Order flow measurement: CVD resets, units and timing
Reconcile CVD resets, estimated delta, base and quote units, delayed trades and unknown classifications using original tapes and timelines.
What you will practise
Determine whether an apparent flow signal survives consistent anchors, units, classification, coverage and information availability.
Before you start
- Reconstruct basic cumulative volume delta from signed trades.
- Distinguish an execution, a candle and an open-interest observation.
Course outline
- 1
CVD resets: distinguish a new session from aggressive selling
A CVD line that drops to zero may just have restarted its session. Reconstruct continuous and session-anchored CVD and spot a false −6 reading.
7 minIn short: CVD adds interval deltas (buy-aggressor minus sell-aggressor quantity) from an anchor, so a session reset can drop the line to zero without any trade. Record the anchor and timezone, and read the change between two CVD values as interval delta only within one uninterrupted accumulation.
- 2
Volume delta: compare candle estimates with taker-side trades
Two tools called volume delta can disagree: +10 estimated from candles versus −2 from taker-side trades. Compare both on the same trades first.
7 minIn short: Taker-side delta signs each execution by its aggressor; a candle-direction estimate assigns all intrabar volume by open versus close. On the same four trades the estimate shows +10 and the executed delta −2, a measurement difference rather than a hidden signal or unreported trades.
- 3
Base versus quote volume delta: why the signs can disagree
Base delta counts coins; quote delta weights each trade by its own price. Calculate both from one tape and see how more coins bought can mean more value sold.
7 minIn short: Base delta sums signed quantities (Σs_iq_i); quote delta weights each signed trade by its own execution price (Σs_iq_ip_i). Buying 3 units at 80 and selling 2 at 130 gives base delta +1 but quote delta −20, so the signs can disagree without any data error.
- 4
Orderflow lead and lag: align event time with arrival time
A 100 ms lead in exchange time can become a 180 ms lag on arrival. Replay trade messages as received and keep event, publication and receipt times apart.
7 minIn short: Keep exchange event time, message-generation time and local receipt time separate, and replay a decision only with data received by its cutoff. In the example, trade A leads price update B by 100 ms in event time but arrives 180 ms after it, so an event-sorted backtest leaks information.
- 5
Unknown trade sides: calculate honest bounds for volume delta
When some trades have no aggressor side, report a range, not a guess. Calculate the lowest and highest possible delta; a range that crosses zero has no sign.
7 minIn short: With known buy quantity B, known sell quantity S and unknown-side quantity U, total delta lies between B − S − U and B − S + U. A range that crosses zero leaves the sign unresolved, and the bound is deterministic, not a confidence interval.
Related courses
- Order flow and CVD: absorption, footprints and data qualityCalculate delta from trades, compare spot and perpetual activity and distinguish a useful observation from a feed or classification artefact.
- Volume profile and TPO: audit how profiles are builtReconstruct POC ties, value-area coverage, composite sessions, TPO counts and price-grid effects using original, checkable datasets.
- Order book liquidity and liquidation heatmapsIdentify what a heatmap measures, reconcile walls with trades, calculate slippage and evaluate sweeps without assuming a guaranteed reversal.
Open full-size diagram- CVD resets: distinguish a new session from aggressive selling
- Volume delta: compare candle estimates with taker-side trades
- Base versus quote volume delta: why the signs can disagree
- Orderflow lead and lag: align event time with arrival time
- Unknown trade sides: calculate honest bounds for volume delta
Educational material. Examples do not establish a profitable strategy. Trading costs, gaps and liquidation can produce losses beyond a planned stop.