I’ve since made some corrections and clarifications so if you have downloaded the rules and or game software on December 31st, please re-download them once again to obtain the current version.

There are deeper strategies to be found in Splay than what the computer game exhibits. For example, developing a cluster of nearby towers in a vacant area is one such strategy.

This game will pose some serious programming problems precisely because humans can work out long term strategies that are not easily implemented in a traditional evaluation function. Monte Carlo evaluation may initially have a better shot, but MC suffers if long but narrow tactical lines play a role, since these seldom show up in random output. So for the foreseeable future don’t make the mistake of underestimating a program’s stupidity ;-)

I’ve noticed that I can get an early lead using a deeper strategy than the computer is able to find. However, the computer eventually catches up and surpasses me at the end of the game as the number of possibilities is reduced and the search depth increases. That shouldn’t come as a big surprise. Axiom’s MC engine is a work in progress, eventually I plan on creating an MC version of Splay.