|
This article originally appeared in the December 2001 issue of GammOnLine. Thank you to Kit Woolsey for his kind permission to reproduce it here. |
|
On a good day, an expert can hope to play through an entire game making
at most two or three errors, hopefully none of them serious blunders.
On a real good day, he might struggle through a full game without any
noticeable mistakes. When we achieve this, we finally think that maybe
we are learning something about this game of backgammon.
Reality soon sets in. There comes that game when we make one whopper after another, showing us once again that we really don't understand backgammon at all. The following is such a game. I don't know what was so difficult about it. We weren't involved in a weird backgame or anything like that. The positions were all normal enough. Yet, move after move the proper play eluded both myself and my opponent. I rolled out all the controversial plays on Snowie. The rollout parameters were as follows: 360 trials, 3-ply with variancd reduction, no truncation until data base bearoff, 50% speed, medium search space, cubeless. These results aren't necessarily conclusive, but they are quite likely to be accurate. I will report the cubeless equity of the rollouts, since other numbers involve Snowie's strange biases when taking the cube into account and involve distorted pictures. It should be noted that Snowie also had plenty of trouble with this game. There were several positions where there was a big discrepancy between Snowie's 3-ply opinion and the rollout results, resulting in Snowie's 3-ply evaluation choosing a whopper of an error. Maybe this game is just too tough for carbon and silicon beings alike. Click HERE to see this interesting game. |
| Return to: |