Matter of Stats

View Original

A Freekick from the Umpires

Every year there’ll be a game or twelve where the fans of the losing team lament a lop-sided freekick count and largely attribute their team’s loss to the ineptitude (or worse) of the officiating. This will especially be the case where the visiting team has been on the thin end of the freekick count.

How much plausibility is there in this feeling?

Today we’ll look at historical freekick data and investigate the following:

  1. Do home teams tend to receive more freekicks?

  2. Does it vary depending on which is the home team?

  3. Does the relative strength of the teams have any impact?

  4. Do teams that win the freekick count also tend to win the game?

THE DATA

For this analysis we’ll use the freekick data available from the www.afltables.com website, accessed using the fitzRoy() package in R, along with the MoSHBODS pre-game team ratings.

We’ll include all home and away games from 1980 to 2021, excluding any where the afltables data shows a zero freekick count to either team. That leaves just over 7,000 games in the dataset.

THE METHODOLOGY AND MODEL

We’ll fit the following Ordinary Least Squares model:

FreeKick Differential ~ Intercept + IsHome + RatingsDifference + HomeTeam

where:

FreeKick Differential = Designated Home Team Freekicks For less Freekicks Against

IsHome = A binary variable that is “Yes” if the designated home team is playing notionally at home at a venue where it has played as the home team at least 25 times across the period (the reference category for this variable is “No”)

RatingsDifference = Difference between the pre-game MoSHBODS Combined Ratings of the home team and the away team

HomeTeam = Categorical variable reflecting the designated Home Team (the reference team for this variable is “Adelaide”)

Details of the fitted model appear at right, and can be interpreted as follows:

  • Adelaide, when playing as the notional home team at somewhere other than Adelaide Oval or Football Park against a team of equal ability would be, on average, expected to receive 1.36 more freekicks than their opponents

  • At Adelaide Oval or Football Park, the expected differential would increase to 1.75

  • For teams other than Adelaide, the expected differential moves up or down by the relevant coefficient for that team. So, for example, Western Bulldogs would be expected to enjoy an extra 1.56 freekicks when playing as the notional home team.

  • Just four teams tend to fare better than Adelaide - Western Bulldogs, West Coast, Collingwood, and Port Adelaide - and only the first two of those to a statistically significant extent

  • A lot more fare worse, especially Essendon, Gold Coast, GWS, Richmond, Brisbane Lions, Hawthorn, and Melbourne. On average these teams earn between 1 and 2 fewer freekicks per game

  • Interestingly, weaker teams do tend to receive more freekicks, but the effect size is quite small. The mean absolute difference in teams’ pre-game ratings is only 17 points, which is worth only 0.3 extra freekicks, according to this model

Overall, the model explains less than 3% of the variability in freekick differentials, which tells us that there are either a lot of other factors influencing the count or that it is largely random.

And, even where the model does provide some statistically significant explanations, the effect sizes are quite small - generally of about 1 to 3 freekicks.

THE IMPLICATION

Even if there is some evidence that freekick differentials are not entirely random, how important have these proven to be, on average, across history?

The chart below provides a boxplot of home team margins for every observed freekick differential.

The overwhelming message from this chart is that there is essentially no relationship between the home team’s freekick differential and final margin. The correlation is -0.0056.

So, even if some teams do tend to receive more freekicks than others, overall this does not translate into higher margins.

THE CONCLUSION

Based on an analysis of freekick data from 1980 to 2021 there is some evidence that, to a statistically significant level:

  • Home teams, generally, receive more freekicks

  • Particular teams playing at home receive slightly more (or slightly less) than the norm for home teams

  • Weaker teams receive more freekicks

In all cases, the size of these effects is quite small - of the order of just 1 to 3 freekicks per game - and, together, they explain only a very small proportion of the variation in freekick differentials.

When we look to estimate what the effect of these differences might be, we find virtually zero correlation between home teams’ margins and freekick differential, which suggests that, at a macro level, freekick differentials don’t matter.

There will always be cases where a specific freekick can be referenced and a convincing argument made about its pivotal nature in a particular game, but I think the more discerning and statistically savvy follower should recognise that, most of the time, the freekick differential is a poor predictor of the outcome of a game.

(They’d also be wise to think about why so many fans seem to assume that a zero freekick differential is somehow the “right” outcome. I’d suggest that’s no more necessarily the case than that the final margin should be zero as well, but that’s maybe another blog for another day).