There are two rollouts posted, my original which was extended from 31k >> 46k and the gammons lost for the S play went from 11.7% to 12.5% in 15k trials and David Rockwell's soon to be finished 15k rollout of his own.
He started the rollout because he saw my posted rollout of 31k and thought the S play wasn't losing enough gammons and he was a little relieved when I told him I'd extended it to 46k and the gammons had jumped up. Still, if you look at his rollout the S play is losing significantly more gammons still.
This isn't the biggest concern though. The rollouts are different settings, his being supremo and mine being modified world class, I don't believe that will have much of an effect. I have ran a few tests to test this before, the tradeoff between speed and accuracy. The biggest scare in my opinion is the absolute equitites. Could you imagine if you David's rollout of the S play extending it to 46k and it stayed in the same ballpark it is now, -0.2870, which I believe it will, and compared it to the R play of my rollout which has an equity of -0.2600, the difference would be huge at .027!
There is very little variance left in these rollouts according to GNU to account for this massive difference. (there's a lot less variance in pure AtS rollouts since the cube is never in play) We need to know what's going on here? The best guess at the moment is not to mix seed results. When I started doing these lengthy opening game rollouts it wasn't a concern of mine but given some of the oddities that have arisen with GNU of late I try to avoid it at all times.
Most of you will never need worry about this, you have a question about which play is better, you highlight two or three plays and you roll them out, they are assigned the same seed, no worries. What happens though is if you roll out two plays for example and later decide to add other candidates. You've rolled out the two original plays, saved the rollout, closed the rollout and now decide maybe there's a third or fourth play that should have been considered. You open the rollout, highlight the new candidate plays and launch the rollout. Whoops, GNU sticks in a new seed number and the seed number of play 3 and 4 will not match the original seed number of plays 1 and 2.
This isn't a big deal now that I'm aware of it, you can manually change the seed before adding any other plays, however it still might be the cause of the difference in rollouts like these two. It has caused me to reroll a few plays already matching up the seed numbers.
The score (after 0 games) is: gnubg 5, user 6 (match to 7 points, Crawford game)
| gnubg | 167 |
|---|---|
| |
| user | 167 |
Position ID: 4HPwATDgc/ABMA Match ID: 8An6AFAAMAAA | |
| # | Ply | Move | Equity | |||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | R | 24/18 13/9 | -0.2590 | |||||||||||||||||||
| ||||||||||||||||||||||
| Full cubeful rollout with var.redn. | ||||||||||||||||||||||
| 46656 games, Mersenne Twister dice gen. with seed 1045316273 and quasi-random dice | ||||||||||||||||||||||
| Play: 2-ply cubeful prune [world class] | ||||||||||||||||||||||
| keep the first 0 0-ply moves and up to 12 more moves within equity 0.16 | ||||||||||||||||||||||
| Skip pruning for 1-ply moves. | ||||||||||||||||||||||
| Cube: 2-ply cubeful prune [world class] | ||||||||||||||||||||||
| • | 2 | R | 24/14 | -0.2600 ( -0.0011) | ||||||||||||||||||
| ||||||||||||||||||||||
| Full cubeful rollout with var.redn. | ||||||||||||||||||||||
| 46656 games, Mersenne Twister dice gen. with seed 1045316273 and quasi-random dice | ||||||||||||||||||||||
| Play: 2-ply cubeful prune [world class] | ||||||||||||||||||||||
| keep the first 0 0-ply moves and up to 12 more moves within equity 0.16 | ||||||||||||||||||||||
| Skip pruning for 1-ply moves. | ||||||||||||||||||||||
| Cube: 2-ply cubeful prune [world class] | ||||||||||||||||||||||
| # | Ply | Move | Equity | |||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | R | 24/14 | -0.2801 | |||||||||||||||||||
| ||||||||||||||||||||||
| Full cubeful rollout with var.redn. | ||||||||||||||||||||||
| 12534 games, Mersenne Twister dice gen. with seed 813463969 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] | ||||||||||||||||||||||
| • | 2 | R | 24/18 13/9 | -0.2870 ( -0.0069) | ||||||||||||||||||
| ||||||||||||||||||||||
| Full cubeful rollout with var.redn. | ||||||||||||||||||||||
| 12533 games, Mersenne Twister dice gen. with seed 813463969 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] | ||||||||||||||||||||||