I've been re-checking my data for checker error rates for the bot comparison as I was concerned I'd logged errors which weren't, simply because I failed to spot best moves that didn't make it past the filters used for the analysis.
I'm now using the biggest filters possible and paying more attention to see if even with the bigger filters, the best move didn't make it.
I expected a bigger filter to produce less errors but for 4-ply it seems to have the opposite effect so far. Where there are many close-ish choices XG 4-ply seems to chose some odd plays which the other plies don't. Typically they can be home board preparation plays.
Here's a different type of play where 4-ply goes a little off the rails.
Also interesting is that XG 4-ply's 32nd play (worst as XG only saves max 32) is the play my gnubg rollouts has as the best play 8/6 8/2.
Ok no bot is perfect and the error isn't gigantic but it's odd.
I checked to see if gnubg 3-ply (equivalent of XG 4-ply) suffers from the same problem but it doesn't.
           
    
         
 |  is Snowie4
score: 25
pip: 119 | Money session
Jacoby | pip: 126
score: 17
 is gnubg |
|
| XGID=---BABB-C-AA-Ac-ddBaaaa---:1:1:1:62:17:25:1:0:10 |
| to play 62 |
| 1. | 4 ply | 18/10 | eq: -0.041 |
| Player:
Opponent: | 41.48% (G:0.85% B:0.02%)
58.52% (G:1.41% B:0.03%) | |
|
| 2. | 4 ply | 18/12 6/4 | eq: -0.041 (-0.001) |
| Player:
Opponent: | 41.57% (G:0.93% B:0.02%)
58.43% (G:1.76% B:0.05%) | |
|
| 3. | 4 ply | 13/7 10/8 | eq: -0.043 (-0.003) |
| Player:
Opponent: | 40.84% (G:0.38% B:0.01%)
59.16% (G:0.39% B:0.01%) | |
|
| 4. | 4 ply | 18/12 4/2 | eq: -0.047 (-0.006) |
| Player:
Opponent: | 41.44% (G:0.88% B:0.02%)
58.56% (G:1.88% B:0.05%) | |
|
| 5. | 4 ply | 13/5 | eq: -0.048 (-0.008) |
| Player:
Opponent: | 40.61% (G:0.42% B:0.00%)
59.39% (G:0.40% B:0.01%) | |
|
| 6. | 4 ply | 13/7 3/1 | eq: -0.049 (-0.008) |
| Player:
Opponent: | 40.55% (G:0.54% B:0.01%)
59.45% (G:0.46% B:0.01%) | |
|
| 7. | 4 ply | 13/7 6/4 | eq: -0.051 (-0.010) |
| Player:
Opponent: | 40.53% (G:0.41% B:0.00%)
59.47% (G:0.45% B:0.01%) | |
|
| 8. | 4 ply | 11/5 10/8 | eq: -0.051 (-0.010) |
| Player:
Opponent: | 40.52% (G:0.44% B:0.00%)
59.48% (G:0.46% B:0.01%) | |
|
| 9. | 4 ply | 13/11 10/4 | eq: -0.053 (-0.012) |
| Player:
Opponent: | 40.41% (G:0.44% B:0.00%)
59.59% (G:0.43% B:0.00%) | |
|
| 10. | 4 ply | 18/12 10/8 | eq: -0.054 (-0.013) |
| Player:
Opponent: | 41.10% (G:0.80% B:0.02%)
58.90% (G:1.65% B:0.03%) | |
|
| 11. | 4 ply | 11/5 3/1 | eq: -0.054 (-0.013) |
| Player:
Opponent: | 40.35% (G:0.58% B:0.01%)
59.65% (G:0.49% B:0.01%) | |
|
| 12. | 4 ply | 13/11 8/2 | eq: -0.055 (-0.014) |
| Player:
Opponent: | 40.33% (G:0.39% B:0.00%)
59.67% (G:0.39% B:0.01%) | |
|
| 13. | 4 ply | 13/7 5/3 | eq: -0.056 (-0.015) |
| Player:
Opponent: | 40.33% (G:0.41% B:0.01%)
59.67% (G:0.42% B:0.01%) | |
|
| 14. | 4 ply | 13/7 4/2 | eq: -0.056 (-0.016) |
| Player:
Opponent: | 40.28% (G:0.38% B:0.00%)
59.72% (G:0.38% B:0.00%) | |
|
| 15. | 4 ply | 8/2 6/4 | eq: -0.056 (-0.016) |
| Player:
Opponent: | 40.29% (G:0.45% B:0.00%)
59.71% (G:0.47% B:0.01%) | |
|
eXtreme Gammon Version: 1.11