Start
What a backtest can prove, and what it cannot
Do this now: Take one strategy you have built or copied. Before you run anything, write three lines - what the rule is meant to capture, the one number you might want to tune, and the date where your held-back data will begin.
Key ideas:
- A backtest answers whether fixed rules would have made or lost money on that history, after costs.
- It cannot tell whether the rules were shaped to fit that history; the grade never knows how many versions you tried.
- Split the history before choosing anything: tune in sample, judge once on the part held back.