Monte Carlo Pass Search: Using Trajectory Generation for 3D Counterfactual Pass Evaluation in Football

arXiv 2026 64.6 method, application

TLDR

Monte Carlo Pass Search uses a world model and value model for counterfactual pass evaluation in football with real Bundesliga data.

Reasoning

The paper introduces a novel MCTS-like framework for pass evaluation, leveraging a world model for trajectory generation and a value model for scoring, with strong empirical validation on a real-world dataset. Weaknesses include domain adaptation from autonomous driving and limited generalizability beyond football.

Read-first score

Read-first score 64.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 44.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=analysis,baseline,dataset,evaluation

Citation impact 18%

87.8

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.87751413

Topical relevance 29%

62.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=checkpoint,code,dataset

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 420.

Keyword Scores

world model

world simulator

world dynamics prediction

generative world model

model-based reinforcement learning world model

interactive world model

video world model

Deep Analysis

Innovations

Recasting pass evaluation in football as a Monte Carlo Tree Search (MCTS)-like evaluation problem using trajectory generation.
Monte Carlo Pass Search (MCPS) method that infers kick parameters, samples execution and option variants, and rolls forward with a ball-conditioned world model.
Distribution-aware attribution with two complementary execution-surplus scores (mean-based and percentile-based) for analysis and ranking.
Adapting a discrete-token, autoregressive trajectory generator (SMART) from autonomous driving to football for sample-efficient world modeling under limited public data.

Methodology

The method uses a high-fidelity 3D tracking dataset from the Bundesliga. For each observed pass, kick parameters are inferred, then execution variants and option variants are sampled. Each candidate is rolled forward using a ball-conditioned world model (adapted SMART) until the next ball interaction, and outcomes are scored with a learned value model to obtain a distribution over gained value. Execution-surplus scores are computed for distribution-aware attribution.

Key Results

The adapted SMART world model achieves strong best-of-20 forecasting accuracy compared to baselines. Model checkpoints and code have been released.

Limitations

Limited public data for training the world model (only one high-fidelity 3D tracking dataset from the Bundesliga).
Reliance on existing components (value model, world model, policy) that may have their own inherent limitations.
Evaluation is conducted on a single league/dataset, which may limit generalizability.

Resources

Paper