THE NUMBERS & MY THINKING

How the model works

It rates every J1 and J2 club from official results since 2003, then turns those ratings into match probabilities and expected goals. For me, thats a useful starting point. It gives us a consistent view across the league.

Its been tested on 4,472 J1 and J2 matches from 2021 onwards, using earlier results to forecast games it hadnt seen. The recorded probability score is 0.219, against 0.231 for a league-average guess. Lower is better. It beats that baseline in the supplied test. The Season so far table keeps backtests separate from forecasts actually published before kick-off.

The underlying ratings and season record use results alone. Match forecasts add team news and ACL congestion adjustments from round 14. Each match shows the adjustments and the results-only odds beside them. My thinking sits alongside the model, and I can disagree with it without changing its numbers.

Forecasts: J League Scout. Bookmaker odds: The Odds API. Results data from the official J.League sites, compiled by mokekuma-git/JLeague_Matches-Bar_Graph (CC BY 4.0) and k-addtimes/jleague_open_data (MIT).