| |
BGonline.org Forums
rollout, not that i agree
Posted By: Daniel Murphy In Response To: rollout, not that i agree (Adam Versaw)
Date: Sunday, 20 December 2009, at 6:20 p.m.
These days, given the speed of computers now, and the frequent unreliability of truncated rollouts, I never run them, and generally ignore all truncated results.
Anyone using Gnubg, however, who might not agree that running truncated rollouts is a waste of computer time, should be aware that their truncated results should be better if they assure that ply+depth=even. That is, if ply is even, then the truncation depth should also be even; if ply is odd, then the truncation depth should also be odd. In my experience (which others have confirmed), this has tended to be true, although I have found counterexamples.
As an example of this, take the play 14/10 13/9 6/2*(2) in Adam's position. In Adam's 2-ply depth 11 rollout, this play wins 55% with 30% gammons. But in a 2-ply depth 10 rollout, this play wins 61% with 33% gammons.
To me, that huge difference suggests that it may be best to avoid truncated rollouts entirely. However, I do believe it's likely that if Adam were to do an untruncated 2-ply rollout, his results would be closer to 2-ply truncated at 10 than to 2-ply truncated at 11.
As for the play, assuming that a full 2-ply rollout would favor 14/10 13/9 6/2*(2), I prefer keeping all my checkers in play before opponent's anchors, and making the most desirable point, with 13/5(2).
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.