Hacker News new | past | comments | ask | show | jobs | submit login

Combinatorial game theory would be the wrong tool for this, I believe, since this isn't a strictly finite turn-based combinatorial game. (And in my opinion combinatorial game theory is more useful for analysis and proof techniques than for creating AI algorithms, but I've only had one class on it.) The blog mentioned "fruitwalk-ab" uses minimax (with the alpha-beta optimization of course), which is the bread-and-butter algorithm for these kinds of games and I expected it to be in 1st place. Sure enough, it is at the moment. (Edit2: No longer.)

In an intro machine learning course you'd learn about minimax and others, but skip paying any money and just read the basic algorithms here and look up the wiki pages for even more examples: http://www.stanford.edu/~msirota/soco/blind.html (The introduction has some term definitions.)

Edit: Also, the obligatory plug for Artificial Intelligence: A Modern Approach http://www.amazon.com/Artificial-Intelligence-Modern-Approac...

Thanks for the info, interesting stuff. I've got more first-gen minimax going right now, sorta works but gets stuck in a repeating move pattern. Interesting stuff to work on though, i really appreciate the pointers. Hopefully this weekend i can work out a better heuristic valuation of a game state, for a first run i'm just counting my objects vs theirs(very simple), but i need to somehow weight it such that it prefers picking objects up sooner than later. And of course it would better match if it also took into account groupings, but thats for after i get basic board-clearing working.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact