Saturday, March 16, 2013

Sack The Coach!: The Effects Of Coaching Changes In The AFL

As someone who tires quickly of the seemingly endless media speculation regarding which AFL coaches are under pressure to hold their jobs, I was really looking forward to showing that coaching changes actually make little difference at all. That’s what I thought anyway. To show this, I was going to compare the changes in teams’ winning percentages the season after changing coach with the changes in teams’ winning percentage in all other seasons.  Well, I did that, and you can see the results in the histogram below.

It turns out that, since 1998 over a quarter of the time when a team changes coach it has improved its winning percentage by over 0.20 percentage points (over 4 wins) the next season! Less than 10 per cent of the time has a team’s winning percentage fallen by over 0.20 percentage points. By contrast, when there has been no coaching change, only around 10 per cent of the time has a team’s winning percentage increased by over 0.20 percentage points the next season, and over 15 per cent of the time the team’s winning percentage has fallen by over 20 percentage points.
To put all this into a neater summary statistic, the average increase in a team’s wins the season after it changes coach is 1.5 wins (compared to -0.2 wins if it doesn’t change). That’s a pretty handy pick-up. However, it could be that a coaching change often occurs when the team just happens to have a stretch when it is performing below its long-run capabilities, and the improvement we see on average when a new coach comes in is just a case of mean reversion. As it turns out, when I looked to see if anyone had suggested this before, I found that a study that looked at coaching changes in Italian football had suggested pretty much the same thing! They concluded that, once you do some controlling for other factors that change when the coach changes, changing a coach does not seem to have a significant impact on team performance.

I’d like to think then, that if I could be bothered to do the econometrics, my first theory would still be proved right in the end. In the meantime though, maybe teams looking to boost membership after a brutal coaching change could send out the above histogram to their fanbases.

No comments: