Why does my backtest not match the replay?

Usually one of three things. First, data resolution. If the backtest is evaluating on a coarser series than the replay, the two see different intrabar paths and can resolve a stop-versus-target race differently. Second, fill assumptions — check that both are using the same rule for what price you'd have received. Third, any state that carries between runs, such as a position left open at the boundary of the period. If you can tell us which direction the difference goes and whether it's consistent, that usually narrows it to one of the three quickly.

The intrabar path explanation is the one that got me. Coarser data can't tell you which came first.

Also check WHICH bar your entry fires on! Backtest might fill on the signal bar close, replay waits for next bar open. That one gap ate me for weeks!!

One more to add to TraderWe's list: indicator warm-up. If the backtest loads more history before the day starts, your moving averages aren't at the same value as the replay at open. Same formula, different starting state.

ok honest question, if the two never fully agree, how do people decide which number to trust for like a real decision? do you just take the worse of the two and assume that's your floor? still fuzzy on this

Neither, probably. Both are approximations, and if the edge only shows up in one of them it wasn't much of an edge. Another perk of holding for days is I stopped caring which tick came first at 10:31.