In my opinion, a backgame position requires a massive simulation with 3 ply settings. An example:
| 1. | Rollout1 | 8/4 6/4(2) | eq: -0,812 |
| Player:
Opponent: | 29,97% (G:0,68% B:0,03%)
70,03% (G:44,15% B:12,11%) | Conf: ± 0,010 (-0,822...-0,802)
Duration: 5 hours 45 minutes |
|
| 2. | Rollout1 | 9/7* 7/5 6/2 | eq: -0,815 (-0,003) |
| Player:
Opponent: | 31,73% (G:0,98% B:0,05%)
68,27% (G:45,27% B:16,11%) | Conf: ± 0,014 (-0,829...-0,801)
Duration: 9 hours 33 minutes |
|
| 3. | Rollout1 | 6/4(2) 6/2 | eq: -0,836 (-0,024) |
| Player:
Opponent: | 28,23% (G:0,73% B:0,03%)
71,77% (G:43,76% B:10,38%) | Conf: ± 0,009 (-0,845...-0,827)
Duration: 4 hours 31 minutes |
|
| 4. | Rollout1 | 8/2 6/4 | eq: -0,853 (-0,041) |
| Player:
Opponent: | 27,91% (G:0,70% B:0,03%)
72,09% (G:44,44% B:10,50%) | Conf: ± 0,010 (-0,863...-0,843)
Duration: 4 hours 28 minutes |
|
| 5. | Rollout1 | 8/6 8/2 | eq: -0,877 (-0,065) |
| Player:
Opponent: | 27,05% (G:0,55% B:0,03%)
72,95% (G:44,83% B:10,26%) | Conf: ± 0,009 (-0,886...-0,868)
Duration: 4 hours 00 minute |
|
| 6. | Rollout1 | 8/4(2) | eq: -0,888 (-0,076) |
| Player:
Opponent: | 26,62% (G:0,51% B:0,02%)
73,38% (G:45,16% B:10,00%) | Conf: ± 0,008 (-0,896...-0,880)
Duration: 3 hours 05 minutes |
|
| |
1 10368 Games rolled with Variance Reduction.
Moves and cube decisions: 3 ply
|
I cranked up to 40k, and then "the phantom prime play" came in as first, 0.003 better, with a confidence interval of 0.001. The predefined 3ply with variance reduction is good, i think. And i crank up, until the added confidence intervals of the two best candidates is lower then their difference. So i will put Pos. 1 with my stone aged core 2 duo under the microscope.
