results on back testing not matching (optimization and simple test)

 

so, ive noticed this quite a few times - i do an optimization of an EA,

i randomly pick a result that has used specific settings, and profited the tested time period, but when i launch it to see either visually, or without - it fails, thus not representing the result given by optimizer.

genetic optimization is off, and it seems that setting spread to 0 [??] fixes this, but then the question - during optimization what spread is used? i presume the same, so im still quite clue-less.

Reason: