Otter: A Time-Aware, History-Conditioned Human Chess AI

Otter：一款具备时间感知与历史条件的人类国际象棋 AI

Abstract: Otter is a 15.3M-parameter human chess AI that predicts human move selection by modeling play as a time-aware, sequential process rather than treating each position in isolation. It combines two conditioning signals: (1) a move history encoder that conditions predictions on the last 20 moves, capturing opening preferences, positional drift, and intra-game behavioral tendencies; and (2) a time control module that modulates predictions based on clock pressure.

摘要： Otter 是一款拥有 1530 万参数的人类国际象棋 AI。它通过将对局建模为一个具备时间感知的序列过程，而非孤立地看待每一个棋局位置，从而预测人类的走法选择。该模型结合了两种条件信号：(1) 一个走法历史编码器，根据过去 20 步的走法来调整预测，从而捕捉开局偏好、局势演变以及对局内的行为倾向；(2) 一个时间控制模块，根据剩余时间压力来调节预测结果。

Otter is trained on 6.1 billion positions from 117 million Lichess rapid games over 30 days on a single T4 GPU. Otter achieves 55.23% top-1 and 90.95% top-5 move-prediction accuracy, surpassing the prior state-of-the-art human chess model, Maia 2, with far fewer parameters and less training data. Across 11 Elo brackets (<1100 to >=2000), accuracy peaks at 57.38% in the 1900-1999 bracket.

Otter 在单块 T4 GPU 上训练了 30 天，使用了来自 1.17 亿场 Lichess 快棋对局中的 61 亿个棋局位置。Otter 的 Top-1 走法预测准确率达到 55.23%，Top-5 准确率达到 90.95%，在参数量更少、训练数据更少的情况下，超越了此前最先进的人类国际象棋模型 Maia 2。在 11 个 Elo 等级区间（从 <1100 到 >=2000）中，其准确率在 1900-1999 区间达到峰值，为 57.38%。

These results show that modeling chess as a time-aware, sequential activity yields more human-accurate move prediction than position-only approaches, using a smaller model. Code, trained models, and complete training logs are publicly released.

这些结果表明，与仅基于位置的方法相比，将国际象棋建模为一种具备时间感知的序列活动，能够在模型更小的情况下，实现更符合人类习惯的走法预测。目前，代码、训练好的模型以及完整的训练日志已公开。