Bot Comparison S4 & Gnubg Snowie ERs
Posted By: Timothy Chow In Response To: Bot Comparison S4 & Gnubg Snowie ERs (neilkaz)
Date: Monday, 26 April 2010, at 9:40 p.m.
In Response To: Bot Comparison S4 & Gnubg Snowie ERs (neilkaz)
Once again I have to say that never in my wildest dreams did I expect XGR+ to be 5.5 Elo stronger than 5-ply.
It's worth remembering that we're using GNU rollouts as "ground truth." It is to be expected that XGR+, being based on rollouts, will tend to agree more with GNU rollouts than a 5-ply analysis would, because they're approaching the problem in the same manner. Conceivably, some of the positions that are being labeled 5-ply "errors" are actually being played correctly by 5-ply and misplayed by both XGR+ and the GNU full rollout.
I still expect that XGR+ is stronger than 5-ply but one has to be a little cautious about drawing precise quantitative conclusions because of the above bias.
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.