Order flow and CVD: absorption, footprints and data quality

Calculate delta from trades, compare spot and perpetual activity and distinguish a useful observation from a feed or classification artefact.

Course outline

1. 1

   Calculate cumulative volume delta from a trade tape

   Reconstruct CVD from a trade tape: four trades, 0.7 BTC of net aggressive buying. Check side classification, units and reset times before reading a divergence.

   7 min

   In short: CVD is a running sum of signed trade size from a declared start: add buyer-initiated size and subtract seller-initiated size. For quote CVD, multiply each trade by its own price. Trades of +2, −0.5, +1 and −1.8 BTC give a CVD of +0.7 BTC on 5.3 BTC of volume.

2. 2

   Distinguish absorption from exhaustion in order flow

   A price that stops falling can mean buyers absorbing sellers or selling simply fading. Tell them apart with executed flow, replenishment and the price response.

   8 min

   In short: Sell-side absorption is substantial seller-initiated trading with limited downward progress, as passive bids meet the selling. Exhaustion is fading aggressive participation. Both can stop a decline, so compare volume, delta and price response over the same window against normal activity for that market.

3. 3

   Compare spot CVD with perpetual flow correctly

   Compare spot CVD with perpetual flow on matching units, windows and venues. A spot–perp divergence raises questions; it does not locate smart money.

   8 min

   In short: Compare spot and perpetual CVD only on matching windows, quote units, classification and venue sets, and read net delta against total volume: delta ÷ total volume gives a signed imbalance between −1 and +1. A spot–perp divergence can reflect hedging, basis trading or short covering; it does not locate smart money.

4. 4

   Read diagonal footprint imbalances with a sample threshold

   Calculate a diagonal footprint imbalance: 160 against 40 is 4×. Set a minimum sample, handle a zero denominator and check how the volume was classified.

   8 min

   In short: A diagonal footprint imbalance compares buyer-initiated volume at one price with seller-initiated volume one row below: 160 against 40 is 4×. Apply a minimum combined volume, treat a zero denominator as undefined, and check whether volume comes from taker sides or intrabar estimates.

5. 5

   Audit order-flow data before trusting a signal

   Detect missing trades, duplicate events, stale timestamps and mismatched classifications. Build a completeness check before trusting a CVD or order-book signal.

   8 min

   In short: Before trusting a CVD or order-book signal, check that the feed could observe the activity: keep event and arrival times, trade IDs and reconnect boundaries, and leave gaps missing rather than zero. In the example, one lost 1.8 BTC sell turns a CVD of +0.7 BTC into +2.5 BTC.

Open full-size diagram

1. Calculate cumulative volume delta from a trade tape
2. Distinguish absorption from exhaustion in order flow
3. Compare spot CVD with perpetual flow correctly
4. Read diagonal footprint imbalances with a sample threshold
5. Audit order-flow data before trusting a signal

Educational material. Examples do not establish a profitable strategy. Trading costs, gaps and liquidation can produce losses beyond a planned stop.