EPTOn May 07 2011 02:53 Cloak wrote:
I'm not saying either/or, just saying that there is a legit way to determine the statistic's validity. I'm just shutting down the notion that we have any idea what's a sufficient sample size without involving math.
http://en.wikipedia.org/wiki/Pearson's_chi-square_test
Point is that the null hypothesis (null= itz balanced guis, itz just randum) gets far less likely when the outcomes get far beyond the standard deviation. I'm too lazy to crunch though. Other people are more inclined.
I'm not saying either/or, just saying that there is a legit way to determine the statistic's validity. I'm just shutting down the notion that we have any idea what's a sufficient sample size without involving math.
http://en.wikipedia.org/wiki/Pearson's_chi-square_test
Point is that the null hypothesis (null= itz balanced guis, itz just randum) gets far less likely when the outcomes get far beyond the standard deviation. I'm too lazy to crunch though. Other people are more inclined.
i'm speaking from poker experience so i could be completely off... but any sample size >4k games had a relatively high %%% chance of being an accurate gauge of one's EV. the thing that makes it tough with SC2 is the vast range of skill between a bronze level nub and some of the top pros. what might appear as "imba" in gold could be easily countered in masters... as a result we have heaps of opinions clashing from all levels of skill into this huge never-ending debate
