Guild Wars 2

I have 1,834 precision. Hero panel shows 47%. Crit Chance formula from wiki says it should be 47.7%. However, I just recorded 15k hits vs subject alpha and tallied up all crits and non-crits. The observed crit chance was 48.1%.

In order for 1,834 precision to be 48.1% crit, the formula has to be (precision – 824)/21 however the wiki says it should be (precision – 832)/21 instead, which doesn’t add up.

Something is not right here…

Ohhhhhh numbers. Lots of numbers. Bad numbers.

I know very little about statistics, so I hope someone will come in and do this right, but just using the first hit on google for a binomial distribution calculator (which actually uses a normal distribution for this many trials), it looks like there’s around a 16.5% chance that you’d get at least that many crits if the actual chance is 47.7%. So maybe the wiki is right, you got a little lucky, and the hero panel just chops off decimals?

(Or maybe something is buggy, if we have conclusive evidence from somewhere else that the hero panel rounds to the nearest integer. But anyway it seems like the wiki formula could still be right.)

K. I’ll do 20k samples every day for a week or so. That should give me a good dataset to work with. If I’m getting 48.1% every time then something is definitely up.

15000 samples and 47.7% critchance result in a expected value of 15000 × 0.477 = 7155 crits and has a standart deviation of sqrt[ 15000 × 0.477 x (1-0.477) ] = 61.

This means the result is between:
7094 and 7216 for 68.3% of the time
7033 and 7277 for 95.4% of the time
6972 and 7338 for 99.7% of the time.

You result of 7215 is in the bounds of the first standart deviation.

(See also this: http://en.wikipedia.org/wiki/Standard_deviation or this: http://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule)