| |
BGonline.org Forums
Plies again
Posted By: Matt Cohn-Geier In Response To: Rollout (ScottyK)
Date: Tuesday, 11 August 2009, at 4:45 p.m.
It is reported that Snowie 3-Ply is same as GNU 2-Ply. It appears that there is substantial deviation.
Yes, this is correct. Snowie uses a different numbering convention than GNU, so GNU's 2-ply is Snowie's 3-ply, GNU's 3-ply is Snowie's (nonexistent) 4-ply, GNU's 4-ply is Snowie's 5-ply, etc. There is substantial deviation because they are different bots.
Is GNU 2-Ply total crap?
No, GNU 2-ply is pretty damn strong...IMHO stronger than all other human players and bots, with the sole exception of myself.
In your example it goes from a 0.052 error to the #1 play by 0.012, not to mention thematically they are very different plays. This is akin to walking into a whorehouse and asking for two big titted tattooed blondes and settling for a single flat chested brunet.
2-ply and 3-ply play very different. It's not a question of "bigger is always better". 3-ply will often come up with the wrong answer (due to the odd/even parity effect in GNU).
However, my rollouts indicate that 3-ply checker is slightly stronger than 2-ply checker, although 2-ply cube is slightly stronger than 3-ply cube. Although these are the strongest settings IMO not involving plies higher than 3, the time difference is such that most people just use 2/2.
I am not sure if 4-ply is significantly stronger than 3/2 or not. It could be that 3-ply checker is better than 4-ply, who knows. But analyzing a match with 3/4 would give you substantially better results than analyzing it with 3/2.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.