How it works: the stats model sets a base expected-goals number, then the AI reads your notes and nudges it up or down with a short rationale. Paste real info - confirmed lineups, injuries, suspensions, a team resting starters in a dead rubber, weather. Example: "Brazil rotating heavily, already qualified; Morocco at full strength and must win." Leave it blank to use the pure-stats model.
No predictions logged yet.
Elo = team strength. Roughly: ~1500 is weak (New Zealand), ~1700 average, ~2000+ elite (Argentina, Spain). These are my starting estimates, not an official feed - edit any value and it saves instantly. Sensitivity: about every 145 Elo points equals one goal of expected supremacy, so a +50 swing moves a team's win probability by roughly 5-8 points. Unlisted teams default to 1700.