Simulation
Simulation
Generative rollouts of tactical scenarios are a research direction; today the same machinery powers an evaluation workbench for the foundation model.
The research direction
Generative rollouts of full tactical scenarios. The generative simulator models the joint distribution of all on-pitch agents, enabling counterfactual analysis and what-if reasoning at the play level.
That is the vision statement, quoted as written — and it is labeled Research deliberately. No such simulator exists at WithoutBall today. What exists is the machinery that would have to underpin one: a generative model of body movement that can already continue, complete, and re-imagine spans of motion it cannot see.
What runs today: an evaluation workbench
Generative rollouts and rigorous evaluation are the same computation viewed from two directions. When the model fills a gap in observed movement, comparing the completion to the ground truth is an evaluation; sampling it forward without ground truth is a simulation. Today that machinery runs exclusively in the first direction — as the workbench behind the evaluation suite, where reconstruction, probing, and clustering results are produced and measured.
- Gap completion scored against ground truth
- Reconstruction error measured per occlusion pattern
- Probe and clustering harness for the embedding
- Free-running rollouts beyond the observed window
- Multi-agent tactical scenes
- Counterfactual what-if queries at play level
What would have to be true
Three things separate the workbench from a simulator: substantially more data than today’s corpus, validated multi-agent dynamics rather than single-body reconstruction, and evaluation protocols for generated scenarios that do not yet exist anywhere in the field. Until those are earned, simulation stays labeled what it is — research.
The data side of that equation is mapped on the data-scale ladder.