RE: [Bug-gnubg] Good beta release! (Suggestions)

[Øystein Johansen]

>12. In a racing situation, it indicated that I should have doubled with a
>pip count lead of 69 to 71 and a two crossover advantage!  This may be
>theoretically correct, but is seems counter-intuitive.  Example attached.

GNU Backgammon is very strong in races and I really think GNU Backgammon is
right to turn the cube in this situation.

GNU Backgammon is probably the strongest bot available when it comes to races.

>16. While playing a well-timed back game, the program likes to cover a blot
>on the 1 point of its own inner board - example attached.  This is
>universally acknowledged to be a bad play.

I think I agree with you in this one. I think the right play is to go from the
midpoint. There is still a lot of training left on these kind of positions.

BTW: have you seen position 99 in "Classic Backgammon Revisited" by Jeremy
Bagai? It discussing position 81 from "Backgammon for Profit" by Joe Dwek. Any
comments on that play? What's you play and what do you think of Jeremys
analysis?

>17. The program missed several obvious hits in a back game - example
>attached.  Perhaps it is still using the regular game analyzer in this
>position?  Or has it given up and is trying to get off the gammon?

Ouch! This is one of GNU Backgammons bad sides. Backgames are not really good
played. If it is any consolation, I can assure you it has been much worse.
Other bot, Jellyish and Snowie doesn't play backgames really good either.
Neural nets are hard to train in these kind of positions.

Thanks again for your comments and suggestions.

-Øystein