baller.fyi publishes football forecasts generated by DARA. For every match you’ll see the probabilities for Home, Draw and Away, along with the number of goals DARA expects each team to score.
The approach is simple: publish the numbers before the match, leave them unchanged, and track what actually happens.
These show how DARA sees the possible outcomes. The outcome with the highest probability is the one used for Accuracy.
Probabilities express uncertainty, not certainty. A 60% chance doesn’t mean something will happen — it simply means DARA considers it more likely than the alternatives.
Expected Goals indicate how many goals DARA expects each team to score on average. They give useful context behind the probabilities — whether the model expects a tight, low-scoring game or something more open.
They are not exact-score predictions and do not form part of the Accuracy measure. They simply help you understand the kind of match DARA is forecasting.
We don’t choose matches because they look easier or because the model is more confident.
Every published match that finishes with a valid full-time result contributes to Accuracy. A difficult fixture counts exactly the same as a strong favourite, and unsuccessful forecasts are never removed from the record.
There is no selective filtering behind the Accuracy figure.
Forecasts are generated before matches using the information available to DARA at the time.
Once a match is published, its probabilities and Expected Goals stay fixed. They are never adjusted after the result is known.
If a match is moved to a different day, its original forecast is retired and a new one can be published for the new date. A change of kick-off time on the same day leaves the published forecast unchanged.
Accuracy measures how often the outcome with the highest published probability matches the full-time result.
Every published match with a valid full-time result adds to the same running total. Accuracy continues to build over time rather than resetting when matches leave the Archive.
Only full-time results are counted. Extra time and penalties are ignored.
The Archive displays the original published numbers next to the finished result, so you can check recent forecasts exactly as they appeared before the match.
It retains the latest five matchdays. Older matches eventually leave the public Archive, but their results remain part of the Accuracy record.
Football schedules sometimes change after a forecast has already been published.
If a match is postponed to another day, its original forecast is retired. A new forecast can then be published for the new date.
If a published match is abandoned, its forecast is retired and does not count towards Accuracy.
Neither situation is treated as a finished match.
baller.fyi offers DARA’s mathematical view of a match. It does not provide betting tips or guarantees.
The probabilities show how DARA assesses the possible outcomes. Expected Goals add scoring context. The Archive preserves what was originally published, and Accuracy shows how those forecasts have performed over time.
The information is yours to interpret.