Stock backtesting that respects the data

Define the historical stock universe

A current watchlist can contain only companies that survived to today. Testing that list in the past can omit delisted, acquired, or failed companies and create survivorship bias. State whether the test covers one stock, a fixed list, or a point-in-time index membership. For screen-based strategies, record when fundamental or membership data became available rather than using a later revision.

Control price adjustments and execution assumptions

Document whether the data is adjusted for splits and dividends and whether the strategy claim is about price return or total return. Match order timing to information available at the decision bar. Add commissions and slippage, then inspect whether low-priced or thinly traded stocks would have supported the simulated position size.

• Record adjusted or unadjusted price treatment.
• Model costs and liquidity assumptions.
• Avoid future index membership or revised data.

Validate across stocks and market conditions

A rule that works on one familiar ticker may reflect that ticker rather than a general stock-market effect. Test the same frozen logic across relevant stocks and periods, then keep an untouched time range for confirmation. Compare trade distribution and drawdown rather than selecting the symbol with the highest net profit.