How to backtest an ORB strategy

Freeze the rules before viewing results

Document the session timezone, ORB window, break confirmation, retest rule, entry price, stop, targets, late-entry cutoff, and whether both directions can qualify. Changing definitions after seeing losses creates a test that explains the past instead of evaluating a repeatable process.

Use clean intraday data

Check timestamps, missing candles, corporate actions, early closes, and feed coverage. Different providers can print different highs and lows, so record the source and avoid combining datasets without reconciliation.

Track every final outcome

Record qualified wins, stopped trades, no-trade sessions, double breaks, and excluded sessions. A strategy can look stronger when failed or late signals quietly disappear from the denominator.

• TP1 through stretch-target hits
• Stop or invalidation
• Maximum favorable and adverse movement
• Entry and exit timestamps
• Rule-based exclusion reason

Test robustness, not just the headline

Split the sample by year, volatility regime, direction, day of week, and range size. A result that depends on one brief market period may not be durable. Treat out-of-sample performance and execution costs as separate checks before risking capital.

Model costs and candle ambiguity

A candle that touches both a stop and a target does not reveal which traded first unless a smaller timeframe or event sequence is available. Define a conservative tie-breaking rule instead of automatically awarding the favorable result.

Commissions, spread, slippage, partial fills, and market impact can turn a positive gross result into a weaker realized result. Report the raw historical path separately from every cost assumption so another reviewer can reproduce both versions.

Separate development from validation

Use one period to define or tune the rules and a later untouched period to evaluate them. Repeatedly selecting the best timeframe, ticker, filter, and target on the same history increases the chance that the result describes noise rather than a durable relationship.

• Publish the complete rule set before the validation run.
• Keep an audit trail for every exclusion and correction.
• Show trade count beside every percentage.
• Compare gross, cost-adjusted, and out-of-sample results separately.