Comment by senekerim

It's not really a cheat, as it produces the MLE (i.e. most likely) estimate under certain assumptions (e.g. errors are normally distributed, which occurs naturally if the errors are large sums of many unrelated measurements errors).