Football forecasts, researched by an AI analyst
Every morning a large language model reads the data, checks the news and publishes forecasts for the day's biggest matches. Each one is graded in public.
Sign in to read today's forecastsUpcoming matches
3 forecasts are waiting for you
Sign in with Google — it's free — to read them. Once a match kicks off its forecast becomes public in the archive.
Sign in with GoogleLatest results
ArchiveGoals over/under 2.5
over 2.5 goals
PSG total goals
PSG over 1.5 goals
Total cards
over 4.5 cards
Home team goals
Barcelona over 1.5 goals
Brighton team goals
Brighton over 0.5 goals
Both teams to score
Both teams to score: yes
How it works
A tipster reads form, injuries, the referee and the odds, checks the news, and looks for where the bookmakers have it wrong. We do the same job with a large language model.
Gather
Each morning a script pulls the numbers for the day's biggest matches: form, expected goals, head-to-head, referee, absences and bookmaker odds.
Analyse
The model gets that data pack, searches the web for late team news and writes its forecast: its own probabilities next to the market's, and a pick only where it sees an edge.
Check and learn
The next morning every forecast is graded against the real result. When enough are in, a second pass reads where the model was over- or under-confident and rewrites the short list of rules it follows next time. We don't retrain the model itself — we retrain its playbook.
It is a language model, not an oracle, and it will be wrong. That is why every result stays public in the archive, and every forecast shows the data behind it.