Perhaps some transfers are worth the money

Readers of Soccernomics know that one of the key conclusions made in the book is that transfer fees explain very little of a team's success via average table finishing position. Authors Simon Kuper and Stefan Szymanski conclude:

In fact, the amount that almost any club spends on transfer fees bears little relation to where it finishes in the league. We studied the spending of forty English clubs between 1978 and 1997, and found that their outlay on transfers explained only 16 percent of their total variation in league position.

The statistician in me wonders what the Pearson correlation coefficient was of that data so, just so I could understand if that variation is even statistically significant. But I digress...

Perhaps the authors of Soccernomics were looking at too wide of a data set when attempting to quantify the effects of transfers. Just like the stock market, transfers are a bet by one team that they can get a better value out of a player than the asking price of the team who currently owns the player's rights. And just like the stock market, a good number of people bet wrong and thus lose money. In soccer, a good number of managers may be betting wrong and thus leading to a poor correlation. Maybe we shouldn't look at all of the clubs, and instead look to the successful clubs. Let me explain.

I have been lucky enough to have Paul Tomkins pick up on my blog-related tweets, and a single re-tweet from this man can elicit ten times the re-tweets from his loyal followers. A week ago he recommended that I check out his Transfer Price Index (TPI) where he has been able to assemble all of the transfers during the Premier League era into a single database for comparison. I finally got a few hours today to work through the posts he has made so far, which I summarize below. In general, I would say it is a must read as it does a more detailed analysis of transfers than Soccernomics and comes to some different conclusions.

In one post, Tomkins analyzes the effect of gross expenditures of teams.

What’s most interesting is what happens to most clubs in the year or two after a particularly big Gross spend (+10% of the total Premier League outlay).

In total, this figure has been exceeded on 31 occasions over the past 18 years, with 12 different clubs managing to do so at least once, and with two (Manchester United and Chelsea) breaching it on no fewer than five occasions.

He then goes on to explain the trophies won by teams the season after such expenditures are made. Without doing a statistical analysis, the case for team improvement is compelling.

In another post, Tomkins then compares the net expenditure of teams, allowing sales that offset purchases to tell a more complete story.

If the fact that Chelsea accounted for 39% of the entire Premiership’s Gross spend in 2003 was incredible, that it rises to 67.9% of all of that season’s Net outlay is quite jaw-dropping. This is a club that was buying, buying and buying, with precious little selling involved. Whereas major clubs normally at least have to sell a star or two for profit to reinvest, this was a pure purchasing machine. A year later it was followed by a further 49.5% of the entire division’s Net outlay.

Looking at net spending helps us understand who's spending far more money than others, who's driving up league debt levels, and which squads are winning in the transfer game by consistently buying low and selling high. I will leave it to readers to click on the links and read Tomkins' awesome conclusions.

Tomkins' approach is superior to that of Soccernomics for one simple reason: transfers are one time fees that are speculative in nature and often used to either plug a short term hole or pay off immediately by shoving a good team to the top. Looking at them on the average, and only looking at the buyer's results, is a bit simplistic and likely to miss the short term payoff of the transfer. Tomkins' approach captures this effect. I am hoping that his partner, Graeme Riley, uses some of his advanced statistics skills to produce solid time-series statistical analysis from the TPI data set. It could be very powerful in revealing the effects of transfers in the Premier League.