Repeatable replay

Understand which inputs make a replay repeatable and what repeatability cannot prove.

Repeatability means using the same recorded inputs, engine version, initialization and settings to reconstruct the same market sequence. It lets you revisit an observation and compare interpretations against a shared record.

Keep the inputs fixed

The replay engine reconstructs trades, quotes and supported derived displays from captured data. A different engine version, warmup window, data coverage or setting can change a derived value. Browser rendering and frame timing are not themselves evidence of identical market state.

Captured exchange publications are also not every event that occurred at the venue. Liquidation publication limits, gaps and missing streams remain limitations when data is replayed faithfully.

Repeatability and selection are different

An author can select a dramatic event even when its replay is perfectly repeatable. A lesson can illustrate a mechanism without demonstrating its frequency or predictive value. Compare selected examples with research counts and baselines.

Research receipts

A Research receipt identifies the query and versions behind a result. Preserve the reported dataset revision and pinned artifact; rerunning against a newer record can produce a new result. The reproducibility reference owns the exact contract.

Start a replay, or browse published reports to inspect a recorded study.