baller2vec++: A Look-Ahead Multi-Entity Transformer For Modeling Coordinated Agents
This addresses the challenge of modeling coordinated agents in domains like sports analytics, where existing methods often assume independence, but it is incremental as it builds on a predecessor model.
The paper tackles the problem of modeling statistically dependent trajectories in multi-agent spatiotemporal systems, such as basketball, by introducing baller2vec++, a multi-entity Transformer that uses a self-attention mask on location and look-ahead sequences. The result shows that baller2vec++ outperforms its predecessor by a wide margin in professional basketball data and can emulate perfectly coordinated agents in a simulated toy dataset.
In many multi-agent spatiotemporal systems, agents operate under the influence of shared, unobserved variables (e.g., the play a team is executing in a game of basketball). As a result, the trajectories of the agents are often statistically dependent at any given time step; however, almost universally, multi-agent models implicitly assume the agents' trajectories are statistically independent at each time step. In this paper, we introduce baller2vec++, a multi-entity Transformer that can effectively model coordinated agents. Specifically, baller2vec++ applies a specially designed self-attention mask to a mixture of location and "look-ahead" trajectory sequences to learn the distributions of statistically dependent agent trajectories. We show that, unlike baller2vec (baller2vec++'s predecessor), baller2vec++ can learn to emulate the behavior of perfectly coordinated agents in a simulated toy dataset. Additionally, when modeling the trajectories of professional basketball players, baller2vec++ outperforms baller2vec by a wide margin.