I don't know what exactly this extra fix will entail, but I hope it doesn't end up being too difficult for you.
thanks. it involves having the Dice special effect reinstate the attacking hand to the top bit of $B6 when it ends, so my bugfix's Jump animation code can look at it. istr concluding that the Dice Toss animation ignores that bit, so it won't be a casualty of the change. accounting for (an)other patch(es) will be difficult in the tedious way.
----------
koala_knight: thanks for the info.
----------
At the very least, I was able to calculate that my algorithm reduced the variance on one of the still-skewed dice by a factor of 256.
yeah, that's close enough. who's to say that the weight of the die paint is perfectly balanced out by the lost weight of the die indentations? :P
that said, i think if you make a random # 0-65535 and MOD that by 36 for your "combined roll" value, you will have an even lower variance from ideal. the downside is, because the spare 16 RNG values are going to a _combined_ roll value, Die X is now being tinily weighted
contingent on Die Y's value, and vice versa. whether that oddity matters more than the advantage of getting closer to 1/6ths, i dunno.
iow, you're squaring both the quantity of extra RNG values,
and the much larger denominator.
now, i'm not suggesting you waste time or code space on this method. it's been purely a thought exercise.