Indeed it does deserve a good look. I saw 6/4 6/2 and thought that it was probably the kind of thing that would appeal to the bot. It creates a solid 4-point table and (with a superficial examination) didn't look to leave as many blots next go as moving our 8-point's spare into the home. But, like zorba, I couldn't go for it either. A deeply ingrained reluctance to give up the 6-point? A sense of horror at having both the 5-point and 6-point empty? Who knows. ;)
It turns out that 6/4 6/2
is the bot move - at
all evaluation levels.
Cubeful 2-ply
# .. Move ...... Equity . (Diff) ... Win ..... WinG .. WinBg -- Lose ... LoseG . LoseBg
1 .. 6/4 6/2 ... -0.408 ............ 40.0% ... 6.4% ... 0.0% -- 60.0% .. 11.4% ... 0.1%
2 .. 9/3 ....... -0.420 (-0.013) ... 39.5% ... 7.6% ... 0.1% -- 60.5% .. 11.6% ... 0.2%
Cubeful 3-ply
# .. Move ...... Equity . (Diff) ... Win ..... WinG .. WinBg -- Lose ... LoseG . LoseBg
1 .. 6/4 6/2 ... -0.512 ............ 37.9% ... 7.3% ... 0.1% -- 62.1% .. 13.8% ... 0.2%
2 .. 9/3 ....... -0.531 (-0.020) ... 37.4% ... 7.8% ... 0.1% -- 62.6% .. 13.8% ... 0.2%
Cubeful 4-ply
# .. Move ...... Equity . (Diff) ... Win ..... WinG .. WinBg -- Lose ... LoseG . LoseBg
1 .. 6/4 6/2 ... -0.456 ............ 39.6% ... 6.7% ... 0.0% -- 60.4% .. 11.9% ... 0.2%
2 .. 9/3 ....... -0.461 (-0.005) ... 39.1% ... 7.8% ... 0.1% -- 60.9% .. 12.1% ... 0.3%
But when the bot is asked "Do you
really think so?", it demurs and admits that 6/4 6/2 would be a
bad move. Furthermore the equities are much more pessimistic than the evaluations.
Rollout 2592 trials, 2-ply throughout
# .. Move ...... Equity . (Diff) ... Win ..... WinG .. WinBg -- Lose ... LoseG . LoseBg
1 .. 9/3 ....... -0.651 ............ 37.2% ... 7.8% ... 0.1% -- 62.8% .. 13.2% ... 1.2%
2 .. 6/4 6/2 ... -0.707 (-0.056) ... 36.7% ... 7.2% ... 0.1% -- 63.3% .. 14.4% ... 0.9%
Std Err ......... 0.007 ............. 0.1% ... 0.1% ... 0.0% --- 0.1% ... 0.2% ... 0.1%