matchprior

Calibrated sports probability · API

When we say 70%,
it happens.

Calibrated win probabilities for 11 sports from one JSON schema, each with the rationale behind the number — Elo rating, recent form, head-to-head, expected goals. Group every backtested call by the confidence we assigned, and the observed rate matches it.

Get a key See the proof →

Don't trust the chart — curl /v1/accuracy and it returns this table live.
67,667 out-of-sample backtested predictions · simulated, not live · a measure of calibration, not a betting or profit claim.

Reliability diagram n = 67,667
050100 050100 predicted % observed % perfect calibration 75.3% 86.3%

predicted vs. observed win rate, six confidence bands · dot size ∝ sample · out-of-sample backtest · /v1/backtest

said 60–70%65.1%won, out-of-sample backtest
said 70–80%75.3%won, out-of-sample backtest
said 80%+86.3%won, out-of-sample backtest
Calibrated, not just confident

A 60% pick should win about six times in ten.

Calibration is the property that makes a probability honest. It is not about being right more often — it is about a number meaning what it says. It is the same standard used to grade weather forecasts and election models: when a well-calibrated source says 70%, the event happens about 70% of the time over many calls.

Calibrated is not the same as accurate. A forecaster who always says 50% can be perfectly calibrated and tell you nothing. Calibration earns trust in the number; being confident and correct is a separate axis — we report both.
Why it matters for your build. A calibrated probability is a number you can reason with — combine it, threshold it, feed it to your own model — and the maths stays honest. An un-calibrated 70% is a vibe with a decimal point.
What it is not. Calibration tells you how often something happens — never that acting on it comes out ahead. These are simulated backtest figures, not a betting edge or profit. For information and entertainment only.
Why builders use it

Breadth, a visible rationale, and a record you can audit.

11

Eleven sports, one schema

Football, tennis, basketball, baseball, hockey, American football, AFL, rugby league, rugby union, netball and cricket — every fixture and prediction in the same shape. Learn it once, consume any sport.

{ }

The rationale object

Most prediction APIs return a bare number. We return the inputs behind it — Elo rating, recent form, head-to-head, expected goals — so you can show your users the why, not just the what.

✓

Published, verifiable calibration

/v1/accuracy and /v1/backtest serve the hit rate, calibration error and reliability-by-tier on this page as a first-class endpoint — so you can check every number yourself.

One header, real JSON

A calibrated prediction, with its evidence.

Authenticate with a single X-API-Key header and read the model's probabilities, the pick and confidence, and the full rationale object — the same signals the model weighed, returned inline.

Read the quickstart Browse the docs
request.shcURL
# Calibrated predictions for the Premier League
curl https://api.matchprior.com/v1/predictions?league=E0&limit=1 \
  -H "X-API-Key: mp_live_8f3c…"
200 · application/jsonResponse
{
  "fixture": { "home": "Arsenal", "away": "Chelsea" },
  "probs": { "H": 0.55, "D": 0.25, "A": 0.20 },
  "pick": "H", "confidence": 0.55,
  "rationale": {
    "ratingEdge": 111, "homeForm": ["W","W","D"],
    "expectedGoals": { "home": 1.7, "away": 1.1 }
  }
}
Get started

Start on the free tier — no card.

Read calibrated probabilities, fixtures, Elo ratings, the accuracy record and the backtest from one schema. Upgrade only when your volume or depth needs it.

Get a key See our calibration

Calibrated probabilities, not predictions of profit. For information and entertainment only.