A community championship showcasing open-source ingenuity, strategic code, and resource-efficient architectures. Run for fun and absolute bragging rights.
Autoregressive generators, sequence architectures, and causal networks predicting structural moves as natural language strings.
Deep policy architectures, Q-learning pipelines, and custom state evaluation models navigating deep decision spaces.
Host your model on Hugging Face and submit the repository link. Every repo must contain these four components.
A model card covering architecture overview, parameter count, training data & methodology, known failure modes, inference speed, license, and optional performance metrics.
Serialized model weights. Standard pickled files (.pth, .bin) are explicitly not allowed for security and rule enforcement.
Architecture configuration used to verify parameter counts and model settings against bracket constraints.
Your inference logic implementing the standard ChessAgent class template so the evaluation engine can drive games automatically.
Your chess_agent.py must implement exactly this class structure.
import chess
class ChessAgent:
def __init__(self, model_dir="./"):
"""
Initialize your model, tokenizer, or custom
neural network here. model_dir points to the
root of your Hugging Face repo download.
"""
pass
def select_move(
self,
board_history_san: str,
legal_moves_san: list[str]
) -> str:
"""
Inputs:
board_history_san — space-separated game so
far, e.g. "e4 e5 Nf3 Nc6"
legal_moves_san — list of legal SAN moves,
e.g. ["d4", "Bc4", "Nxe5"]
Output:
One string from legal_moves_san.
"""
chosen_move = legal_moves_san[0]
return chosen_moveGames run asynchronously on Kaggle / Colab via the python-chess framework, keeping hardware access open and reproducible.
If your agent returns an unparseable or illegal move, the engine substitutes a random legal move so play continues. Repeated reliance on the fallback will hurt strategic performance.
Swiss-system group stage leading into a single-elimination knockout. The LM bracket champion faces the RL bracket champion in the Grand Finale.
Full PGN match histories and bracket summaries will be published once the tournament simulation concludes.
Join the Discord to stay updated, share your submission, and discuss architectures with other participants.