GNU 2-ply surprised me by playing bar/22 6/5* here. As some people here have noted, one might consider playing this move against certain opponents, but it's hard to believe that it's technically best. Furthermore, it looks more like an anti-bot move than a bot move. I figured that GNU would repent upon rollout. Due to my limited computer power and the large amount of variance, I have so far rolled out only two plays with GNU, namely bar/22 6/5* and bar/22 24/23. GNU insists that it prefers bar/22 6/5*. Next time I go out of town for a week I might try rolling out bar/22 11/10.
I tried rolling the position out with XG 3-ply. Nine CPU-days later, XG says that bar/22 6/5* is a whopper and that there are four other plausible plays that are much better. Note also the rather large discrepancy in the absolute equity estimates between the two bots. If this position catches the fancy of someone with a powerful computer, I'd be curious to see what an XG 4-ply rollout looks like.
Anyway, as I said before, my main point here is that even long bot rollouts can result in sharply contradictory results, and not just for extremely contrived positions.
| # | Ply | Move | Equity |
| • | 1 | R | bar/22 6/5* | -0.157 |
| | | 0.530 | 0.106 | 0.004 | - | 0.470 | 0.316 | 0.045 | -0.192 | -0.157 | | 0.002 | 0.002 | 0.001 | - | 0.002 | 0.003 | 0.002 | 0.008 | 0.016 | | |
| | Full cubeful rollout with var.redn. | |
| | 15552 games, Mersenne Twister dice gen. with seed 314159265 and quasi-random dice | |
| | Play: supremo 2-ply cubeful prune [world class] | |
| | keep the first 0 0-ply moves and up to 16 more moves within equity 0.32 | |
| | Skip pruning for 1-ply moves. | |
| | Cube: 2-ply cubeful prune [world class] | |
| # | Ply | Move | Equity |
| • | 1 | R | bar/22 24/23 | -0.215 |
| | | 0.487 | 0.103 | 0.004 | - | 0.513 | 0.262 | 0.027 | -0.209 | -0.215 | | 0.002 | 0.001 | 0.001 | - | 0.002 | 0.002 | 0.001 | 0.005 | 0.011 | | |
| | Full cubeful rollout with var.redn. | |
| | 15552 games, Mersenne Twister dice gen. with seed 314159265 and quasi-random dice | |
| | Play: supremo 2-ply cubeful prune [world class] | |
| | keep the first 0 0-ply moves and up to 16 more moves within equity 0.32 | |
| | Skip pruning for 1-ply moves. | |
| | Cube: 2-ply cubeful prune [world class] | |
          
 
          
 | | White is Player 2
score: 0
pip: 160 | Money session
Jacoby | pip: 214
score: 0
Blue is Player 1 |
|
| XGID=--a--aB-C--AdDa--c-eA-A-BA:0:0:1:31:0:0:1:0:10 |
| Blue to play 31 |
| 1. | Rollout1 | Bar/24 8/5* | eq: -0.316 |
| Player:
Opponent: | 46.21% (G:9.07% B:0.40%)
53.79% (G:25.52% B:3.08%) | Conf: ± 0.017 (-0.333...-0.299)
Duration: 1 day 20 hours 37 minutes |
|
| 2. | Rollout1 | Bar/22 24/23 | eq: -0.316 |
| Player:
Opponent: | 45.52% (G:9.12% B:0.42%)
54.48% (G:24.51% B:1.70%) | Conf: ± 0.016 (-0.332...-0.300)
Duration: 1 day 18 hours 07 minutes |
|
| 3. | Rollout1 | Bar/22 11/10 | eq: -0.320 (-0.004) |
| Player:
Opponent: | 45.67% (G:8.60% B:0.38%)
54.33% (G:25.11% B:2.32%) | Conf: ± 0.018 (-0.338...-0.302)
Duration: 1 day 18 hours 56 minutes |
|
| 4. | Rollout1 | Bar/22 8/7 | eq: -0.346 (-0.030) |
| Player:
Opponent: | 45.94% (G:7.84% B:0.36%)
54.06% (G:27.17% B:3.35%) | Conf: ± 0.022 (-0.368...-0.324)
Duration: 1 day 23 hours 29 minutes |
|
| 5. | Rollout1 | Bar/22 6/5* | eq: -0.427 (-0.111) |
| Player:
Opponent: | 45.26% (G:7.00% B:0.32%)
54.74% (G:30.18% B:4.67%) | Conf: ± 0.029 (-0.456...-0.398)
Duration: 2 days 10 hours 58 minutes |
|
| |
1 7776 Games rolled with Variance Reduction.
Moves and cube decisions: 3 ply
|
eXtreme Gammon Version: 1.21
