NFR - Why 'Moneyball' Wouldn't Work In Football

[White Noise]

http://www.timesonline.co.uk/tol/sport/columnists/gabriele_marcotti/article7000809.ece

Why the statistician can never provide all of the answers

Gabriele Marcotti wonders whether the theories of Michael Lewis's famous book on baseball, Moneyball, could be applied to football

Gabriele Marcotti, European Football Correspondent

3 Comments

Michael Lewis is a former bond trader turned financial journalist, who, in 2003, sent shock waves through the world of sport with Moneyball. The book examined the work of the Oakland A's, the baseball team, and how they managed to compete with bigger, wealthier franchises largely by avoiding conventional wisdom in favour of innovative statistical analysis.

The basic premise was that teams have finite resources (read: money) and are therefore forever looking for value. While that "value" was generally assessed by evaluating traditional parameters, many of which were subjective — pace, strength, character, etc — you could identify value more efficiently by analysing certain previously obscure statistics.

His work spawned criticism and scepticism from the baseball establishment — the old-school types who trusted their eyeballs and believed you could "look" at a guy and determine whether he could play — but also scores of imitators, and not just in baseball. Lewis became the darling of a certain kind of progressive executive and, to this day, enjoys plenty of admirers in football as well.

Finding a way to apply to football the principles laid out in Moneyball has been something of a holy grail. The most obvious obstacle is that while baseball has easy-to-measure individual match-ups, football is a free-flowing game.

Last year Lewis wrote a piece on Shane Battier, the basketball player, and how his team, the Houston Rockets, come closest to replicating the "Moneyball principles" by studying a raft of new statistical categories. And because basketball's ebb and flow is closer to football than baseball, it raises the question of how soon football will have its own Moneyball types.

The Rockets believe Battier is one of the best players around, even though his statistics — he does not score or rebound much and gets few assists — suggest otherwise. Indeed, there are only two statistical categories in which he is above average.

The first is "plus/minus", which basically means that his team tend to outscore the opposition when he is on court. The other has to do with the scoring average of the players he is guarding. Battier generally man-marks the opposing team's best player, who generally performs less well when being guarded by Battier than when facing other teams.

The upshot of those stats tells you that Battier is, presumably, a very good defender — either that or extremely lucky.

But they do not tell you why. And that is what the Rockets' internal statisticians try to measure. They do not share the information, but they try to quantify things such as his movement off the ball, the way he sets himself when defending and even his little unseen tugs or shoves aimed at breaking opponents' rhythm.

Is this stuff that valuable? And can it be applied to football? In theory, you could measure many of the same things. For instance, rather than simply saying, "Xavi Hernández is a creative passer", you could measure the percentage of his passes that find a team-mate and leave Barcelona in a better position. Or, instead of saying, "John Terry reads the game well", you could quantify the occasions on which his placement not only leads to an interception but prompts the opponent not to make a certain pass.

Along the way, you would have to deconstruct matches into their most basic parts, recording the actions of every player during every second of the game. No doubt, there would be some value in all this.

Except — and here's the twist — the more you delve into this kind of analysis, the more it necessarily shifts from the objective to the subjective. Who decides if a striker's run really did suck a defender out of position or if the latter wandered off of his own volition? Who determines whether a pass is accurate or whether the team-mate simply made the wrong run?

The point is that a computer could not compile this kind of analysis, certainly not if it is to have any value; it takes a human being, with his own biases and judgments. Which, when you think about it, brings us back to square one: personal opinions based on conventional wisdom. Ultimately there may be a role for this kind of objective microanalysis in football, but it will be only as useful as the subjective judgments of the compilers.

[RidgeRider]

Lewis wrote a book, but Bill James came up with the measurements/statistics originally and Billy Bean of the Oakland Athletics chose to use them as competitive advantage. 

Reading the piece it kind of sounded like Lewis was getting the credit for the idea but he is just a writer.  Lewis also wrote a famously good book named "Liar's Poker" that chronicled the lead up to the market crash in October of 1987, and it's aftermath and evil characters (remember Michael Milken??) and the derivative instruments that were involved. 

Sound eerily familiar to anything recently?   

Junk bonds = Sub Prime Mortgage Backed Securities. 

Anyone? 

Buehler?

[finnster01]

This is eactly why Monte Carlo simulations (random computer simulation of all possible outcomes) in finance and economics does not work properly. The reverse is also true in the sense that history is not necessarily the best predictor of the future, and a human can't do it all on his own depnding on the decision sample that is available to him.

So what you really need in all of these things including football, is historical data, loads of computer horse power and a human brain or two to make sense of it all at the end of the day. And you'll still be buggered!!

[dancrawford]

Very interesting. Really enjoyed reading Moneyball when I got into baseball.

Why England Lose is a very good read too.

[Steve_orino]

Pretty long read but for those who enjoy the analysis of sport, it's a must.

http://www.nytimes.com/2009/02/15/magazine/15Battier-t.html