Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Agent Representations for Ice Hockey
Guiliang Liu, Oliver Schulte, Pascal Poupart, Mike Rudd, Mehrsan Javan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add hierarchical ladder latent variables (vs single-layer CVRNN) (ablation) | VaRLAE · Player identification, standard | accuracy: 46.61% -> 50.01% | Improved |
| add hierarchical ladder latent variables (vs CVRNN) (ablation) | VaRLAE · Player identification, sparse participation | accuracy: 24.30% -> 36.65% | Improved |
| use variational instead of deterministic encoder (CAERNN to CVRNN) (ablation) | recurrent encoder · Player identification, standard | accuracy: 43.64% -> 46.61% | Improved |
| use variational instead of deterministic encoder (CAERNN to CVRNN) (ablation) | recurrent encoder · Player identification, sparse participation | accuracy: 11.65% -> 24.30% | Improved |
| add play history via recurrence (CVAE to CVRNN) (ablation) | conditional VAE · Player identification, standard | accuracy: 11.94% -> 46.61% | Improved |
| constrain candidates to recent on-puck players | MA-BE · Player identification | accuracy: 19.74% -> 51.75% | Improved |
| constrain candidates to recent on-puck players | VaRLAE · Player identification | accuracy: 50.01% -> 78.54% | Improved |
| replace variational encoder with deterministic encoder (CVAE to DE) (ablation) | CVAE · Player identification, standard | accuracy: 11.94% -> 9.40% | Hurt |
| add one-hot player ids to the XG model (ablation) | XG model · Expected goals estimation | F1-score: 0.21 -> 0.15 | Hurt |
| add VaRLAE player embeddings to XG model (vs no player info) (ablation) | XG model · Expected goals estimation | F1-score: 0.21 -> 0.54 | Improved |
| use VaRLAE instead of CVRNN embeddings (ablation) | XG model · Expected goals estimation | F1-score: 0.54 -> 0.54 | No clear effect |
| use VaRLAE instead of CVRNN embeddings (ablation) | XG model · Expected goals estimation | precision: 0.40 -> 0.37 | Hurt |
Closest papers by meaning
- Uncertainty-Aware Reinforcement Learning for Risk-Sensitive Player Evaluation in Sports Game (NeurIPS 2022 · similarity 0.86)
- Event2Tracking: Reconstructing Multi-Agent Soccer Trajectories Using Long-Term Multimodal Context (AAAI 2025 · similarity 0.86)
- Semi-Supervised Generative Models for Multiagent Trajectories (NeurIPS 2022 · similarity 0.86)
- Diverse Generation for Multi-Agent Sports Games (CVPR 2019 · similarity 0.86)
- SportR: A Benchmark for Multimodal Large Language Model Reasoning in Sports (ICLR 2026 · similarity 0.83)
- Sports-Traj: A Unified Trajectory Generation Model for Multi-Agent Movement in Sports (ICLR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
