OddsLab · public methodology

How OddsLab football predictions are produced

This page explains what goes into a selection, the checks applied before publication, and how results are counted. It never presents a bet as certain.

Editorial lead: Chafy Game · Last updated:

01

The data we observe

OddsLab combines fixture schedules, historical results, available contextual information, and pre-kick-off odds snapshots. Each source has its own update frequency; missing information is never made up.

02

From data to probability

The model estimates probabilities for several supported markets, then applies calibration and consistency checks. The result remains a statistical estimate; it cannot describe everything that may happen during a match.

03

The decision to publish

A public selection links a specific market to odds that were actually observed. OddsLab compares the model’s probability with the reference odds, checks the data for freshness and known anomalies, then separately records the odds observation time and the detection creation time. The content update date marks a change to the public view; it is neither the odds time nor the detection time. Publication can be disabled globally or separately for selections, daily accumulators and high-risk tickets. A high-risk ticket never appears unless it has been explicitly published; its editorial stage - Building, Validated or Finished - remains separate from its sporting outcome.

Probability
The model estimate after its corrections.
Observed odds
The price observed at a given time, which may subsequently change.
Model–odds gap
The estimated difference from the reference odds, never a promise of profit.
04

Settlement and success rate

After the match, the official score determines the selection’s status. Won and Lost describe its sporting outcome. Refunded means the stake was returned under the market line, Removed means the detection left active tracking, and Cancelled is reserved for a genuine match or market cancellation. An accumulator wins only if every required selection wins. A refunded selection carries settled odds of 1.00 in the ticket without erasing its observed odds. If all selections are refunded, the accumulator is also shown as Refunded with cumulative odds of 1.00. The displayed rate is calculated as follows:

won predictions ÷ (won predictions + lost predictions)

Pending, refunded, removed or cancelled events remain visible in the history, but only won and lost predictions enter this percentage. The number of settled cases is shown next to the rate so that a small sample is not mistaken for an established trend.

05

What the model cannot guarantee

A late team sheet, a card, an injury, or an incident during play can change the result. Odds can also move after detection. OddsLab does not accept stakes, is not a bookmaker, and guarantees neither results nor returns.