How it works

How we price a football match

No tipster's hunch, no insider whisper. Every number we publish comes from the same pipeline: twenty-six seasons of evidence, an expected-goals model, and a decision layer that stays quiet unless the price is genuinely wrong.

26seasons of history
742K+fixtures, growing daily
162K+with real xG
7markets priced
The pipeline

Eight steps, every match, every day

The same path runs for a Champions League tie and a second-division fixture. Nothing is hand-picked.

1

Collect the evidence

Every match is stored with its full odds board and match statistics — 742K+ fixtures and growing every day, 97.8% of them carrying the bookmaker's pre-kick-off prices, and 162K+ with real expected-goals data. Where xG was never recorded, a model fills the gap rather than leaving a hole.

2

Compare against history

The fixture is measured against comparable matches from our history — several independent reference views rather than one. If they disagree, that disagreement is itself information, and it makes us more cautious, not less.

3

Model the goals

In parallel, a goal model estimates how many goals each side should score, weighting the quality of chances created over lucky finishing. From those estimates, every possible scoreline gets a probability.

4

Blend, then respect the market

Those views are combined and then deliberately pulled back toward the market's own price, once its margin is removed. The market is the best-informed forecast available; we move away from it only as far as the evidence justifies.

5

Strip the margin honestly

Every quoted price carries the bookmaker's built-in margin. Removing it carelessly flatters outsiders and invents edges that were never there — we learned that the hard way and withdrew the results. The method we use now is more conservative on longshots, deliberately.

6

Let a trained layer decide

A probability is not a recommendation. Each candidate is checked against the fair price and approved only when a decision layer, trained on hundreds of thousands of settled outcomes, judges the value real — after a deliberately cautious reading of its own confidence. The overwhelming majority is rejected. That silence is the product.

7

Size it to survive

Approved selections are sized by a disciplined bankroll rule and capped by a hard daily limit set from our own worst historical runs. The aim is not the biggest possible win; it is surviving the losing streaks that are certain to come.

8

Publish, then leave it published

Every selection is recorded before kick-off and settled automatically afterwards — winners and losers alike. Nothing is quietly deleted after a bad weekend. Open any past date and you see exactly what we said beforehand next to what actually happened.

“We don't try to predict who wins. We try to find prices that are wrong.”

A 46% outcome priced like a 39% one is value. A 70% favourite priced like an 80% one is a trap. Most of the time neither is true, and the honest answer is to say nothing.

What we do not claim

  • We are not a bookmaker. We do not accept, place or settle bets, and we never handle stakes.
  • These are probability estimates, not predictions of fact. A 70% call loses three times in ten — that is what 70% means.
  • No profit is promised or implied. Past performance, ours included, does not guarantee anything about tomorrow.
  • Losing months are expected and normal. Any system claiming otherwise is not being straight with you.
  • Nothing here is financial advice. 18+. If betting stops being fun, stop.

Questions? Write to us: bilgi@iddaakolik.com