A wider football world.
Move between leagues and cups. Find fixtures, results, tables and team pages without losing your place.
FOOTBALL, WITH THE WORKING SHOWN
From the first result to the final probability. A football workspace that makes the data, the model views and the method easier to follow.
history → strengths → score grid
goals + julia → two model views
evidence → explanation + unknowns
01 THE ORIGINAL PREVIEW
Follow a match from the loading pipeline to its probability, score grid and Julia’s on-device reading.
The original Süper Lig walkthrough. The app now explores leagues and competitions beyond Türkiye.
This silent film opens with MatchRater’s loading steps. It then shows a list of fixtures, a match’s home/draw/away probabilities, fair odds and price history, a scoreline grid, Julia’s written reading, and the model’s method and historical evaluation screens.
02 INSIDE THE APP
Move between leagues and cups. Find fixtures, results, tables and team pages without losing your place.
Explore outcome probabilities, scorelines and model history. Read Julia’s estimate beside the goals model.
Ask an optional API analyst to explain the supplied evidence, show its sources and identify missing information.
03 DIFFERENT ROLES. ONE WORKSPACE.
Each part has a job. The estimates stay visible, and the explanation stays connected to the evidence.
Turns previous results and team strength into a distribution of possible scores.
A separately trained model reads the match history and offers another probability estimate.
Gemini or LongCat explains supplied facts and model agreement when you ask.
Explanations do not change the models’ probabilities. New weights are evaluated before they become an app update.
04 FROM CLUB FOOTBALL TO THE WORLD STAGE
The competition browser includes 41 competitions. Available fixtures, history and model coverage vary by competition.
05 A FEW USEFUL DETAILS
MatchRater is currently a personal research project. This page shares its design and a working app preview; a public download is not available.
The goals model and Julia run on the iPhone. Cached match data can be explored offline. Fresh fixtures and signed model updates need a connection, and optional API explanations send the selected evidence to the chosen provider.
As probability estimates with uncertainty. MatchRater makes the method, competing model views and evaluation visible. Historical test results do not establish future performance.
No. Julia is trained against real match results through a separate evaluation process. The API analyst’s role is to explain the supplied evidence and its limitations.
∵ MATCHRATER / A MESUTCY PROJECT