## How repeatable is scoring/conceding goals from penalties?

Thanks, once more, are due to Infostrada (website), who’ve provided the data I’m using in this post. The work I’ve done previously with this penalty data can be found here.

Edit – 26AUG13

I made some errors in the previous plots (for transparency I’ve left the entire original post in it’s original condition below) by including all relegated teams as having zero Premiership penalties the following season (which is strictly speaking true, but they weren’t exactly given the opportunity to win any).

A few days back I posted two articles looking at the repeatability of shooting percentage from inside and outside the box are the beginning of a long series looking at which metrics are repeatable at the team level over the course of a Premiership season. At the end there’ll be a summary, which I’m aiming to have as a useful resource.

Today I’m looking at scoring and conceding goals from penalties. Note – this post only considers penalties that result in a goal. In the future I’ll do the same study but incorporating all penalties that are awarded. So if we take the erroneous points out then the data set is correct – it spans the 2001-02 – 2011-12 seasons. In total there are 220 team seasons, and 179 pairs of back-to-back seasons (17 non relegated teams per season X (11-1) seasons).

How repeatable is scoring penalties?

Remarkably this plot is even more of a mess than the original (which may be found below). In short, the ability at the team level to score penalties over the course of a Premiership season is 99.5% luck driven and a mere 0.5% skill driven. In other words the number of penalties a team score in one year bears absolutely no relation to the number that it will win the next time around.

How about the number of penalties a team concedes?

This time we have a larger skill component, but it’s still only 3%, whereas the other 97% is luck at the team level over the course of a Premiership season. Once more luck is by far the dominant factor, and we’re best off assuming that a team will concede the league average number of goals from penalties in a given season.

And finally, how about a teams penalty differential (goals scored from penalties minus goals conceded from penalties) – is there any repeatability there?

Well there’s certainly more. But the breakdown is still a meagre 8% skill/92% luck at the team level over the course of a Premiership season. In short, even the best penalty differential in year 1 (+8) is still worth less than a single goal when we’re predicting that teams penalty differential in year 2 (the +8 team actually had a -1 penalty differential in year 2).

Basically, variance in the number of penalties scored/conceded by a given team over the course of a season is very much luck driven.

Original post

Yesterday I posted two articles looking at the repeatability of shooting percentage from inside and outside the box are the beginning of a long series looking at which metrics are repeatable at the team level over the course of a Premiership season. At the end there’ll be a summary, which I’m aiming to have as a useful resource.

Today I’m looking at scoring and conceding goals from penalties. Note – this post only considers penalties that result in a goal. In the future I’ll do the same study but incorporating all penalties that are awarded. The data set used here spans the 2001-02 – 2011-12 seasons. In total there are 220 team seasons, and 179 pairs of back-to-back seasons (17 non relegated teams per season X (11-1) seasons).

There are two plots here. Firstly, how repeatable is scoring a given number of penalties from season to season?

Now that is a beautiful mess of a plot. Not only is there almost zero positive slope to the line of best fit, the correlation is also terrible. To be specific, scoring a given number of penalties is 4% skill driven and 96% luck driven at the team level over the course of a Premiership season. You’re best off assuming that each team in the league will score 3.6 penalties, or one every 10-11 games.

So does the same hold true for conceding goals from penalties over the course of a season?

There are two remarkable findings here. Firstly there’s a negative correlation between goals conceded from penalties between years 1 and 2, and somewhat remarkably the correlation here is even worse than for goals scored from penalties. Specifically the gist is that conceding a given number of penalties is 3% skill driven and 97% luck driven at the team level over the course of a Premiership season. Once more, you’re best off just assuming that a given team will concede 3.6 goals from penalties over the course of a season.

Finally, let’s look at penalty goal differential. We know that some teams win more penalties than others, and have a larger differential in terms of penalties won vs penalties lost, but is that a repeatable metric when it comes to scoring/conceding goals from penalties?

Now that’s interesting. The plot is still a mess, but the correlation has improved. Albeit only slightly. The breakdown is that penalty goal differential is driven 7% by skill and 93% by luck over the course of a Premiership season.

Now it is worth saying that, although 800 penalties have gone into producing each plot, a given point still represents only a small number of penalties. In other words there’s scope for these numbers to move, but the correlations are so tiny that any movement would still leave luck as the overwhelmingly dominating factor.