| |
BGonline.org Forums
The right way to measure which player played better in a match
Posted By: Maik Stiebler In Response To: The right way to measure which player played better in a match (Casper van der Tak)
Date: Wednesday, 1 June 2011, at 5:19 p.m.
Unbiasedness is the property Tim is describing in his OP. Total error difference in MWC is unbiased if given by a perfect bot. Currently, the only easy way to get unbiased skill estimates from real bots (gnubg and XG, specifically) is asking them for luck adjusted results. Those have some drawbacks: They aren't, in an obvious way, related to the error analyses done by the same bots, and they are much more noisy than the total error differences in MWC. I think for those reasons most players don't care about luck adjusted results. In this post (is that the thread you were referring to?) I hinted at a procedure to coax an unbiased estimate out of real bots that suffers much less from the mentioned drawbacks - it has other drawbacks though, namely that it is much more expensive to compute than the luck adjusted result (that shouldn't matter too much if you are doing a high quality analysis of the match anyway), and of course that it is not implemented in any bot.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.