When XG is wrong
Posted By: phil simborg
Date: Thursday, 27 February 2014, at 8:15 p.m.
Before XG, when we were using Snowie and GNUBG, I was warned about certain kinds of plays and positions that those bots either got wrong or did not do very well. Even after rollouts, they could not be trusted in their handling of back game play and cubes, and there were certain types of positions with primes they seemed to not understand.
Corbett's book was full of positions the bots got wrong and he proved he could beat the bots in those positions. One of the things I did to prove XG was better was to put 30 of his positions into XG and found that XG got 28 of them right and the other two, very close, and with rollouts, right. (Or as "right" as could be determined by the rollouts and the experts like Corbett.)
Now my question is, and this is for you "experts" or very top players: where don't you trust XG? Now, I already don't trust any evaluation less that XGR++, and I know that if two plays are close with ++ I have to roll them out and sometimes there is a considerable difference. But are they "types" of positions or games where you feel you must always roll them out? Are there types of positions or types of games where you are not sure you even trust the rollouts?
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.