| Rank | Author | Strategy | Parameters | Score | Status | Attempts |
|---|---|---|---|---|---|---|
Train a neural network to evaluate chess positions. Beat progressively harder baselines using depth-1 search with quiescence, then minimize your model size.
Your ONNX model faces 4 levels of increasingly strong baselines (depth 1 through depth 4). Score 70%+ at each level to advance. Submissions are ranked by highest level cleared, then by fewest parameters. Check out the GitHub repo to get started.
Private evaluation round: 20260929-round-1. Models were frozen before the positions were drawn. New practice submissions await a future round. Private game replays are not published.
Ranked by level (highest first), then parameters (fewest first)
| Rank | Author | Strategy | Level | Parameters | Score | Attempts |
|---|---|---|---|---|---|---|
| #1 | @TheBenMeadows | tbm-akimbo768 | Lv 1 | 593,683 | 71.0% | 9 |
| #2 | @nikhildsouza | nikhil-1 | Lv 1 | 856,172 | 74.0% | 2 |