A backtest replays a strategy against historical data to estimate how it would have performed. Its usefulness depends almost entirely on whether the data reflects what was actually observable and executable at each moment, rather than what is knowable in hindsight.
The three failures that account for most inflated backtests are look-ahead (using information that did not exist yet), fill assumptions that ignore spread and depth, and resolving markets against a different price basis than the exchange used.
The third is specific to this venue and easy to miss: resolving historical crypto cycles against spot instead of the 60-second TWAP produces a strategy that looks profitable and is not.
Look-ahead that does not look like look-ahead — for example filtering to markets that ended up liquid, or that resolved cleanly. The filter itself uses information that did not exist when the trade would have been placed.
Browse the full glossary (32 terms), the strategy database, or the build guides.
