Why are swings from one move been shown as 100% to another as 99.9% certain so commonplace with XG? The one below is just the latest that I rolled out today and the cube is not even in play. If the answer is that with too few trials XG cannot accurately calculate the CI, why report one that is bogus at all?

              
 
              
 | | White is GammonBot_VI
score: 0 pip: 158 | 5 point match Crawford | pip: 163 score: 4
Blue is You | |
| XGID=-a---aE-CA--eD---c-d---AaA:0:0:1:62:4:0:1:5:10 |
| Blue to play 62 |
| 1. | Rollout1 | Bar/23 13/7 | eq: -0.035 |
| Player: Opponent: | 49.12% (G:16.45% B:1.50%) 50.88% (G:16.70% B:0.32%) | Conf: ± 0.028 (-0.063...-0.007) Duration: 1 minute 57 seconds |
|
| 2. | Rollout1 | Bar/23 9/3 | eq: -0.101 (-0.066) |
| Player: Opponent: | 45.89% (G:17.17% B:1.63%) 54.11% (G:16.18% B:0.48%) | Conf: ± 0.019 (-0.120...-0.082) Duration: 1 minute 49 seconds |
|
| |
1 80 Games rolled with Variance Reduction. Dice Seed: 31247 First 2 moves and cube decisions: 4 ply Remaining moves and cube decisions: 3 ply
|
eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross

              
 
              
 | | White is GammonBot_VI
score: 0 pip: 158 | 5 point match Crawford | pip: 163 score: 4
Blue is You | |
| XGID=-a---aE-CA--eD---c-d---AaA:0:0:1:62:4:0:1:5:10 |
| Blue to play 62 |
| 1. | Rollout1 | Bar/23 9/3 | eq: -0.063 |
| Player: Opponent: | 47.76% (G:18.72% B:1.37%) 52.24% (G:15.32% B:0.57%) | Conf: ± 0.004 (-0.067...-0.059) Duration: 1 hour 44 minutes |
|
| 2. | Rollout2 | Bar/23 13/7 | eq: -0.073 (-0.010) |
| Player: Opponent: | 47.39% (G:18.42% B:1.77%) 52.61% (G:16.70% B:0.70%) | Conf: ± 0.005 (-0.078...-0.068) Duration: 2 hours 00 minute |
|
| |
1 4481 Games rolled with Variance Reduction. Dice Seed: 31247 First 2 moves and cube decisions: 4 ply Remaining moves and cube decisions: 3 ply
2 4480 Games rolled with Variance Reduction. Dice Seed: 31247 First 2 moves and cube decisions: 4 ply Remaining moves and cube decisions: 3 ply
|
eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross