'Great Success' Rates

It would either be the first case, or the second case with adjusted rates to match the first. This might be the same question about how music performances are decided.

That said Hematin, the samples are significantly imbalanced and by your own admission the data has impurities. You should do at least 100 on thursday with 100 on not thursday for a more appropriate (and statistically significant) comparison.

The rates are specifically different, which is the problem & why we need to record failures. In both cases, the displayed success rate is accurate since a great success is still a success. Not sure what/how you want to adjust. I would expect it to work in the same way as music success, though, if we have the data to figure out which that one is.

I did not read the rng() calls as separate values. Did you mean he should record the failures so we can determine if there are fewer great failures? Otherwise i don't see how recording failures could possibly help determine the method of calculating success; we only have access to the result of calculation unless we dig real deep and determine how many rand() calls the game makes during enchanting.

I don't know if it can be used to determine which version is used (that's the limit of my mathematical ability), but they're important in one of them, so they need to be recorded, anyway.