Why risk isn’t always the enemy in player recruitment
Let’s pretend for moment you’re a technical scout, preparing to add to a dossier of reasonable transfer targets for your club to cover the next three seasons. Being an analytics minded person, you are armed in your search with statistics on the effects of age on performance, advanced metrics that measure innate skill in individual players, informal data on player “agreeability”, and filters for realistic transfer fees etc. etc.
Your ostensible aim here is to use data to objectively measure innate skill and adaptability for the purpose of reducing the risk the transfers will ‘fail.’ After all, failure in the transfer market is expensive, in more ways than one. Not only does a ‘dud’ present something of an opportunity cost with funds spent that could have been used on a better player, but they may also negatively affect an otherwise decent team. Many believe that the use of accurate predictive data to reduce the number of bad player transfers should be one of the chief aims of the football analytics movement.
Of course, even the most advanced data can reduce the risk of a poor transfer, but they cannot eliminate it altogether. Even star players who are perceived as “sure things” and priced accordingly can fail, and fail spectacularly. One thinks here of Chelsea’s infamous £50 million Fernando Torres transfer in January 2011–Torres scored 81 goals in 142 appearances with Liverpool, but only managed to net 45 in 172 games for Chelsea.
This can cut both ways of course. Sometimes a bet on a player might turn into something special, as with Fulham’s largely successful £2 million transfer of Clint Dempsey in January 2007. Often a player will “click” at a particular club at a particular time for a host of reasons that no level of predictive data can foresee.
However these are relatively extreme examples; despite how transfers are generally portrayed in most football media, I’m not sure the “lottery” model of assessing transfers as either “successes” or “failures” is a particularly useful one. Some January transfers hit the ground running for, say, six months, and then quickly fade away. Others are initial disappointments who take a while to break into the first team, perhaps needing more time to adapt to their new manager’s general approach and coaching philosophy. Some players add enormous value to a team even if they themselves aren’t much better than average, say, by allowing another star player to skip some cup ties or a league match or two to recuperate and avoid injury without being so bad as to negatively affect the team.
Though I’m not certain, I suspect even the most gifted technical scouts can’t always predict how exactly new additions will work out at the club. That’s why good teams will be proactive in working with a new player to best suit the needs of the collective. Even if a major signing is not an instant success, they may still offer value down the line or in a different role. At the same time, teams should also have a contingency plan in place should a transfer deal prove a total disaster. As Seth Godin recently wrote in a post on uncertainty vs risk, “…a range of results, all uncertain, does not mean you are exposing yourself to risk. It merely means you’re exposing yourself to an outcome you didn’t have a chance to fall in love with in advance.”
This is also why I think savvy technical scouts might include in their dossiers (and perhaps already do) statistically riskier transfers that nonetheless offer potentially big (and relatively cheap) upside with potential little downside. They might for example use data to identifier a class of slightly older, more injury prone strikers whom they believe could be effective with a smaller number of starts and could still be potentially sold on within a season or two. They could use that information to set a reasonable transfer/salary ceiling which takes into account things like relative resale value, or lack thereof.
The idea here would be to use data to assess the transfer market as a whole, not cut entire, potentially valuable swathes of it out in the name of reducing or eliminating risk. This approach offers several advantages, including offering cash-strapped teams more affordable options. While in the near term analytics-minded clubs will potentially gain a ‘moneyball’ advantage by using data to identify talented, low risk prospects that other teams overlook, as predictive stats permeate football in the coming decades, smarter clubs will need to find other ways to beat the market. That includes using data to carefully identify riskier but potentially effective transfers.
It helps to think of the connection between risk and transfer market fees. A 23 year old star striker with 40 goals last season will command a double digit transfer fee because they are a star, yes, but because they are perceived are more of a ‘sure thing.’ In other words, transfer fees have an inverse relationship to (perceived) risk. This, and because even disappointing transfer can still offer potential value, is why it’s important for recruiters to first to know and quantify the risky business of the transfer market, and not get caught up in a dogged and misguided search to remove it altogether.