expected-threat

Professional football scouting platform — xT maps, pizza charts, tactical clustering & AI scout annotations for Top 5 European leagues.

Deriving Expected threat as first introduced by Karun Singh in his blog post. Using this metric to further visualize the difference between the top 5 teams and players in the La-Liga 2017-2018 season.

Expected Threat (xT) model for football, built on a sequence-modeling analogy - match events as tokens, possessions as sentences - using a Transformer + Dual Mixture Density Network.

Data-driven football possession value model (xT/VAEP) to quantify ball progression and player impact.

Production-grade football intelligence platform — Expected Threat (xT), trained xG model, VAEP-style possession value, pitch control from StatsBomb 360 freeze frames, packing, data-driven player role discovery, AI tactical commentary. Real World Cup 2022 / Premier League data, FastAPI + React.

Elite Football Data Visualization & Computer Vision Tracking Suite | Gornik Zabrze (Ekstraklasa 2025-2026) & Turkish Super Lig Applications. Benchmarked against roboflow/sports, eddwebster, golazo, and worldfootballR.

Temporal expected threat modeling with semi-Markov processes, survival analysis and StatsBomb 360 data.

Gradient-boosted pass-danger model on StatsBomb event data — flags passes likely to create a shot and ranks passers across five leagues, with pitch-control and circular-statistics run analysis. Uppsala MSc (1RT001).

Spatial decision-making analytics on 2022 World Cup tracking data — pitch control, trained xT, per-pass decision quality (Streamlit)

H-VAEP and H-xT: Valuing Offensive On-the-Ball Actions in Handball by Estimating Probabilities

Interactive Streamlit app evaluating 2022 World Cup players from raw event data, with Expected Threat (xT) and Bayesian player ratings built from scratch.

Football scouting and player evaluation using statistical analysis and visualization in R.

Self-contained football analytics dashboard for all 104 matches of the 2026 FIFA World Cup — custom xG, expected threat fitted to the tournament, tactical profiles and head-to-head comparison. Vanilla JS, no dependencies, one offline HTML file.