Well done to all the participants in the GECCO 2025 competition!
The clear winner both in the full and partial observability leagues are the entries by TeamTitans: well done to the team at the Frankfurt University of Applied Sciences!
Read more about their approach in their Slide Deck: Team Titans GECCO 2025 Entry
League results are as follows:
| Rank | Agent Name | Avg Win Rate % | Total Played |
|---|---|---|---|
| 1 | TeamTitansAgentV3 (Remote) | 99.0 | 1200 |
| 2 | Greedy Lookahead Agent (Remote) | 62.7 | 1200 |
| 3 | FlexibleEvoAgent-200-50-0.5 : OpponentWrapperFactory | 60.3 | 1200 |
| 4 | Greedy Heuristic Agent | 58.8 | 1200 |
| 5 | Better Greedy Agent (nHold=200) | 55.5 | 1200 |
| 6 | FlexibleEvoAgent-200-50-0.5 : FixedMappingWrapperFactory | 55.4 | 1200 |
| 7 | Greedy Heuristic Agent (Remote) | 52.2 | 1200 |
| 8 | SmarterAgent4 (Remote) | 48.2 | 1200 |
| 9 | Dynamic RHEA Phase-Aware Agent v2 (Remote) | 21.3 | 1200 |
| 10 | Careful Random Agent | 12.2 | 1200 |
| 11 | TorchAgent PlanetWarsAgentGNN models/PlanetWarsForwardModelGNN__ppo__random__adj_False__1__1751564155_final.pt (Remote) | 0.3 | 1200 |
(single league with fixed default game params - not many entries)
| Rank | Agent Name | Win Rate % | Played |
|---|---|---|---|
| 1 | TeamTitansPartialAgentV2 | 74.7 | 300 |
| 2 | GreedyPartialObservableAgent | 53.3 | 300 |
| 3 | Partial RHEA (Phase-Aware) Agent | 44.3 | 300 |
| 4 | Careful Partial Random Agent | 27.7 | 300 |
These show the game parameters for each league, and the results for each league - these were aggregated into the overall league above.
See results in this folder: GECCO 2025 Fully Observable League Results