A Dixon-Coles scoreline engine priced by a Poisson goal-rate model — one scoreline matrix prices the whole match book, then a minutes-mixture layer prices player props (shots, shots on target, goals, assists). Built for the 2026 World Cup across 18 competitions of training data. Pre-launch: the board you see is example data until the model is forward-validated and approved.
A Dixon-Coles bivariate-Poisson scoreline matrix prices the entire match book — result, totals, both-teams-to-score — from a single fitted goal-rate surface, with a GBM correction on each team's attack and defence λ.
Shots, shots on target, goals and assists are priced through a minutes-mixture layer on top of the scoreline engine — each prop conditioned on expected minutes and role, so a substitute and a starter never share a line.
The vertical is in build: data gathering and the Wave-1 model are complete and calibrated, with serving, a forward CLV gate, and the SportAdapter still to come. It goes public as its own violet tier only once it clears forward validation — fail-closed, same as every channel.
Computed live from the public settled archive — losses included. ROI is on staked units, halved-unit basis.
World Cup player research runs on the same per-market spine — shots, shots on target, goals, assists — built on the minutes-mixture props layer and coming online with the channel.
How the model worksThe finder and alt ladders extend to soccer markets the moment the book is approved — every rung priced, best-EV flagged, NO LINE where a book is missing.
About the finderFree through the NBA Playoffs — evening delivery, public record, cancel anytime.
