Cluster Volume and Order Flow: From Price-Time Concentration to Trading Features

Cluster analysis asks where unusually large activity occurred simultaneously in price and time. For automation, the important step is converting that visual observation into normalized, testable features.

Price and time together

A cluster is not simply a high-volume candle. It identifies concentrated activity around particular price levels during a defined interval. The same raw number can mean very different things across instruments and sessions, so normalization is essential.

Aggressive and passive flow

Trades executed at ask and bid can be used to describe aggressive buying and selling, while the resulting price response helps reveal whether that aggression was accepted or absorbed. A large aggressive print near an extreme should therefore be evaluated together with subsequent price movement.

Useful bot features

Candidate features include volume percentile, bid/ask imbalance, distance from swing extreme, distance from a structural zone, post-cluster displacement, rejection size, session, volatility regime and whether the event occurred during a retest.

Do not use raw thresholds blindly

A fixed volume threshold is fragile across symbols and contract periods. Prefer rolling percentiles or z-scores with a minimum sample requirement. If the feed is incomplete or the sample is too young, the order-flow feature should be marked unavailable rather than guessed.

Role inside a strategy

Cluster volume is best tested as confirmation or a veto around an existing setup. It should not bypass the structural stop, portfolio gate or execution approval process.

Automation pipeline

Educational software-engineering material only. Market heuristics are models to validate, not investment advice or guarantees of profitability.