Aria was built around a simple frustration: football is full of people explaining results after they happen. We wanted something different — a way of measuring the information available during the match, before the outcome gets the chance to rewrite the story.
I measure. I don't predict.
That distinction matters. Aria isn't trying to tell you what must happen next. She produces a probabilistic read from the match state, converts that read into fair prices and lets the market decide whether there is anything worth doing.
Possession, shot counts and scorelines can describe a match without explaining it. Aria is interested in the quality and context of the chances being created, and in what that information says about the next phase of the game.
Aria's number comes first. The exchange price comes afterwards. A selection is only interesting when the market offers enough compensation for the probability we think we're buying. Winning and losing are outcomes; value is the decision.
We track what Aria measures against what subsequently happens. Large samples matter more than memorable matches. The aim isn't to create a model that sounds clever — it's to find out where the read is reliable, where it isn't and whether it improves as the evidence grows.
Research, blind results and league calibration belong on the site because claims should be testable. Aria will evolve, and when the evidence changes the model should change with it.
This was the game that crystallised the idea. Liverpool took 34 shots and failed to score. Depending on the xG provider, their total was around the mid-twos — not above four — yet the reaction around the match captured exactly the problem that interested us: a large cumulative xG number could coexist with a game in which many observers felt Liverpool had created surprisingly little genuine danger.
The point wasn't that xG was useless. The opposite. xG was valuable enough to make the contradiction interesting. Thirty-four attempts can accumulate a substantial expected-goals total while being composed largely of individually modest chances. That raised the question that became Aria: can we get closer to the quality of what is actually happening, then use that measurement to price what comes next?
Aria grew from that question. Not as an argument against xG, but as an attempt to build on what xG taught us — and to measure a match in a way that is useful before we know how the story ends.
The betting application matters because a market gives the measurement a price and therefore a test. But the underlying job remains the same: measure, calibrate, challenge the assumptions and keep the outcome from fooling us.