Define the investable universe
Say whether the test is for one stock or a changing universe. If you use a list of stocks, ask how membership was determined at each point in time. A list of present-day survivors can make history look easier than it was.
Check the price series
Record the exchange, currency, timeframe and data period. Verify split and dividend treatment. An adjusted close can be useful for a total-return question but may differ from the price at which an actual order could have filled.
Separate signal and execution
State when information becomes available and when the order can fill. A rule that uses a candle’s final closing value should not silently execute earlier in that same candle. Include an indicator warm-up period and do not count it as a tradable test window.
Read costs and exposure
Record commission, spread or slippage assumptions, maximum position size and the treatment of idle cash. A high return with much more market exposure is different from a similar return using less capital.
Use a matched baseline
Compare with a relevant benchmark over the same evaluated dates and initial capital. Inspect the actual trades, drawdown duration and losing periods as well as the final return.
Sources and next steps
- Premiss — supported research workflow ↗AI-assisted strategy creation, historical tests and inspectable code.
- Premiss current plans ↗Product allowances and execution boundaries.
- SEC Investor.gov ↗Public investor education and risk information.