TL;DR: My game AI kept freezing. It wasn't a bug in the AI, but a feature of the game allowing infinite move loops. I fixed it by adding a move limit rule, and now the AI runs faster and beats me even more effectively.

Fun little story about that: once in a while, the Alpha-Beta AI would just hang. So I needed to debug an AI-generated Alpha-Beta search that I before handn't bothered to understand.

Turns out there is nothing wrong with the AI.

Turns out that this variation of a game can have positions where a move simply loops.

Only after I asked ChatGPT to create a python script that does these sequences of "sowing steps" that make the heart of the game, it told me that after a sequence of 343 such steps, we're back at the initial position.

Then added a "you may only sow 100 hands of stones in a move rule" and now all is fine.

(this also explained why a 6-ply search would take forever - these kind of positions are fairly frequent).

The good news: now I have not only the 4-ply search that destroys me but the app can do up to 8-ply without taking too much time, and destroying me even more.

My dev learnings:
- asking ChatGPT to visualize a JSON is fast.
- asking ChatGPT to simulate a given issue in python is fun and gives insights fast.

#mancala #kahala #boardgames #development #gemini25pro