Neural Sabermetrics with World Model: Play-by-play Predictive Modeling with Large Language Model
Young Jin Ahn, Yiyang Du, Zheyuan Zhang, Haisen Kang · 2026
Read Neural Sabermetrics with World Model: Play-by-play Predictive Modeling with Large Language Model by Young Jin Ahn, Yiyang Du, Zheyuan Zhang, Haisen Kang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-07030.txt.
Description
Young Jin Ahn and three co-authors treat baseball as a world a language model can simulate, pitch by pitch. Classic sabermetrics boils long histories into compact stats. Those stats serve valuation and hindsight, but they do not generate how a game unfolds. The team's world model casts each game as a long autoregressive event sequence. One model is pretrained without pause on over a decade of Major League Baseball tracking data, covering more than seven million pitch sequences and near three billion tokens.
In one frame it predicts many facets of game evolution. On regular-season data and on postseason games outside the training mix, it beat strong neural rivals. It got about 64 percent of next pitches right within a plate appearance, and 78 percent of batter swing choices. Language models, the work suggests, can serve as unified world models for sports.
Catalog record
| CREATOR | Young Jin Ahn, Yiyang Du, Zheyuan Zhang, Haisen Kang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 18 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-07030