Paper Overview
Research Area: Computer Vision Authors: Akihiro Kubota, Tomoya Hasegawa, Ryo Kawahara, Ko Nishino Published: 2026-03-26 arXiv: 2603.25736
Original Abstract
Gauging an individual's skill level is crucial, as it inherently shapes their behavior. Quantifying skill, however, is challenging because it is latent to the observed actions. To explore skill understanding in human behavior, we focus on dyadic sports -- specifically table tennis -- where skill manifests not just in complex movements, but in the subtle nuances of execution conditioned on game context. Our key idea is to learn a generative model of each player's tactical racket strokes and jointly embed them in a common latent space that encodes individual characteristics, including those pertaining to skill levels. By training these player models on a large-scale dataset of 3D-reconstructed professional matches and conditioning them on comprehensive game context -- including player positioning and opponent behavior -- these models capture individual tactical characteristics within their latent space. We probe this learned player space and find it reflects distinct playing styles and attributes that collectively represent skill levels. By training a simple relative ranking network on these embeddings, we show that both relative and absolute skill prediction can be achieved. These results demonstrate that the learned player space effectively quantifies skill levels, laying the foundation for automated skill assessment in complex interactive behaviors.
Key Contributions
- A generative modeling approach for player-specific tactical racket strokes in table tennis
- A joint latent embedding space encoding individual characteristics, including skill levels
- Conditioning on comprehensive game context (player positioning, opponent behavior)
- Validation on a large-scale dataset of 3D-reconstructed professional matches
- Relative and absolute skill prediction via a simple relative ranking network trained on the embeddings