Methodology · 22 Aug 2026

What whale CVD actually tells you

Every trade has two sides, so volume alone tells you almost nothing. For every buyer there was a seller. What matters is who crossed the spread: who was impatient enough to take the price on offer rather than wait. That impatience is information.

Cumulative volume delta, CVD, measures exactly that. Add up the dollar value of trades where the buyer was the aggressor, subtract the ones where the seller was, and you get a running line of net urgency. Rising CVD means buyers keep reaching. Falling CVD means sellers keep hitting.

Useful. But standard CVD has a blind spot you could drive a bus through: it treats a $200 market order and a $2 million market order as morally equivalent. It adds them to the same line. And those two orders are simply not the same animal. One is somebody on their phone at the bus stop. The other is a desk with a mandate and a rather good reason.

Split the tape by size and it starts talking

BTC Pulse splits every trade into three buckets by the size of the aggressing order: retail under $25k, mid-size in between, and whale at $250k and above. Each bucket gets its own CVD line. The moment you do this, the market stops being one anonymous crowd and becomes three groups with very different track records.

One technical detail matters a lot here, and skipping it quietly corrupts the data. Exchanges report fills, not orders. A single $3 million market order can print as three hundred small fills, and if you bucket each fill separately, your whale disappears into the retail line. So we group fills back into logical orders first, by timestamp and side, on every exchange we track. Same rule everywhere: Binance, Bybit, OKX, Kraken, Hyperliquid. Only then do we ask how big the order really was.

Why five exchanges? Because flow migrates. A move that looks whale-driven on one venue and absent everywhere else means something very different from coordinated pressure across all five. The app shows each exchange separately, and its alerts name the venues driving an event.

The one rule the data keeps teaching

When we tested years of this data, one finding kept surviving every attempt we made to kill it: when the whale line and the retail line disagree, you would really rather not be standing with retail.

Price falling whilst whales quietly accumulate has resolved upward far more often than chance would allow. Price rising whilst whales sell into the strength has been one of the most dependable warnings in the entire dataset. Size is the tell, because size implies information, infrastructure and patience that a bus-stop order does not.

“Whale CVD is not a signal. It is a filter. It rarely tells you what to do. It very often tells you what not to believe.”

That distinction matters. A green whale line does not mean up tomorrow. It means the players with the deepest pockets are leaning one way, and reads that fight them carry extra risk. In BTC Pulse the Engine treats whale flow exactly that way: as the most load-bearing input in its read, and as a veto on conclusions that oppose it.

How to read it in the app

  • Pulse tab. The three CVD lines per exchange, on any timeframe. Watch for the whale line bending before price does.
  • Engine tab. The current whale reading over the last hour and four hours, weighed into the overall read, with the exact flip condition stated.
  • Alerts. When whale flow flips sides, surges to top-decile size, or fights a rising price, your phone hears about it once, with the driving exchanges named.

None of this predicts the future, and we would be suspicious of anyone claiming otherwise. It does something more modest and considerably more useful: it tells you what the biggest participants are actually doing, whilst everyone else is busy telling you how to feel.