Hi! I hope some of you will give me some of your insight to this. I have been using gnu to improve my game for a while now, and I have found it really good in most aspects of the game. But lately, I have come across situations, where I could not see why it analyzed as it did. The position below is a moneygame, and for me this should be a CLEAR reject, not a take. Can anyone tell me, why I am wrong?
The score (after 0 games) is: gnubg 0, Sousuke 0
Move number 4: gnubg doubles to 2
| gnubg | 125 | ||
|---|---|---|---|
                         | |||
| Sousuke | 115 | ||
| Position ID: c/aAAQy4bYMFBA Match ID: MBkAAAAAAAAA | |||
• Sousuke accepts
| Cube decision | |||
|---|---|---|---|
| 2-ply cubeless equity | +0.457 | ||
| 0.697 0.123 0.005 - 0.303 0.064 0.001 | |||
| Cubeful equities: | |||
| 1. | No double | +0.713 | |
| 2. | Double, pass | +1.000 | +0.287 |
| 3. | Double, take | +0.708 | -0.005 |
| Proper cube action: | No double, take (1.8%) | ||
Situations like this happens often. I have the level on "world class" with "grandmaster" in cube decisions (3-ply). I then wanted to test, if it should be because of the 21. but nope:
The score (after 0 games) is: gnubg 0, Sousuke 0
Move number 4: gnubg doubles to 2
| gnubg | 99 | ||
|---|---|---|---|
    | |||
| Sousuke | 110 | ||
| Position ID: s22JAAy4c3cAAA Match ID: MBkAAAAAAAAA | |||
• Sousuke accepts
| Cube decision | |||
|---|---|---|---|
| 2-ply cubeless equity | +0.535 | ||
| 0.758 0.027 0.000 - 0.242 0.008 0.000 | |||
| Cubeful equities: | |||
| 1. | Double, take | +0.906 | |
| 2. | Double, pass | +1.000 | +0.094 |
| 3. | No double | +0.876 | -0.031 |
| Proper cube action: | Double, take | ||
Maybe it is me who is worse with the cube than I thought... I also asked one friend who I believe to be a very strong player. We both agreed thaa Gnu is taking a double that should not be taken. I hope you will help answering this doubt of mine. I feel that Gnu is superior with the checkers, but takes like these make me doubt whether its evaluations for the cube makes sense.