Friday, April 07, 2006
Moneyball Meets the Midfield
I'm a big fan of the book Moneyball and its approach to determining the worth of a baseball player. I agree that the prime attribute for a batter is "don't get out". Just get on base, any way possible, and keep the inning alive. And while you're at it, don't fall in love with stealing bases because there's usually not a good reason to give the other team an out 25% - 50% of the time. And don't worry if a defensive player yields 10 more errors a year as long as they are on base at least 10 more times than the next best option. Love it or hate it, Moneyball is a seminal book in the progress of evaluating baseball talent.
I want to bring Moneyball into the soccer discussion.
In my mind, the closest analogy to Moneyball's "don't get out" axiom is a soccer player who doesn't lose possession. However, since soccer is such a different game in the sense that it's played with a running clock, scoring is so much more infrequent and possessing the ball for 90 minutes doesn't necessarily mean that you'll score goals, I find that a focus on possession is not even close to the metric I hope to apply to soccer performance.
But modern baseball does have a more translatable statistic - it's OPS (On Base Plus Slugging). That measure rewards hitters who "don't get out" but then adds the degree of success they've had with their chances (their slugging percentage). Similarly, in soccer, I would like a statistic that rewards a player for possessing the ball but also adds points for creating goal scoring opportunities.
The best way to measure pure "goal creation" is via assists and scores. But, in my mind, that only tells half the story. I would like to add the number of times an offensive player "beats" a defender in his defensive half of the field. That statistic is very meaningful because it is difficult to break down a defense and a prime instigator is simply when someone gets past his defender. It seems like a basic premise but I've never seen the statistic anywhere. Similarly, a well placed through ball or a 40 yard change of direction pass can also have the desired result of beating a defender or defenders.
For the soccer statistic that I'm trying to create, I would add the possession rate of a player, the points via goals and assists, and the number of instances when defender(s) were beaten via skillful dribbling or inspired passing. The last element of the statistic would require some measure of subjectivity, but that's better than not adding it at all. This statistic would work well for attacking players and could be modified easily for defensive players.
I don't have the detailed formula worked out yet (hey, I just began thinking about it tonight!) but there is the rationale to fuel the algorithm. It would be very interesting to see the results! (Bostrom)
I want to bring Moneyball into the soccer discussion.
In my mind, the closest analogy to Moneyball's "don't get out" axiom is a soccer player who doesn't lose possession. However, since soccer is such a different game in the sense that it's played with a running clock, scoring is so much more infrequent and possessing the ball for 90 minutes doesn't necessarily mean that you'll score goals, I find that a focus on possession is not even close to the metric I hope to apply to soccer performance.
But modern baseball does have a more translatable statistic - it's OPS (On Base Plus Slugging). That measure rewards hitters who "don't get out" but then adds the degree of success they've had with their chances (their slugging percentage). Similarly, in soccer, I would like a statistic that rewards a player for possessing the ball but also adds points for creating goal scoring opportunities.
The best way to measure pure "goal creation" is via assists and scores. But, in my mind, that only tells half the story. I would like to add the number of times an offensive player "beats" a defender in his defensive half of the field. That statistic is very meaningful because it is difficult to break down a defense and a prime instigator is simply when someone gets past his defender. It seems like a basic premise but I've never seen the statistic anywhere. Similarly, a well placed through ball or a 40 yard change of direction pass can also have the desired result of beating a defender or defenders.
For the soccer statistic that I'm trying to create, I would add the possession rate of a player, the points via goals and assists, and the number of instances when defender(s) were beaten via skillful dribbling or inspired passing. The last element of the statistic would require some measure of subjectivity, but that's better than not adding it at all. This statistic would work well for attacking players and could be modified easily for defensive players.
I don't have the detailed formula worked out yet (hey, I just began thinking about it tonight!) but there is the rationale to fuel the algorithm. It would be very interesting to see the results! (Bostrom)
