- Updated Wednesday, August 24th 2016 @ 06:30:34
In my opinion Its interesting to discuss what all we done right after lockdown.
- Created Wednesday, January 27th 2016 @ 15:35:50
@HowDoesItWork: Well, I agree that you should keep things for your self of course. :)
- Created Wednesday, January 27th 2016 @ 19:53:28
I will bet on alpha-beta and a lot of optimization...
- Created Saturday, January 30th 2016 @ 00:59:01
I haven't looked into the competition yet however I don't really like it when strategy becomes a synonym for "brute forcing" or "smart way to brute force". (Bitboard representation to brute force efficiently,...)
The winner of this competition will be (of course) determined by the quality of the evaluation function and the ability to look deep into further configurations.
- Updated Saturday, January 30th 2016 @ 19:17:57
Well, Monte Carlo simulation plays at master-level in Go, and I think there are ways to make it work for games with lower branching factors as well. With such a search, you don't need any evaluation function at all. I have thus far been using negamax, but even if I were to implement alpha-beta pruning I suspect I won't be able to get enough plies deep; also, finding good heuristics for the evaluation function is dependent upon good domain knowledge, which can in turn be difficult to formulate.
I think I will try to implement a bot which uses the Monte Carlo method and see how it fares, after all it relies mostly on many methods I have already implemented in my current bot.
EDIT: I rewrote it using Monte Carlo, but there doesn't seem to be remotely enough time to compute enough playouts without heavy optimization. If anyone finds a smart way of implementing which performs well within the time limits, I'd love to know.
- Created Friday, January 20th 2017 @ 22:04:28
Isn't it time yet to discuss? I'm down for some in-depth discussion although I didn't manage to implement everything I wanted and didn't place well. I believe the bots are now frozen, amirite?