Binance cumulative volume delta: trade direction, volume and data gaps

By Cruscy Data · 2026-10-30

Cumulative volume delta, or CVD, accumulates the difference between buyer-initiated and seller-initiated traded volume. It describes an observed trade-flow imbalance. It is not the same as the imbalance of resting bids and asks, and it does not by itself establish what price will do next.

Get the direction convention right

On Binance, the buyer-maker flag identifies whether the buyer was the resting maker. When the buyer is the maker, the seller is the aggressive side; treating that flag as an aggressive-buy flag reverses the sign of delta.

Normalized data products may expose a direction field instead of the raw exchange flag. Read the field dictionary for the specific stream before applying a formula. Do not copy a side mapping from a different exchange or a differently normalized table.

Choose units and a reset rule

Base-volume delta measures the traded asset, such as BTC. Quote-volume delta measures notional, such as USDT. A cumulative series also needs a stated starting point: the beginning of a session, a UTC day, or an explicit research window. Two charts with different units or resets are not directly comparable.

The calculation has two stages: add positive volume for the buyer-initiated trades and negative volume for the seller-initiated trades, then cumulatively sum those signed amounts in event order. Keep the trade identifier and time conventions available so that you can detect duplicate or missing records.

Separate missing data from quiet trading

An interval with no observed prints may be a quiet interval or a recording problem. Check source coverage before representing it as zero traded volume. Deduplicate overlapping recordings before aggregation; counting the same trade twice can create a convincing but incorrect flow signal.

Cruscy Data provides the spot BTC trade tape and separate USD-M futures trades. Those markets have different participants and books, so maintain separate CVD series. The derived feature stream also exposes CVD; check its schema for the product's precise aggregation convention.

Keep the underlying records inspectable

Start with a short window, compare signed volume with total volume, and inspect the original prints around large changes. Use SQL to answer those questions before using the series as a model input. The order-book guide explains why a trade-flow measure and a liquidity measure provide different information.