Things happen when a top club draws with a squad at the bottom of the fourth division, even when the top club played a team of young prospects. People start wondering if those young prospects really have what it takes. People start to speculate about how the young prospects arrived at the club. And people start to point fingers.
A few months ago I asked if it was possible that the vaunted Liverpool analytics department, with its seat on the club’s transfer committee, simply wasn’t very good. After all, the public knows almost nothing about its activities and its influence within the club; it could be good and ignored, or bad and heeded, or just unlucky. I was duly assured by Garry Gelade, another consultant-cum-academic, that Liverpool’s analytics supremo Ian Graham was “the best in the business by a long shot.” Well, I haven’t seen any of Ian’s stuff recently (and I’m pretty sure Garry hasn’t seen much of mine), but I have an idea about what might really be going on at Liverpool.
We know that clubs with higher wage bills in the Premier League tend to do better, especially in the upper reaches of the table. But we like to think that clubs using analytics can get more bang for their buck – by identifying talent more precisely, finding bargains in the market for players, honing tactics, and the like. So a quick test of whether a team’s analytics are working is to ask whether, given its wage bill, performance was above or below what we’d expect. (Of course, managers may also be responsible for over- or underperformance, but their effect should be relatively constant; we know underlying performances still vary even with the same manager.)
One of the strongest relationships between wages and points in the Premier League table is based on the log of the ratio of wages to median wages, though log wages alone have been about as good in the past several seasons. I’ll use the latter here, adjusted for inflation, since I don’t have median wages for 2014-15 yet. Below is a chart of log wages (in millions of pounds) versus points for the past five seasons, roughly since Fenway Sports Group took over Liverpool. I’ve highlighted both Liverpool and that other analytics-heavy club, Arsenal:
Liverpool finished below par for four seasons and well above in one – I think everyone knows which – while Arsenal were at par or slightly above in all five seasons. The two clubs had fairly similar wage bills throughout this period, but their aims may have been very different.
Arsenal is well known for being the Mr. Consistency of the Premier League, always finishing in the top four yet rarely winning the title. Liverpool’s owners, by contrast, like to win things, and its fans still dream of repeating past glories. Yet with an overall wage bill lagging the big spenders in Manchester and London, Liverpool’s only hope of taking home a trophy is to gamble. As someone from the financial world might say, they need variance.
To win something, Liverpool must roll the dice every year. In some years the results may be terrible, but in others they just might be fantastic. So how do they do it?
In my evaluation of players during this transfer window, I’ve noticed several different risk profiles. Just as with financial assets, there are players with high and low risk-reward tradeoffs. A Silicon Valley tech stock is like a young striker who’s had spectacular performances over a small number of minutes, while a U.S. Treasury bond might be a steady defensive midfielder who’s shown himself to be neither a star nor a liability over several seasons. And just as in finance, the returns on different assets may be related.
Every team needs a few Treasury bonds, but Liverpool in particular needs tech stocks. Not surprisingly, the analytics department catches a lot of flak from fans when transfers and results seem to go wrong. But Liverpool fans who want to win something have to realize that gambling is the only way, and sometimes gambles don’t pay off. Likewise, if Arsenal fails to win the title this season, its fans may find themselves asking whether they wouldn’t like just a bit more variance.