| Rank | Author | Strategy | Elo | Win% | Kill% | Attempts |
|---|---|---|---|---|---|---|
Program an AI agent that outmaneuvers opponents in aerial combat. Upload an ONNX model that observes the battlefield and decides actions in real time.
Your model receives a 224-dimensional observation and outputs 3 continuous actions each tick. It plays 100 matches against other submissions and is scored by Elo rating. Elimination wins count more than HP wins. Check out the GitHub repo to get started.
Results from the Apr 9, 2026 tournament
| Placement | Author | Strategy | Rounds Won | Win % | Seed |
|---|---|---|---|---|---|
| #1 | @adrianleb | ProtoSnoopy-X9 Sentinel Unit II | 3 | 94.0% | #1 |
| #2 | @0xschelling | follow me | 2 | 38.3% | #3 |
| #3 | @artemis_onchain | hehe | 2 | 76.0% | #4 |
| #4 | @bqbrady | Kibble v2 | 1 | 54.5% | #2 |
| #5 | @CryptoNdee | v7_fast | 1 | 67.5% | #6 |
| #6 | @cameronspirals | self_play_1 | 1 | 40.5% | #12 |
| #7 | @samaltman43 | best | 1 | 33.5% | #7 |
| #8 | @danrobinson | Slop Gun | 1 | 30.5% | #8 |
| #9 | @onblueroses | smol | 0 | 37.0% | #9 |
| #10 | @josusanmartin | p80_hg_a465 | 0 | 30.0% | #5 |
Strategies ranked by cross-play Elo. Ratings start provisional (vs reference pool) and refine via head-to-head Swiss rounds every 30 minutes.
| Rank | Author | Strategy | Elo | Attempts | Last Submitted |
|---|---|---|---|---|---|
| #1 | Baseline Strategy | Starter | 1000 | — | — |
| #2 | @adrianleb | ProtoSnoopy-X9 Sentinel Unit II | -1948 | 21 | 39d ago |
| #3 | @bqbrady | Kibble v2 | -2300 | 4 | 38d ago |
| #4 | @0xschelling | follow me | -2376 | 30 | 8d ago |
| #5 | @artemis_onchain | hehe | -2447 | 20 | 39d ago |
| #6 | @josusanmartin | p80_hg_a465 | -2524 | 49 | 30d ago |
| #7 | @CryptoNdee | v7_fast | -2561 | 2 | 36d ago |
| #8 | @danrobinson | Slop Gun | -2597 | 5 | 42d ago |
| #9 | @samaltman43 | best | -2603 | 3 | 30d ago |
| #10 | @onblueroses | smol | -2704 | 3 | 8h ago |