Play against @AllieTheChessBot, yet another human-like chess bot

Chess botChess engineChess
Allie is a human-like chess bot that learns to move, think and resign like humans.

What is Allie?

@AllieTheChessBot is a human-like chess bot that learns to play chess like humans. One component of "human-likeness" is in the moves Allie play: we want Allie to play moves a human would play. Besides what moves to play, Allie also learns how much time she should use to make a move in every position, and when to resign!

How does Allie work?

Allie is a neural network (more specifically, a transformer model like ChatGPT but on a much smaller scale). We trained Allie on roughly 90 million blitz games played on lichess in 2022, and she learns the full trajectory of human chess (moves, think time and resignation), which can be modeled sequentially like human language. Allie also knows the strength of her opponents, and tries to adapt to different levels of strength.

We are planning to release data, code and model weights soon!

How is Allie different from Maia?

Allie is inspired by Maia. Maia and Allie are both chess bots that are trained to model human behavior. One key difference is that Allie is a single model that adapts to different levels of play, and Allie models human behavior beyond just the moves played.

In offline evaluations, we find that Allie consistently outperforms Maia in predicting the moves humans would play, and we are interested to see whether this offline performance translates to online games.

Help us improve Allie by playing!

Allie is part of a research project, and we would like to collect online games for understanding and improving Allie. It would mean a lot to us if you are interested in playing against Allie! If you play, please try to complete three games, since we change the configuration of Allie after every game. There will also be a survey sent via chat after each game.

You can play against @AllieTheChessBot using this link.

Note: Allie currently only accepts casual blitz games. For any questions or feedback, please DM @yimingz3, or shoot us an email at