Data sources
Fixtures, standings, squads and injury data are collected from professional sports data providers. Team-level advanced metrics — including expected goals (xG), expected goals against (xGA), possession and defensive actions — are synchronized daily from a second independent provider.
Market context (1X2, totals and handicap movement) is snapshotted at multiple points before kickoff to detect significant shifts.
Report generation
Each report is generated by a large language model that receives a fixed, structured data package and a strict output schema. The model's role is reasoning and expression — never data invention. Reports that fail structural or factual validation are rejected automatically.
Settlement rules
The public track record measures exactly two outputs: the 1X2 direction (home win / draw / away win) and the Over/Under 2.5 goals direction. A prediction is a HIT when the full-time result matches the published direction.
Market and handicap data may inform the analysis, but the published predictions cover ordinary 1X2 direction, scoreline, total goals and half-time/full-time only; handicap results are not part of the settled ledger.
Settlement runs automatically after full-time. Once settled, a report is frozen permanently — we never edit, delete or re-write history.
Statistical integrity
Hit rates are computed as hits divided by total settled matches, with no minimum-sample filtering and no league cherry-picking in the headline numbers. Losing streaks are displayed exactly as they happen.