| |
BGonline.org Forums
Slight bot content
Posted By: Timothy Chow In Response To: DMP backgame checkerplay decision (Stick)
Date: Tuesday, 9 March 2010, at 3:34 p.m.
Stick wrote:
13/11 6/3 would be my first choice.
Then maybe you will care what the rollout says because this is GNU 2-ply's first choice.
It looks like there is a gap in the backgammon literature here. As you hinted at, one should probably work backwards, starting with a more advanced stage of the backgame, where the anti-backgamer has one last chance to hit. A rollout of such a position will probably be fairly reliable because there probably won't be any difficult decisions after that one for a while. After tinkering with a few variants of that position to get a good sense of when it's right to hit and when it's not, one can wind back a step, seeing if 2-ply or 3-ply makes correct decisions about hitting. And so on.
Some experts have probably gone through this exercise privately but I've never seen anything published along these lines. In my opinion, this kind of thing is where the future of backgammon analysis lies. Ballard and Weaver have convincingly demonstrated what an excellent book one can produce by taking a bunch of related and commonly-occurring checker-play decisions and showing that what's right in Position A is wrong in Position B and vice versa. There's no reason this approach can't be applied to other situations. Checker-play problem books like New Ideas in Backgammon and Boards, Blots, and Double Shots are very good, but they roll out only one position at a time instead of a suite of related positions. Now that bots are so much faster than they used to be, carefully chosen suites of positions are much easier to generate and I'm positive that they will yield enormously valuable new insights.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.