Battlesnake AI Agent using a hybrid approach, combining dynamic heuristics with logit quantal response equilibrium. The agent is written in Python and was developed as part of a university class at LUH.
- Layered heuristics: Finds best field to aim for considering general favorability of fields, food, distance and enemy positions
- Path planning: Efficient path planning using modified A*, considering potential dangers by avoiding unfavorable fields
- Close combat mode: Uses quantal response equilibrium to approximate Nash equilibrium and make decisions considering future enemy decisions
- Can deal with limited information (i.e. in special game mode where only area with radius of 3 fields around head is visible), estimating the current game state (including own strength and enemy strenghts)
Placed first in internal competition for the university course. Peak position on Battlesnakes standard mode leaderboard: 13. Currently, the agent is no longer active.