Gian-Carlo Pascutto wrote: > Magnus Persson wrote: >> Quoting Don Dailey <[EMAIL PROTECTED]>: >> >>> Just to make it clear, the case we want to "fix" is the case where many >>> bots are programmed to resign. Lazarus will resign when the score is >>> below 1% (and has remained so for a couple of moves in a row which is >>> probably just a superstition on my part to delay it.) >> >> Valkyria is an aggressive resigner and does so below 10% already, and >> I have no memory of it resigning in a position where it was >> meaningful to play on. > > When playing against humans, there's the question of what to do if > winrate is <10%, we have 15 minutes left on the clock, and the (human) > opponent has less than 1 minute left and we're playing absolute time. > > Leela right now resigns, which is nice. But maybe too naive. I think we all have some elements of opponent modeling in our programs. One thing you could do is to play the game without resigning. Even this is a form of opponent modeling because if you play to the end your opponent might get angry and make a mistake, or even if it's another computer there is always the chance that the computer will crash or something.
If you detect that your opponent is short on time and react to this by playing on, it is also a weak form of opponent modeling. I was taught in chess (by a master friend of mine) that if you are dead lost, instead of resigning _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
