[Bahamut]

Who's to say that the 36% one is not a potentially misleading stat?

Assuming the 18%/20% stats are correct, it is a 34.4% success rate from the 18% and 20% stat.

The 36% stat may not be out of the realm of possibility with the difference in values here - there may be context missing from all the numbers cited here. They don't seem to disagree with each other significantly though.

[graeme]

I have an explanation. I got the math wrong. I meant to do:

18 + (82 * .2) but I calculated 16.4 and then added 20 instead of 18.