Talk:Guides:Beyond average damage

Wow, very nice work. Keep it up. --Xiombarg 05:40, 30 Nov 2005 (GMT)

Are the estimates you cite based on a normal distribution or a binomial distribution? For large sample sizes, the normal does a good job of approximating the binomial, but we're specifically looking at small sample sizes here. --Smoov B 1:06, 18 March 2006 Pacific

The sample sizes aren't nearly large enough to call for a normal approximation to the binomial. It might be worth making some tables. --Theblackgecko 00:45, 19 March 2006 (GMT)

Adding some tables to this page would be easy enough. Even better though, this could be added to one of the firefox plugins. Given your present AP and munitions, what are the odds on a) dealing damage or b) killing a particular target. I don't know anything about making firefox plugins, but I have the math pseudocoded in my head. --Smoov B 11:50, 30 March 2006 Pacific

The two-thirds/95% rules are based on a normal approximation. Obviously the sample size is too small for this to be anywhere near exact, but I wanted to give people something relatively simple that they could do in their heads if need be. IIRC, I tested a few examples with a binomial calculation (that's the "more precise calculation" referenced here and there in the guide) and overall, it seemed to work well enough for decision-making purposes - its "95% confidence interval" might actually be 90% or 99%, but as long as it identifies the best option, you probably don't need to know just how much better it is.

If you're using a computer to calculate odds, it'd probably be better to ignore variances and go straight to the exact binomial calculation (with some sort of sensible cutoff when dealing with large numbers of attacks). I'm not around much these days and can't commit to anything here, so anybody who feels like improving on this guide is very welcome to do so (and to add their name on the credit over at Guides if it seems appropriate). --William Gordon 07:07, 25 October 2006 (BST)