In-Play Over/Under Goals: A Remaining-Goals Method

Translate the line into remaining goals

The current score is only one input. Record match clock, added-time convention, player count, line-ups, event reviews, current line, qualifying period and the timestamp of every live input.

Worked remaining-goals calculation

Suppose an illustrative model estimates a Poisson mean of 1.20 additional goals from the decision time to the end of the qualifying period. The example match is 1-0, so Over 2.5 requires at least two more goals.

Using the OpenStax Poisson probability mass function:

P(0 remaining goals) = exp(-1.20) = 0.301194

P(1 remaining goal) = 0.301194 * 1.20 = 0.361433

P(2 or more remaining goals) = 1 - 0.301194 - 0.361433 = 0.337373, or 33.74%

Illustrative fair decimal odds before margin = 1 / 0.337373 = 2.96

This illustrative result is not a forecast for a real fixture. It is valid only under the invented mean and Poisson assumptions, and it is highly sensitive to the remaining-goals input.

Estimate the remaining mean without leakage

The mean can be produced by a prevalidated model using pre-match team strength plus live score, time, player count and other fields available at the cutoff. The peer-reviewed Bayesian in-play football model is one example of conditional outcome updating and posterior checking. It does not supply a universal remaining-goals rate.

Do not use a full-time statistic, post-match correction, final line-up event, or later price when reconstructing an earlier decision. Train and validate chronologically, report calibration by clock and score state, and compare with a simple baseline.

Settlement and execution checks

Betfair's football rules and DraftKings' soccer rules illustrate that period, line, abandonment, awarded-result and live-settlement treatment can differ. Preserve the rule attached to the accepted product.

For whole, half and split lines, enumerate each settlement state rather than labelling everything over/under. Record displayed and accepted price separately, and mark market suspensions around material events.

Validation checklist

1. The current score and exact total line are saved.
2. Remaining time includes a declared added-time treatment.
3. Live inputs existed at the decision timestamp.
4. The model estimates a full remaining-goals distribution.
5. Calibration is checked on later matches by time and score state.
6. The accepted price and settlement rule are preserved.
7. No action is recorded when a material event or feed state is unresolved.

The Gambling Commission's in-play standard warns that live information can be delayed, so price and event timestamps cannot be assumed to match.

Next step

Use Over Under Goals Betting for the next part of this topic.

Continue learning

• Next guide: Late-Goal Statistics
• Related guide: Red-Card Impact in Football

Assumptions and limitations

The example assumes one Poisson remaining-goals count, a stable mean after the cutoff and no dependence correction. Real scoring intensity can vary with time, score, teams, substitutions and player count. A well-calibrated probability still does not establish a favourable executable price.