My reasoning was identical to that of David Presser. We have a questionably timed backgame (most of the timing is achieved via deep spares which we might not be able to recirculate due to white prime), so it looks like our main priority is to not let white complete the full prime. Hence bar/23 14/13*. Here are the results of the rollout:
          
 
         
 | | White is Player 2
score: 0
pip: 124 | Unlimited Game
Jacoby Beaver | pip: 211
score: 0
Blue is Player 1 |
|
| XGID=-----ACbBAA--aA-babcbbC-BA:1:1:1:21:0:0:3:0:10 |
| Blue to play 21 |
| 1. | Rollout1 | Bar/23 10/9 | eq: -0.473 |
| Player:
Opponent: | 37.63% (G:3.25% B:0.16%)
62.37% (G:37.34% B:5.58%) | Conf.: ± 0.011 (-0.484...-0.462) - [100.0%]
Duration: 57 minutes 51 seconds |
|
| 2. | Rollout1 | Bar/23 6/5 | eq: -0.522 (-0.049) |
| Player:
Opponent: | 37.51% (G:2.51% B:0.11%)
62.49% (G:39.33% B:8.25%) | Conf.: ± 0.013 (-0.536...-0.509) - [0.0%]
Duration: 1 hour 03 minutes |
|
| 3. | Rollout1 | Bar/23 24/23 | eq: -0.551 (-0.078) |
| Player:
Opponent: | 36.40% (G:3.30% B:0.17%)
63.60% (G:43.20% B:5.28%) | Conf.: ± 0.013 (-0.565...-0.538) - [0.0%]
Duration: 1 hour 02 minutes |
|
| 4. | Rollout1 | Bar/22 | eq: -0.584 (-0.111) |
| Player:
Opponent: | 36.08% (G:2.08% B:0.08%)
63.92% (G:41.47% B:7.93%) | Conf.: ± 0.013 (-0.597...-0.571) - [0.0%]
Duration: 56 minutes 21 seconds |
|
| 5. | Rollout1 | Bar/23 14/13* | eq: -0.672 (-0.199) |
| Player:
Opponent: | 33.42% (G:2.30% B:0.11%)
66.58% (G:42.19% B:9.46%) | Conf.: ± 0.016 (-0.688...-0.656) - [0.0%]
Duration: 1 hour 12 minutes |
|
| |
1 1296 Games rolled with Variance Reduction.
Dice Seed: 36410945
Moves: 4-ply, cube decisions: XG Roller+
|
eXtreme Gammon Version: 2.10
I can try to explain this post factum as follows: since white is about to complete the prime, having more checkers sent back increases our gammon losses. So, we should try to not let white hit a lot of our blots.
However, to be honest, I don't really understand that well what is going on here. That's why I was hoping some of our resident experts would explain this position to us.