There were multiple attempts at properly capturing the stochastics behind an Action Roll, but as far as I know all of them use either test data (which is pretty good but not precise hurr my mathematics) or are a tiny bit off.
There was an anydice script, but that too isn’t perfect.