I love it when good folks complain, "XG only tells us the right play; but it doesn't tell us WHY." I'm an old guy. How old am I? I am old enough to remember playing backgammon back when none of us dreamed that there would ever be BG-playing software within our lifetimes. I am so happy to have XG, which almost always can be used to determine the best play with a rollout, that I would never whine, "But it doesn't tell us WHY." I feel like telling these good folks, "Well, spend a few minutes studying it and try to figure out the WHY."
Why do you take off a checker in one of the positions below but not in the other? I have a great idea. Xavier is very responsive to our requests. Let's ask him to make sure that the next version of XG tells us the "WHY" and not just what the best play is. In the meantime, my only analysis of these positions is the plagiarized line, "XG tells us the right play, but it doesn't tell us WHY."
To figure out the WHY, let's start by looking at the rollout data.
After the best non-hitting play on the left (8/3), Blue is 8.3%, while after the best non-hitting play on the right (8/2), Blue is only 7.8%. Meanwhile, after taking off a checker, Blue is an intermediate 8.1% in both. Blue's reduced winning chance with the deeper spare on the right (2pt vs 3pt) is enough to swing the best play from safe on the left to shot on the right.
Incidentally, if Blue rolls 31 instead of 51 or 41 (and plays 8/4), he is 8.4%. If Blue rolls 53 and takes off a checker (8/0), he is 9.2%. If Blue manages to take off three checkers while maintaining a closed board, he improves to 11.1%, assuming that XG knows how to bear off optimally.
eXtreme Gammon Version: 2.10 |
eXtreme Gammon Version: 2.10 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
