AI algorithm for “RaceTrack” game

Question

does anyone know (or can suggest) a good algorithm for an AI for the RaceTrack pencil-paper game?

since you have 9 possible choices in each step and you need to look

Answer 1

You mention the idea of "assigning each choice some quality value" - this is called a heuristic function. Many AI algorithms (such as A* and alpha-beta pruning, mentioned by others) are only as good as the heuristic function you plug into them.

However, if you do manage to create a good heuristic function, then these algorithms will automatically perform much better "for free" - so it is very much worth your while to invest some time in developing a good one.

Another angle is to try to pre-compute your entire race from the start. Then it is a question of minimization of number-of-turns to cross the finish line. One handy minimum-finding algorithm is simulated annealing.

That aside, it would also be cool to see some genetic algorithm solution to a game like this. Not sure if it's quite the right fit, but I could imagine creating a 'brain' that takes various inputs - expected distance from wall at a few turns in the future, speed, distance to other racers, etc - and evolving the logic of that brain with a genetic algorithm. The trick is to break the problem down into parts that can be mutated meaningfully.

Actually, you could even combine them, and use a genetic algo. to develop a heuristic function which is plugged into a standard AI search algorithm.

Honestly though, brute force would probably work fine on a constrained track, since you can toss out a subtree if you crash (and crashes are common for bad paths).