FC 100: Explaining Player Value, new metric with Football Whispers
What is Lionel Messi worth to Barcelona? What would he be worth to Manchester City? How much would it cost to bring him to Manchester? These questions may initially seem almost identical but in reality they are anything but.
The concept of the "value" of football players dominates coverage and debate, but it is very difficult to nail down conceptually. Is a player's value what a club should pay, or what they would have to?
In an article for the New York Times, Rory Smith describes how an expert in the field, Henry Stott of Decision Technology, distinguishes between "intrinsic value" (what a player should be worth) and an "intuitive price" (what a team may end up actually paying). For instance, when Paul Pogba became the world's most expensive player at £89 million following his move from Juventus to Manchester United in the summer of 2016, few would have valued Liverpool's Philippe Coutinho, one of the best players in the Premier League, at anything close to that. Yet after a series of events, most notably Neymar's world-record switch from Barcelona to Paris Saint-Germain, the Reds reportedly rejected bids of approximately £120m from Barcelona this summer.
Were Liverpool right to turn down those offers? Were Barcelona simply throwing money around in a desperate attempt to replace Neymar? Ask the average football fan and they will have strong opinions about that transfer as well as the activities and rumours of the club they support.
At Football Whispers, we want to bring some objectivity to these debates, without providing a concrete answer. To do so, we have started with our version of "intrinsic value" -- our "Player Value" figure is what a player should sell for in the current market based mainly on their performance, age and position.
While it makes sense for ability or performance to dominate a player's market valuation, it is also the most difficult aspect to model. Luckily for us, it is a task that we work on outside of market valuations.
Our Team Performance and Player Performance models take a top-down approach. We separated team performance into attack and defence, looking at the key features and characteristics that make a good and bad side. We then boiled these down to the player level, crediting players for positive contributions while also penalising them for actions that were not in the interest of their team. Through holistic consideration of a player's actions, we get a more comprehensive profile than simple individual statistics such as goals, assists or appearances can provide.
We also contextualise performance by competition: therefore the best player in the Premier League is rated significantly higher than his counterpart in the Scottish Premier League, for example. But defence is, both statistically and generally, much harder to model in football than attack.
Match data -- on-the-ball actions recorded by companies like Opta, Stats, and Stratagem -- is limited by only recording information when something happens. This works well for modelling a team or player's attacking ability as a strong offensive contribution has to, by definition, involve some sort of action. Defending, however, is more complicated. If anything, good defending at a team level is the absence of "something happening." Pressurising an attacker successfully might delay their attack and limit their passing options without resulting in a tackle, interception or ball recovery.
So it is unsurprising we found our measure of attacking performance is significantly more important to a player's market value than their defensive ability, even for positions like central midfield or full-back where these responsibilities might be equally shared.
To some extent, we're simply reflecting the unfairness that comes from our sport-wide inability to value defenders; fans prefer to watch attackers and it is easier for us to see their contribution, so they tend to go for and earn more money. There are a few exceptions but the genuine superstars of football are the players that get goals rather than the ones that stop them.
The relationship between age and value is more complicated than is generally appreciated. The assumption is a player's value should rise to a peak age and then depreciate. Generally, a footballer's maximum value is at around 27 years old before dropping off into their 30s.
However, our "wonder-kid premium" can affect value. Take three strikers, all of identical ability who play in the same league. One is 18, one is 24, and one is 30. If the 24-year-old is worth £20m: should the 30-year-old be worth more or less? The answer to that one is clear because the older player is likely to have fewer years in him and so has a lower resale value, but the youngster is worth more than the 24-year-old because he will have more years in which to improve and continue the same output.
Our "wonder-kid premium" persists until about 21 years old, with values dropping off a bit before a peak-age premium eases a player into the steady decline we expect in value for their 30s. Currently our age effect is consistent across positions and abilities, but this is a point of research for us in the future.
An inflated market?
As we are trying to model Player Value, not their price, we've decided not to include contract length information. Alexis Sanchez and Mesut Ozil, for example, are available for clubs on a free transfer this summer despite their value standing at £83m and £65m in our model respectively. If and when anyone manages to sign them on a free, without an astronomical premium in wages compared to players of similar abilities, they are undoubtedly bargains.
Another important feature of our model is factoring in the consistent inflation of transfer values from season to season. Although there is a lot of commentary from pundits, managers and fans alike about how the market has gone "crazy," this inflation is largely explainable through the increase in club revenues due to TV deals and increased global demand.
In short, the Neymar transfer didn't break the market; it just epitomised it.