| |
BGonline.org Forums
truncated@6 won't be a heck of a lot better...
Posted By: Robert-Jan Veldhuizen (Zorba) In Response To: You mean XG Roller setting? Couldn't it be made a bit stronger..ie it is so fast (Neil Robins)
Date: Sunday, 12 July 2009, at 10:50 p.m.
Just from the settings itself, it seems clear that xG roller will show marginal improvements at best.
Any kind of strategical errors, which are still all bots' major weak spot, will not often get reversed in truncated at 6 rollouts, just as GnuBG 4-ply (essentially a truncated@4 rollout, no VR, 1679616 trials) does not improve much at all over GnuBG 2-ply.
Also, the standard errors of the xG roller rollout are high enough that it can sometimes lead to xG roller picking inferior moves because of statistical uncertainties.
Then there's also the fact that in these 6 moves, xG will play the checkers at 1-ply, which could introduce errors in the follow-up. It could also make wrong cube actions at 2-ply resp. 1-ply, inside the rollout.
I expect that xG roller won't do anything at all for most backgame problems and most standard opening/middle game play. However, it could show some improvement in containment situations, blot hitting contests and perhaps, some blitz problems and mutual priming games approaching their climax.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.