This is a followup to an old post by nori. I'm posting it for a couple of reasons. I was reminded of it when composing my recent checkerplay quiz, and when I went back and searched for it, sure enough, as Stick predicted, the diagram was broken. So here is a diagram that uses local images.
The other point of interest is that nori said that a rollout was needed here. For example, GNU 2-ply plays 21/12 but puts it ahead of 8/2 7/4 by only a tiny margin, whereas a rollout indicates that covering is a huge whopper. It seems that XG 3-ply understands this sort of position much better. The equity difference between the top two plays might not be quite the same as indicated by a rollout, but at least XG 3-ply says that anything besides running is a whopper.
        
 
        
 | | White is Player 2
score: 3
pip: 121 | | 11 point match | pip: 106
score: 7
Blue is Player 1 |
|
| XGID=a-BBABCBA------bbd-bbB-b--:0:0:1:36:7:3:0:11:10 |
| Blue to play 36 |
| 1. | 3 ply | 21/12 | eq: +0.320 |
| Player:
Opponent: | 59.13% (G:25.20% B:0.46%)
40.87% (G:12.15% B:0.91%) | |
|
| 2. | 3 ply | 8/2 7/4 | eq: +0.206 (-0.115) |
| Player:
Opponent: | 54.63% (G:22.72% B:0.19%)
45.37% (G:7.82% B:0.25%) | |
|
| 3. | 2 ply | 7/1 4/1 | eq: +0.023 (-0.297) |
| Player:
Opponent: | 51.23% (G:14.39% B:0.05%)
48.77% (G:8.17% B:0.35%) | |
|
eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross