Second-Half Goal Statistics: A Period-Aware Method

Use two complementary measures

The share can be useful descriptively, but it ignores different exposure. Under the IFAB duration law, allowance for time lost varies, and second halves can have different elapsed duration from first halves.

Worked descriptive example

Suppose an illustrative sample contains 240 first-half goals in 45,000 actual first-half minutes and 300 second-half goals in 48,000 actual second-half minutes.

First-half share = 240 / 540 = 0.4444, or 44.44%

Second-half share = 300 / 540 = 0.5556, or 55.56%

First-half exposure rate = 240 / 45,000 = 0.00533 goals per minute

Second-half exposure rate = 300 / 48,000 = 0.00625 goals per minute

This illustrative example demonstrates why a goal share and an exposure rate answer different questions; neither is a fixture forecast.

Construct the data by period

The Wyscout open-data paper documents a public spatio-temporal event dataset. A period analysis should preserve match, period, event time, actual period end, team, score before event, player count, competition and correction version.

Remove disallowed events consistently and declare treatment of extra time. Do not classify a goal from extra time as a second-half goal merely because its minute exceeds the regulation clock.

Report the denominator beside every split

A league comparison should publish matches, goals and actual minutes for both periods, plus uncertainty around each rate. Check whether conclusions change when using nominal minutes, observed elapsed time, or matches with complete clocks only. Large differences between those versions indicate that clock coverage or added-time treatment is influencing the result.

Condition on half-time state

The second half begins with information unavailable before kick-off: current score, player count, substitutions used, injuries and observed performance. A 0-0 match and a 3-0 match at half-time are different conditional states. The Bayesian in-play football model provides one example of updating outcome probabilities from live state rather than applying a universal period share.

Test a forecast rather than a slogan

1. Define whether the target is second-half count, any goal, or a stated total line.
2. Freeze the half-time cutoff and available features.
3. Fit on earlier matches and retain later matches for evaluation.
4. Compare against a pre-match plus half-time-score baseline.
5. Report calibration and proper scores by competition and score state.
6. Repeat with actual elapsed-time and nominal-period denominators.
7. Keep every eligible match, including missing data and unavailable prices.

Next step

Use Half Time Betting Strategy for the next part of this topic.

Continue learning

• Next guide: Weather in Live Football Analysis
• Related guide: In-Play Betting Explained

Assumptions and limitations

The sample is illustrative and omits match-level dependence in its simple arithmetic. Competition formats, added time, team strength and tactical responses differ. A period-level historical pattern does not establish that a specific live total is mispriced.