Footballing Trends II
A second story on trends, following on from this, which comes out of what I wrote last night, relates to seasonal patterns in goals per game.
Should we expect particular points in a season to have more goals? What trends might we observe in total goals per football match?
Well, Soccerbase has over 600,000 football matches in its historical record of results. It’s England back to 1872, and a much broader range of countries in recent years. I’ve collected these and taken a look at total goals.
It’s a really simple linear regression model. That is, regress total goals in a match on a bunch of really simple explanatory variables. By simple I mean: a dummy variable that is 1 for:
- Each competition (e.g. Premier League, Estonian cup)
- Each calendar year.
- Each month of the year.
This results in a lot of coefficients. So data plots are useful. Two are below, for the fixed effects (dummies) for calendar years on the left, and for months of the year on the right. Lines are for standard error bounds, and circles are the coefficients themselves. Hence if both lines are above or below zero, that coefficient is statistically significant.
The very early years, and the periods up to the end of the 1960s are sometimes significant and there is considerable variation. But since around 1970 things have settled down with not much variation, and no statistical significant.
For the month of the year, though, relative to January, we see that September, October and November have significantly more goals (conditional on the year and the competition). December is borderline significantly different, while February, March and April are significantly negative — so fewer goals. So don’t bet on the current crazy goalscoring to carry on too much longer.
For completeness, what about the deviations by competition? They are plotted below (small font for the horizontal axis, solid circles significant, dotted lines added to significant one for reference to axis labels):
The English Football League and Premier League, generally, deliver fewer goals than the average match (a rag tag collection of matches across a range of divisions and countries over the years), Bundesliga and Eredivisie more, Serie A and Ligue 1 fewer, non-league competitions in England and Scotland more, smaller leagues across Europe fewer.
The baseline average goals per game? 2.6.