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

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.