| |
BGonline.org Forums
XG roller really seems super strong..and...
Posted By: neilkaz
Date: Wednesday, 16 September 2009, at 3:20 a.m.
Noting how well XGR performed on the first 100 Depreli positions...ie better than anything else in MCG's test. I am checking out XG for the next 100 positions.
I have the first 25 done and XGR made a total error of .097 !! All of that being on the 21 checker plays.
Xavier says "XG roller makes a rollout in 1-ply for 3600 games. The first two cube decisions are made in 2-ply. It is truncated after 6 moves (so 6-ply). It truncates to the bearoff database if the cube is dead."
Neil asks does XGR use varience reduction?
Neil points out that those using XG should set the 3 ply search space to huge since XG is superfast anyhow and the default search space failed to pick up the correct play ( I had to put plays 6-10 thru 3 ply manually) whereas huge found it. Options/settings/analysis search interval.
Neil asks under options/settings/analysis ... how much cache should I set for fastest performance?
Neil states that he is quickly becoming of the opinion that use of XGR roller to analyse matches may become the new standard.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.