| |
BGonline.org Forums
What would it take??
Posted By: neilkaz In Response To: What would it take?? (Chuck Bower)
Date: Tuesday, 25 March 2008, at 5:41 p.m.
g11 was done by Zorba with the last version of unpruned GNU I believe.
Generated from GNUBG Full Rollouts -- 6480 trials (play: 0-ply, cube: 2-ply/25%) by Robert-Jan Veldhuizen
Here's details on Mec26 Automatically generated by Mec26, using values of 26% gammons, 50% winning chances The post-crawford values by Joseph Heled using GNUBGfs rollouts
When I look at Mec26 I am sure that it assumed 100% cube efficiency which I feel is very wrong and distorts scores like -4-2 and even -5-2 and -3-2 IMHO. With 100% cube efficiency, the trailers chances seem to be a bit too good to me, since he has the cube vig.
I don't care for calculated MET's but at far away scores, I think they can be pretty good for cueb decisions since there's decent consistancy, whereas I fear that 6480 trials for g11 is a somewhat too few. My own MET rolled out for a 5 pt match did at least 40,000 trials for each score.
Ian, correct me if I am wrong as it was several years ago, but I recall you doing work on the full 2 ply play MET, rather than the 0-ply play one I did.
I continued my rollout since the D/T decision is not conclusive and less than 1 sd at the moment. I'll stop it later.
Yes..you could extend g11 like you say..and I think David Rockwell may be doing that.
At some point if the mood hits me I'll look at my rolled out 5 pt match MET, compare to g11 (fewer trials) and see if the 11 pt match needs some smoothing while comparing to Mec26 and Woolsey and Jacobs/Trice out to 17 and come up with what I'd think would be a better extension to g11.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.