I was playing XG here, and actually XG was playing Blue, so I did not have the privilege of blundering these decisions OTB. However, I was surprised enough by the bot's plays to study the positions further.
In the first position, XG played bar/24*/22. I would have thought the double hit was a next and that anything else was a blunder. I suppose a longer and stronger rollout might put the double hit on top, but presumably bar/24*/22 has some merit to it? I don't see it.

              
 
              
 | | White is Player 1
score: 0 pip: 156 | Money session Jacoby Beaver | pip: 150 score: 0
Blue is Player 2 | |
| XGID=-B-b-Da-E---ba-b--bbBb-AaA:0:0:1:21:0:0:3:0:10 |
| Blue to play 21 |
| 1. | Rollout1 | Bar/24* 24/22 | eq: -0.310 |
| Player: Opponent: | 41.46% (G:12.69% B:0.38%) 58.54% (G:13.33% B:0.58%) | Conf: ± 0.013 (-0.323...-0.297) Duration: 15 hours 15 minutes |
|
| 2. | Rollout1 | Bar/24* 8/6* | eq: -0.339 (-0.029) |
| Player: Opponent: | 40.54% (G:14.75% B:0.52%) 59.46% (G:14.76% B:0.69%) | Conf: ± 0.014 (-0.353...-0.325) Duration: 15 hours 22 minutes |
|
| 3. | Rollout2 | Bar/22 | eq: -0.353 (-0.043) |
| Player: Opponent: | 41.03% (G:9.84% B:0.30%) 58.97% (G:13.33% B:0.46%) | Conf: ± 0.012 (-0.365...-0.341) Duration: 7 hours 18 minutes |
|
| |
1 5184 Games rolled with Variance Reduction. Dice Seed: 2 Moves and cube decisions: 3 ply
2 3888 Games rolled with Variance Reduction. Dice Seed: 2 Moves and cube decisions: 3 ply
|
eXtreme Gammon Version: 1.21
In the second position, I figured that all three plays would be close, and I think I would have probably played 22/17 16/13. I can accept that it's not the top play, but why is it so far behind 22/14? With hindsight, I can see that 5's play poorly, but is that enough to account for the magnitude of the difference? Maybe Stick's comment about the cube decision before the roll is part of the explanation (these positions originally arose in a match where I think it was correct not to double before the roll).
I originally rolled this out using XG1 but it seemed that XG was screwing up the cube decision on the next roll if White danced, so I asked a friend to roll it out with XG2 (shown below). He tells me that XG Roller assesses the cube decision after 22/17 20/17, dance as D/P but that a rollout assesses it as D/T. Assuming D/T is correct, the equity of 22/17 20/17 below is presumably inflated, so that 22/17 20/17 and 22/14 are probably almost equally strong.
After 22/14, dance, both XG Roller and a full rollout say it's D/P, and similarly after 22/17 16/13, dance, both XG Roller and a full rollout say it's D/T.

              
 
              
 | | White is Player 1
score: 0 pip: 171 | Money session Jacoby Beaver | pip: 126 score: 0
Blue is Player 2 | |
| XGID=aBBbBB--Ca--b--bA-bcBbA---:0:0:1:53:0:0:3:0:10 |
| Blue to play 53 |
| 1. | Rollout1 | 22/17 20/17 | eq: +0.654 |
| Player: Opponent: | 61.04% (G:31.82% B:0.86%) 38.96% (G:9.00% B:0.41%) | Conf: ± 0.011 (+0.643...+0.665) Duration: 57 minutes 35 seconds |
|
| 2. | Rollout1 | 22/14 | eq: +0.632 (-0.022) |
| Player: Opponent: | 59.67% (G:32.06% B:0.97%) 40.33% (G:8.17% B:0.37%) | Conf: ± 0.008 (+0.624...+0.640) Duration: 34 minutes 37 seconds |
|
| 3. | Rollout1 | 22/17 16/13 | eq: +0.570 (-0.084) |
| Player: Opponent: | 59.15% (G:30.49% B:0.88%) 40.85% (G:8.39% B:0.37%) | Conf: ± 0.010 (+0.560...+0.580) Duration: 54 minutes 46 seconds |
|
| |
1 5184 Games rolled with Variance Reduction. Moves: 3 ply, cube decisions: XG Roller
|