| |
BGonline.org Forums
overview + conclusion
Posted By: Robert-Jan Veldhuizen (Zorba) In Response To: xG evals + rollouts + quirk (Robert-Jan Veldhuizen (Zorba))
Date: Sunday, 5 July 2009, at 11:27 a.m.
That's a lot of numbers to swallow, so I made a quick overview below. xG is definitely NOT a dressed-up GnuBG, as it behaves clearly differently among different settings.
xG evaluations 3/4/5-ply and xG roller all close to each other. Pretty close to GnuBG's 2-ply evaluation too. They all show a bigger pass than GnuBG's rollout.
The xG rollouts on various settings (xG ply-notation) show an interesting picture:
1-ply checker play rollouts: closer to too good, even bigger pass, but only slightly higher cubeless.
2-ply checker play rollouts: lower cubeless, ND close to evaluations, much smaller pass, very close to GnuBG's rollout result(!)
3-ply checker play: this is unexpected, not sure what to think of it. Cubeless much higher, ND much higher, DT way higher. Double Whopper to take, and borderline Too Good.
BTW: I turned Jacoby off, as I always do for money. Certainly for rollouts I think it gives better and more interesting results. This could be important for the ND value.
CL....ND....DT....Settings (cube/checker)
0.667 0.951 1.143 3-ply
0.674 0.948 1.160 4-ply
0.668 0.950 1.148 5-ply
0.666 0.952 1.150 xG roller
0.674 0.978 1.195 1/1-ply rollout
0.677 0.972 1.187 1/3-ply red. rollout
0.675 0.967 1.188 1/3-ply rollout
0.659 0.957 1.091 2/3-ply red. rollout
0.659 0.958 1.088 2/3-ply rollout
0.690 0.993 1.217 3-ply red. rollout
0.693 0.993 1.264 3-ply rollout
?.??? 0.010 0.020 typical 95% CI +/- range
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.