| White | 104 | ||
|---|---|---|---|
                        | |||
| Blue | 94 | ||
| Position ID: bPcRAhC4220AAA Match ID: cIkWACAACAAE | |||
Note that the blitzing play wins more games, more gammons, and more backgammons while losing fewer gammons. Again, I think the reason is that blue always has a cash after the safe play, but sometimes plays on and loses when blitzing.
I think it's worth reiterating that with the cube in play, weighing cubeless game outcomes (which are what the rollout displays) is irrelevant to choosing the best play. If Blue had already doubled, anything but 8/3(2) 7/2*(2) would be a blunder, since that play would win more games and more gammons (and lose only a trivial number of gammons). But with the cube centered, 8/3(2) 7/2*(2) does not win more games. It does, of course, win more gammons, but not 11.3%. Gammon losses after 8/3(2) 7/2*(2) are only about 0.4%, but after the plays they are nearly zero.
Cubeless, according to the posted rollout:
| Play | Wins | Gammon Wins | Gammon Losses | |
| 9/4(2) 6/1(2) | 80.9% | 1.9% | 0.7% | |
| 9/4(2) 8/3(2) | 85.9% | 2.9% | 0.4% | |
| 8/3(2) 7/2*(2) | 89.9% | 11.3% | 0.3% |
Cubeful, according to my Gnubg rollouts with actual results, and my calculations:
| Play | Wins | Gammon Wins | Gammon Losses | Notes |
| 9/4(2) 6/1(2) | 99.3% | 0.02% | 0.008% | Double/pass after all responses except 5-5.
At the score 6-away 5-away, 5-5 gives White about a 0.940 EMG take, 4-4 about a 1.045 EMG pass, and 6-6 is worse, about 1.070. In a money game, 4-4 gives White about a 0.980 take, while 6-6 is probably a borderline pass. The reason for these perhaps surprising details is that with 5-5 and 4-4 White can play for much improved board distribution -- 13/8 9/4 6/1(2) and 13/5 6/2(2) -- while 6-6 must be wastefully played 13/1 9/3. |
| 9/4(2) 8/3(2) | 99.2% | 0% | 0% | Double/pass after all responses except 6-6 (huge pass after 5-5). Double/take after 6-6, so White wins about (32% * 1/36) = 0.8%. |
| 8/3(2) 7/2*(2) | 93.1% | 8.5% | 0.4% | This play is more complicated. Blue's cubeful losses total about 6.9% (not 10.1%). These come from an immediate 6-1 followed by a loss (about 65% * 2/36 = 3.6%), and from losing after playing on (about 4.8% * 25/36 = 3.3%, of which about 2% will be on a 2-cube or higher). Blue has a small but clear play-on only when White fans, so gammon wins are about 12.25% * 25/36 = 8.5%. On nonhitting aces Blue has a marginal cash. Trading 6.2% additional losses for 8.1% net gammon wins is bad value, and that's why this is not the best play. |