| |
BGonline.org Forums
The Perfect Bot--Will we Every Have one?
Posted By: Keene In Response To: The Perfect Bot--Will we Every Have one? (Phil Simborg)
Date: Tuesday, 30 April 2013, at 9:39 p.m.
Consider this: XG can tell you which plays are wrong, and normally a corollary can be made along the lines of: this play is wrong because it leaves too many shots; this play is wrong because you bear off two checkers less often next turn etc.
If you can definitively state that there are a finite number of reasons for a play to be wrong (which I think we can), and we know that once you eliminate all the wrong plays, then you must be left with the correct play. Then, by definition, you could have a long list of logical reasons why a play could be wrong, and write an opposing reason (why a play is correct) for each of those wrong statements. Thusly, by elimination, its a simple lookup of the one reason why a play isnt wrong, and its corollary idea of therefore why its right.
By elimination of the proper 'wrong plays', you can probably even break down the numbers of the type of game in place (i.e. racing / priming etc), and have themed comments, and a deductive reason and ability to perform your negative lookup.
It might not be easy to write these out, and to properly conceptualize the ideas in english at least, but theoretically its possible.
Of course, this presupposes that backgammon can be absolutely solved, and that your future bot wont be wishy washy with plays (like a backgame for example).
Its an ugly way to do it, but I think it can be done.
Keene
PS: Theoretically, you could have your bot say the reason for a play to be correct is "Because its better than every other option", and that would qualify as perfect too, would it not?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.