Skip to crypto research
Crypto Trading Strategy GuideVenue / Contract / Settlement

Crypto Backtest Data and Fill Checklist

Build an auditable experiment around the exact venue, symbol and order assumptions.

Original desk review: 8 October 2026. Examples are fictional scenarios; no live crypto strategy or paid account was tested. Expanded edition: 14 October 2026.

Freeze the Market Identity

Write the exchange, spot or derivative designation, base and quote assets, contract multiplier and timestamp convention at the top of the experiment. BTC quoted in a stablecoin on one venue is not automatically the same execution series as a dollar contract elsewhere. Record data acquisition time, vendor, missing intervals and whether the bars represent trades, a mark or an index. Keep the original export unchanged.

A strategy can appear more consistent when inconvenient markets or delisted assets disappear from the sample. Specify the eligible universe at each historical date where the data permits it. If historical membership is unavailable, label the experiment as a fixed-survivor study and limit the conclusion. Do not call a current list of tradable assets a point-in-time universe.

Declare What Is Known at Decision Time

Suppose the fictional rule checks a completed hourly bar at 12:00 UTC and enters after that bar closes. The closing value can inform the decision, but a fill at the best earlier price inside the same bar cannot. Store decision time, first eligible order time and the permitted fill window separately. A slower data feed or an operational delay should move execution forward, not silently improve the entry.

Indicators must use only observations available at that decision. A missing bar should trigger the stated skip or recovery policy. Filling a gap with future observations creates information that the original operator did not have. Keep feature calculation, entry selection and exit evaluation in distinct records so a later exit experiment cannot rewrite the entry identity.

Resolve Ambiguous Bars Honestly

Imagine a long entry at 100 with a stop at 98 and an objective at 104. A later bar reports a high of 105 and low of 97. That bar does not establish which level came first. Obtain a sufficiently detailed sequence, apply a predeclared conservative rule, or retain an unresolved outcome. Do not choose the favourable sequence separately for each trade.

Likewise, a touched limit price does not establish that your size was filled. Queue position, available liquidity and order arrival are missing from ordinary OHLC data. A partial fill needs a quantity and residual-order policy. Your experiment should say whether it simulates those mechanics or simply uses a research reference price. That distinction matters more than another decimal place in a summary.

Reconcile Friction and Exclusions

Attach a row for each entry fee, exit fee, slippage assumption and relevant funding event. Avoid counting spread twice if your fill series already uses executable bid and ask prices. Conversely, a mid-price series does not already include the spread merely because it is labelled market data. Apply the same cost convention to winners and losers.

Keep expired orders, cancellations and rejected candidates in the experiment log. Removing them after seeing outcomes can turn an operational constraint into selection bias. Separate an exchange outage from a strategy loss, but do not erase the outage when judging deployability. Publish the count of unresolved observations and the exact handling rule.

Reserve Evidence You Have Not Optimised On

Define chronological development and untouched evaluation periods before tuning the rule. Write down the candidate variants tried and the reason for selecting one. A result from a period inspected repeatedly is development evidence even if the file calls it validation. Nearby parameter checks can show fragility but do not create independent evidence.

Finish with a reproducibility bundle: frozen rule, data identity, code version, costs, period boundaries, candidate history and trade ledger. This site provides a checklist, not a certified crypto backtest. No provider in the directory has supplied the complete bundle for an independently reproduced result here. A documented negative or inconclusive experiment is more informative than a favourable chart with no reconstructable inputs.

Primary Sources

Crypto Friction Budget Worksheet

Continue the Research

Add the settlement asset to the fill contract

The expanded settlement cases distinguish a declared full-position price assumption from an actual fill. Linear reference base exposure and inverse fixed quote face use different quantity bases. Both require separate fee assets and funding cash; neither authenticates liquidity, minimum lots or account state. Keep actual unique execution identifiers and partial quantities alongside causal availability. A price printed somewhere in a historical bar is not proof that the intended quantity filled during the permitted observation window. The original checklist remains the starting point; this addition narrows the amount identity rather than supplying missing market data.