you choose.

Positive means aggressive buyers executed more than aggressive sellers. Negative means the reverse.

The trouble starts with what delta leaves out: how much passive liquidity was on the other side, whether the aggression achieved anything, and where price went.

Why delta is not direction

If more was bought than sold, it seems logical that price should rise. The mechanics of what moves price say otherwise: price moves when the resting liquidity at a level disappears, whoever is doing the buying.

Closes higher · negative delta Bid Ask 120 90 210 140 180 100 Delta −180 Price rises because passive sellers step away, with no buying needed. Closes lower · positive delta Bid Ask 80 190 110 220 95 155 Delta +280 Price falls while aggressive buyers keep executing. Delta measures the imbalance between aggressors; the close shows where price went.
Schematic of two NQ candles whose delta contradicts their close. Values are illustrative.

A candle can close higher on negative delta if passive sellers pulled their orders instead of letting them fill. Price rises, and the record shows more selling aggression than buying, because the advance came from nobody standing in the way.

A candle can also close lower on positive delta if aggressive buyers kept executing against a passive offer that would not give. Delta records their effort and the close records the result. When effort and result do not match, you are looking at absorption, and the right reading is the opposite of what the sign of the delta suggests.

Delta measures effort and price gives the result. The information appears when you put the two side by side.

What accumulating it adds

Cumulative delta, or CVD, is the running sum of delta over a period. Each candle adds its delta to the previous total, and the result is drawn as a line.

What it adds is time. A candle's delta says who was more aggressive in that minute; the cumulative line says who has been pushing through the session. A stretch with a sustained slope shows pressure that holds up candle after candle.

Before reading it, keep in mind that its absolute level carries no meaning. It is a running sum, so it depends on where the count started. The book calls that choice the anchor and lists the usual ones: the regular-session open at 9:30 ET, the London open or a relevant swing.

Overnight Regular session 0 9:30 ET −2,000 +2,000 CVD from the start of the overnight CVD from the 9:30 ET open Delta of each candle, the same for both lines Same deltas, two anchors: from 9:30 the lines run parallel, 4,000 apart.
Schematic of one series of NQ candle deltas accumulated from two different starting points. Values are illustrative.

In the schematic, the same 12 candle deltas end at −2,000 when the count starts with the overnight and at +2,000 when it starts at the open. Neither line is more bearish or more bullish than the other. From 9:30 they move in step, 4,000 apart, which is exactly the net delta of the overnight candles, −4,000. Only the shape can be read: slope, turns, and how it relates to what price is doing.

A divergence depends on what you compare

The most popular reading is the divergence: price makes a higher high, CVD does not follow, and the conclusion is that the advance lacks support.

The reasoning is sensible, but it skips a step. Declaring a divergence means choosing two highs and comparing them, and that choice is discretionary. On the same data, comparing the first high with the third can show no divergence at all, and comparing the second with the third can show a clear one. How much that choice decides, and the swing threshold sitting behind it, is the subject of delta divergence and swing selection.

The tool stays useful with one discipline: pick the highs to compare for a reason that existed before the comparison, a level that was already on your map for instance, and only then look for a divergence. A divergence that appears only when you go looking for it cannot be told apart from chance.

Something similar happens when reading CVD after a breakout. What separates a breakout that starts a move from one that exhausts it is whether CVD keeps building afterwards or peaks on the breakout candle, as the page on trend days and failed breakouts explains.

One divergence, two opposite mechanics

Suppose the divergence is properly set up, on reference points decided in advance. Price rises and CVD falls. That leaves the question of what produced it, and there are two answers that cannot both be true.

A · Absorption Price CVD Volume Heavy selling aggression held back by passive bids that will not budge. B · Thin market Price CVD Volume Little aggression from anyone. Price rises because nobody is offering. The same divergence. Volume is what separates one mechanic from the other.
Schematic of the same price and CVD divergence with two different levels of traded volume.

In the first case volume is high. There was plenty of aggression that did not turn into displacement, which means someone on the other side was containing it. That is the reading most order flow writing takes for granted.

In the second case volume is low. Price moved with almost nobody executing, in a thinned-out book where a handful of contracts is enough to cross several levels. Nobody is containing anything. The divergence looks the same on the chart and means something completely different.

What separates the two cases is total volume and how it is distributed across levels, which is what the footprint shows. The CVD line on its own cannot tell them apart.

Two warnings about the data

Where the classification comes from. The whole calculation depends on knowing whether each trade executed at the bid or at the ask. On CME futures that information exists and is exact. Where it does not, it is approximated from tick direction, and the approximation fails at the busiest moments, which are the ones you most want to read. The page on bid/ask data and tick-direction data shows where.

Which market is being discussed. Much of the material on CVD is written for crypto or spot forex. Those markets have no centralised book, so the CVD a user sees comes from one exchange or from their platform's provider, not from the market as a whole. Its conclusions do not carry over automatically to a CME contract, where every trade goes through the same book.

What it takes for it to mean something

Delta and CVD measure effort. They become meaningful when compared with the result, and when the point of comparison was decided beforehand.

In practice the reading has to happen at a level you already cared about, on volume that lifts it out of the noise, and an observable response has to follow: a reaction from the other side, or pressure that keeps going.

None of the three is in the CVD line. At the right place, a divergence tells you only that it is worth looking more closely.