Hi.
> I've missed the start of the course:
> https://www.coursera.org/course/ggp
> I may enlist Toss as a player in the course, but I probably
> won't do my best toward Toss winning...
I'm sure it is not too late to make up for stuff in the course :).
As for Toss, there is one thing I'm recently interested in. I added
a way to generate QBF formulas saying "player 1 can win in N steps"
for "easy" games (2 locations only for now, and ?P not allowed in rules).
This works quite ok e.g. for Connect-4. And in SatSolver ml we now
have an interface to depqbf, a very good qbf solver. So e.g. for Connect4
it can check quite efficiently whether player 1 wins in a few steps (solving
the whole game, i.e. 7*6 steps, is more difficult of course). But it seems
that knowing "I can win in 4 steps" should be good for a player - esp. if
we could combine it with a heuristic in case of not-yet-winning. And this
should be quite easy in Toss now. So if you ever get to making a player
again, you can consider adding such a check :).
My best!
Lukasz
|
