I am really flattered and humbled by the attention drawn to my piece on Defensive Efficiency Rating (DER) a couple weeks back. It demonstrates that there clearly is a demand for soccer metrics.
Before I go any further, I need to stress the fact that this statistic is very much a work in progress. This is a theme I will be coming back to several times in this post.
The formula I have used is three separate equations which I'll call DR, DR, and DR.
The point of this stat is to put a value on being able to disrupt ball movement. The numbers are arbitrary and to be honest passing accuracy is somewhat misleading: a five foot pass between two centerbacks under no pressure counts the same as a contested cross in the opponent's box.
It's arguable if this is even a factor. Take Chivas USA for instance, in their five games their passing accuracy has been 68% with the league average roughly 74%. Meanwhile the Seattle Sounders are sitting on one point in four games despite having a passing accuracy that has never dipped below 76%.
This measures the defense's ability to deny possession but then also take into account how successful the defense was in allowing attempts on goal. For instance, if the opponent had 60% of possession but only a handful of attempts on goal, that's a good thing.
As I mentioned last time, it's disingenuous to credit a defense with a clean sheet when a goalkeeper has been marooned and forced to make multiple saves. The biggest problem here is that in clean sheet games, you're multiplying by zero. I am going to say it right now--this is really stupid equation on my part. So already: we found our first thing to fix!!
Tying it all together:
After summing up equations, multiply by 100 and divide that by six. This part was taken straight from the quarterback rating in football. Each team's score is averaged with their other performances.
DER is experimental, and has some flaws and is far from a finished product. As the season progresses, it is almost certain that the DER formula will be tweaked. Frankly, I'm excited by opening it up to people smarter than me. Together we hopefully can construct tools to make a robust and useful gauge of MLS performance.
If you are interesting in participating, please join in the the project's Google group at https://groups.google.com/d/forum/soccermetrics. Code sources can be found at: http://code.google.com/p/mls-advanced-stats/
Head on over and jump right in!
What do you think? Leave a comment below!