Can’t you just run the heuristic over many many runs to build up a probability model? i.e given we simulated a trillion times while varying inputs and we got ma

Can’t you just run the heuristic over many many runs to build up a probability model? i.e given we simulated a trillion times while varying inputs and we got marginal error, we’re 99.99999% certain that this function follows what we observe 100% of the time etc?