[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Is XG confidence confident?

Posted By: Nack Ballard
Date: Thursday, 6 May 2010, at 11:19 p.m.

In Response To: Is XG confidence confident? (mtuhtan)

Based on loosely observing thousands of rollout results, I believe Snowie substantially underestimates the size of the confidence interval (i.e., overestimates percentage of confidence). I don't see the Gnu confidence interval numbers in output files (where are they hiding?), but if the intervals are anything like Snowie's, I believe Gnu underestimates the size of the interval as well.

If a similar algorithm (which Tim seems to know more about than I do) is used for XG, then I assume the same paradigm exists (if not more so). For non-to mid-volatile early game plays, Snowie's intervals for 1k, 2k, 5k and 10k (i.e., 1296, 2592, 5184 and 10368 trials) average (+/-) .026, .018+, .013, and .009+, whereas the XG intervals you cited for the 31P-65R-51S play are even lower: .023, .016, .011 and .008.

It seems to me that bots underestimate the confidence interval by conservatively a factor of sqrt2 (i.e., 1.4+), though as I indicated I have not attempted to back that up beyond the anecdotal observation of so many instances of wild results and fluctuations.

Nack

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.