Friday, October 5, 2012
The Houston problem
Just the other day I came across a piece Michael Lewis published in early 2009. Lewis, as you may know, is the guy who wrote Moneyball, now made into a movie starring Brad Pitt. One of Lewis' pet subjects appears to be the role of stats and figures and numbers and data in sports.
Said article, you see, is called "The No-Stats All-Star" and profiles Shane Battier, an NBA player then with the Houston Rockets, now with Miami Heat. (Or maybe that should read "with the Miami Heat" - I can't seem to figure that one out.)
What fascinates Lewis about Battier is that the player seems to be nothing special and that his stats are at best mediocre. "And yet every team he has ever played on has acquired some magical ability to win," Lewis adds. Of course magic doesn't really come into play here. Rather, the conventional methods of stat-keeping in basketball fail to register what Battier brings to the game.
To use a football analogy, it's almost as if Battier is one of the greatest providers in the game but hardly anyone knows about it because the stats only tell you who scored the goal, not who played the final pass or crossed the ball - who, in the vernacular of the trade, "made" the goal.
You may think that this analogy is askew, because assists are such a vital football statistic that nobody would forget to note them. Yet kicker magazine, the bible of German football which has covered games since 1920, began publishing this particular statistic only on July 25, 1988. Until then, the only yardstick we had to measure performances of players was the number of goals scored or conceded. And, of course, the legendary player ratings dished out by kicker since August 1963.
Naturally, these were as subjective as another statistic kicker introduced together with the so-called "assists" in 1988, namely the number of goal-scoring opportunities a team created during a game. Obviously, it was left to the discretion of the magazine's beat writer to decide what qualified as a proper chance.
As goes without saying, one reason the game has a long history of preferring subjective opinion to objective analyses is that the sheer number of players in combination with the flow of the game makes it hard to break football/soccer down into easily observable situations and quantifiable behaviour. It's very much unlike baseball and gridiron/American football - sports you could call a succession of set pieces interrupted by only a few seconds of general motion.
It appears that even basketball, which has only five players per team and follows a rhythmic pattern of attack and defence which makes for predictable player positioning, cannot be accurately described through data. Or, as Lewis put it in his piece, "basketball has measured not so much what is important as what is easy to measure" and the resulting figures, he argues, "have warped perceptions of the game".
Yet there seems to be one stat that is meaningful: plus-minus. It describes "what happens to the score when any given player is on the court", Lewis says in the article. It is the one category in which Battier does not look like an at best average NBA player but like a perennial All-Star. It also happens to be the statistic the Rockets value the most because it reflects the only thing that counts: your value to the team.
All the other stats that look good on paper and will have a player ask for a pay rise, Lewis says, could actually be accumulated to the detriment of the team in very subtle ways. He quotes Daryl Morey, the general manager of the Rockets and a big admirer of Battier, as saying: "We think about this deeply whenever we're talking about contractual incentives. We don't want to incent a guy to do things that hurt the team."
A few years before Morey was signed by the Houston Rockets to change the club's approach to building a team, there was a young lad in Bristol who looked very closely at European basketball. Not so much because he liked the game but because he felt it lent itself to investigation. He said he wanted to "use some of the advanced statistical analysis that has been developed in the past two years mainly relating to NBA basketball to evaluate player performance in Europe". His name was Steve Houston. (I'm not making this up - it's just a coincidence.)
In 2004, Houston applied for a traineeship with the Rockets and was assisting the club's European scouts within six months. In early 2006, the Rockets signed that great plus-minus advocate, Daryl Morey. "I learned a lot from Morey about how to translate data into information that will really help you to identify talent and find new players," Steven Houston told the German footballer writer Christoph Biermann a few months ago.
In 2008, Houston - the man, not the club - was approached by Mike Forde, the director of football operations at Chelsea FC. At that time, the Dane Frank Arnesen was also working for Chelsea, first as head talent scout and then as director of football. When Arnesen moved on to Hamburg in May 2011, he took Chelsea's chief scout, Lee Congerton, and no less than five young players with him. Oh, and Steven Houston.
Since then, Congerton's trying to improve Hamburg's scouting system with the help of Houston's data. Congerton told The Guardian that statistics played an important role when Hamburg signed the Latvian striker Artjoms Rudnevs from Lech Poznan, as his numbers were "exceptionally high". And Houston told Biermann in June that the club had signed goalkeeper René Adler in part because he excelled in certain statistical categories.
Back then, many people questioned the logic behind signing Adler, considering Hamburg already had a pretty good goalkeeper in Jaroslav Drobny. But now, after six rounds of games, the transfer looks like the most inspired deal Hamburg have made in a long time. Unless, of course, Rudnevs keeps up his pace of improvement. He was a sub in the first game and not very good in the second, but he's scored two goals and set up one more in the four matches since.
However, the numbers men at Hamburg appear to have been using fairly normal stats, such as goals scored and shots parried. It seems safe to say that Houston, the Morey disciple, would love to apply plus-minus to football.
Naturally, plus-minus is perfect for games like basketball and ice hockey (where it's been in use since the 1950s), games that have an unlimited number of substitutions, so that players get into and out of action all the time, and in which the score often changes. You'd think it's useless in football, where there are only few changes to the line-up during a game, where a player who has been taken out of the game can't get back into it and where the score might change only one or two times over the course of 90 minutes.
But perhaps you can tweak plus-minus to suit football? There is now a German company called score.chart (makes you wonder why even number crunchers have to use silly spellings) that has come up with a method not unlike plus-minus to evaluate football players. Mario Gotze has a score.chart index of 106.0 and Mike Hanke's is 102.8, while Hans Sarpei comes in at an even 100.
This seems to indicate that Gotze is better than Hanke, who in turn is better than Sarpei. You could say one doesn't have to do maths to arrive at this result, but if you replace these three names with those of, say, three Latvian players you've never heard of, let alone seen in action, you can understand why some scouts are interested in such charts.
But much more intriguing is another question. Will the stats wizards one day find the footballing equivalents of Shane Battier? Are there players out there whom nobody rates particularly highly because they don't do any one thing particularly well but who should actually be starting players for the very simple reason that the team as a whole performs better when they are on the pitch?
After all, as many of us will know from their working lives, there are some people who bring something intangible, unclassifiable to the workplace - "some magical ability", if you will - that greatly improves the results. And that's what it's all about, isn't it?