How MatchIQ Works

MatchIQ is a personal project that generates football match predictions by combining real statistical and contextual data with an AI model (Claude, by Anthropic). Full AI predictions currently run for the 2026 FIFA World Cup, with live fixtures, results, standings and team form also available for Ligat Ha'Al, Liga Leumit, and the top two divisions of England, Germany, Italy, Spain, France and Portugal — with predictions expanding to more leagues as their seasons resume.

Before each upcoming match, the system gathers a wide picture of both teams and the match context. That picture is then passed to an AI model that weighs everything and writes a prediction. Here's what goes in:

• Team form & recent performance. The last five matches, results, and expected goals (xG) — a statistical measure of chance quality, not just the scoreline.
• Playing style. Possession, shots, corners, and fouls patterns that hint at how a team approaches a game.
• Head-to-head history. Past meetings between the two teams, including recent and long-term trends.
• League standings & group context. Where each team sits in the table or group stage, and what they still need from the fixture.
• FIFA world rankings. The official international ranking for national-team fixtures.
• Weather, altitude & travel. Conditions at the venue, plus rest days and travel distance for the visiting side.
• Betting market odds. Pre-match odds averaged across multiple bookmakers, used as an external reference point — not as the source of the prediction.
• Injuries & suspensions. Known absences, including automatically detected red-card suspensions from previous matchdays.
• A Poisson statistical model. An independent maths-only model runs alongside the AI as a cross-check on the projected scoreline.

All of that is sent to Claude, which weighs the factors and produces a final prediction with a written rationale. To reduce randomness, each prediction is actually generated three times and averaged for stability before it's published.

Every prediction is scored against the real result once the match is over, alongside the confidence rating it was published with. You can see exactly how often the model is right — and how often it's not — on the Predictions page.

MatchIQ was built as a personal project exploring AI-assisted football analysis — what happens when you give a capable language model a rich, structured view of a match and ask it to reason about the outcome. It's an experiment, shared openly, with its wins and losses on display.