2008-10-05, 08:14 PM
Yes, I've been looking into it the past couple of days. So far my results have led to the 'Experimental version' under your quote.
I can't say it's significantly more accurate than the Japanese version (it's actually less accurate on their samples, but that's a side effect of using regression), but it gets rid of the messy decimals, and never deviates from the Japanese by more than 0.5% on the samples I have.
My main concern with both of them is that they shear off some of the damage ranges. What I mean is, for example, mine predicts 1495~2154 for one sample, but 1492~2144 was observed. If it was correct, the person should never have been able to hit that 1492.
I can't say it's significantly more accurate than the Japanese version (it's actually less accurate on their samples, but that's a side effect of using regression), but it gets rid of the messy decimals, and never deviates from the Japanese by more than 0.5% on the samples I have.
My main concern with both of them is that they shear off some of the damage ranges. What I mean is, for example, mine predicts 1495~2154 for one sample, but 1492~2144 was observed. If it was correct, the person should never have been able to hit that 1492.