| |
BGonline.org Forums
A better approch?
Posted By: eXtreme Gammon In Response To: A better approch? (Bob Koca)
Date: Sunday, 1 November 2009, at 10:03 p.m.
What I meant is that comparing the value game per game gives you a better approach to determined how each move compare to each other, but it does not help to give the odd of each move being the best.
in the example I gave, I have the following result 6/2 5/4 better than 6/1*: delta=0.0103 s=0.2713 prob=68.5% 6/2 5/4 better than 13/8: delta=0.0144 s=0.2933 prob=73.3% 13/8 better than 6/1*: delta=0.0041 s=0.2362 prob=58.7%
Compare to the number given by assuming independence and normal: repectivly 71.5%, 77.2% and 58.3%... that's quite close, except for the 1st one.
However to determine the best move using which one came on top for each 162 games is not efficient: in that case 6/2 5/4 and 6/1* came up on top 55 times each and 13/8 52 times.
the raw data can be downloaded from www.extremegammon.com/misc/rawRoresults.txt (format: Game #, 6/1*, 13/8, 6/2 5/4)
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.