Run as a competition in conjunction with - GECCO 2025
We strongly recommend submitting your entry well before the deadline; non-functioning entries will not be accepted.
For the GECCO 2025 competition, we will run both full and partial observability modes. You may enter either or both variants.
See the main Submission Instructions for details on how to submit an entry.
This page only covers the specifics for GECCO 2025.
Game parameters will be as per the provided defaults, except the following which will be uniformly randomly sampled from the following ranges:
Game Duration will be fixed at a maximum of 2000 ticks. Games will run at 20hz, giving up 50ms per decision for agent to respond to the game state, but it runs in real-time, so the agents can take longer if they wish, albeit making decisions with stale information (actually if both agents respond more rapidly game will run faster).
We will also run leagues using the exact same parameters used for the sample leagues.
The submitted bots will be evaluated against a set of provided agents as well as each other.
For this competition there are no prizes, but we will rank the agents based on their average win rates, the winner being the agent with the highest average win rate across all games played.
We aim to run enough games to ensure a statistically significant result, but this will depend on the number of entries and the true underlying difference between the best agents.
Results will be published in this repo after the competition workshop at GECCO 2025.
For this competition, we will not require entries to be open sourced. However, we encourage teams to share their code if they are willing to do so via a public GitHub repository.