Game Solver
Report
This “Game Solver” was the first time I ever made a tech demo showcasing a system of my choice, not assigned to me by a professor or to serve a greater role within a larger project. It uses logic I studied in a course covering combinatorial game theory, a class dedicated to proofs that can be obtained from 1v1 games with no unknown values.
As this project was not for a game, it has a significantly lower quality of play in comparison to my other works. This is apparent all the way from the poor user feedback to the AI that is next to impossible to beat.
AI is precisely the purpose of the project, where I created a system that looks at a defined game (in the case of the project, Hackenbush) and determines the optimal way to play after considering every possible action and every possible outcome.
Naturally, such an algorithm would be notably slow (n! in the example game), thus it can’t be optimally used in highly complex scenarios. However, even in simpler games it is fast enough to outperform an experienced human player, and has very few weaknesses. Though such an AI wouldn’t fit well in most modern games, the logic can still be applied. Say for instance instead of using every action and every outcome, work with predefined actions and expected outcomes. Yes this results in a less perfect play, but the AI will still make highly educated strategies.
All in all, I can’t push that such a complex algorithm be used in the everyday game, but I do find pride in my ability to program it on my own in a timely manner.
h