Pedronicus wrote:I left this site because I was sick of the dice. I posted info after manual dice counting and was dismissed for not having a big enough sample to accurately determine a conclusion.
Since I rejoined, I got dice analyzer working. 11K throws should be a decent enough sample of throws and every dice should be around 16.6%
But if you can't defend against a 6 and attacking with a 1 won't win, guess whats being thrown by me and the defender?
correct....
Attacker dice distribution Defender dice distribution
1s 1934 / 11250 (17.19%) 934 / 5716 (16.34%)
2s 1867 / 11250 (16.6%) 971 / 5716 (16.99%)
3s 1918 / 11250 (17.05%) 965 / 5716 (16.88%)
4s 1797 / 11250 (15.97%) 967 / 5716 (16.92%)
5s 1855 / 11250 (16.49%) 900 / 5716 (15.75%)
6s 1879 / 11250 (16.7%) 979 / 5716 (17.13%)
actually 11k really isn't a large enough sample. you'll see variation for even 100k throws. you need a huge sample to get everything at .166
Though a 95% confidence interval does contain all your values
attacker (.1591, .1728)
defender (.1564, .1756)
which shows that the dice are most likely following the proper distribution. We're 95% confident a distribution where .166 is the p^ falls into this interval.
how to calculate:
p^ +- z(alpha/2) * sqrt[(p^)*(1-p^)/n]
for the attacker:
.166 +- 1.96 * sqrt(.166*.834/11250) = .166 +- sqrt(.000012306) = .166 +- 1.96(.00350801) = .166 +- .006875699
lets round to .166 +- .00688 = (.15912, .17288)
for the defender:
.166 +- .009645995 = (.1564, .1756)