Hat tip to Sam Gregory for giving me the idea for today’s post.
When Cristiano Ronaldo won the Ballon D’Or, this blog’s resident pro-Messi troll came out of the woodwork using a crude homophobic play on Ronaldo’s name to make a series of rambling arguments as to why the Portugal and Real Madrid winger shouldn’t have won the top prize in world football.
I won’t reprint those comments but I’ll give you the gist of them—Ronaldo shouldn’t have won because he was sent off in the Copa Del Rey final against Atletico, Ronaldo shouldn’t have won because Mesut Ozil created most of his goal-scoring chances (which of course doesn’t explain his incredible goal-scoring record since the German international went to Arsenal), Ronaldo shouldn’t have won because he didn’t make enough assists, didn’t have the same goal-per-game average as Messi (not even a decent metric as we’ll see), didn’t win “trophies”.
All of this selective nonsense highlights why I not only don’t think much of player awards, or po-faced attempts to objectively rank footballers in general. Not because I think all players are equal (which is obviously not true), but because finding objective markers to easily rank one player above another is an extremely difficult thing to do. I think Johan Cruyff said it best back in 1994 in an interview with ESPN, remarks which were recently Tweeted out by Dutch football dude Mohamed Moallim:
Ballon d'Or gala looming; extract from a Cruyff interview (1994): 'best player of the world' doesn't exist pic.twitter.com/HjIHLXz3E3
— Mohamed Moallim (@jouracule) January 11, 2014
What Cruyff is elegantly saying here is that you can’t simply point to a single, Platonic form—footballer—and then look among those playing today for a perfect or near-perfect exemplar. That’s just not how the sport works.
For example, at first glance, voting for players based on trophies won—the reason some believe put Ribery on top of most journalists’ ballots—makes sense. If a team won more trophies than other teams in Europe, and that team has a “best” player, surely that player is therefore the best in the world.
There are a few problems with this and they should be obvious. One is that teams field eleven players under the leadership of a manager. That team’s ‘best’ player might have been a key factor in the team’s success, but the onus is on the voter to make the case that player’s ability exceeds not only the contribution of his team-mates and the tactical influence of the manager, but also every other player in the world. When those players include the likes of Lionel Messi and Cristiano Ronaldo, that’s a tall order.
Put another way, if Bayern traded Ribery for either Ronaldo or Messi, would they be a better team? Of course they would, dummy! So Ribery can’t be the best in the world, right? Except even then one could make the argument that sometimes a particular player is best for the a certain team at the right time. Perhaps that player was Ribery. The point is, it’s not obvious at all (I haven’t even mentioned how knockout competition trophies like the European Cup are aided in large part by random variation, which means we shouldn’t give them that much importance anyway).
This is perhaps where you’d expect me to argue that analytics can help offer a reliable, objective way to rank players where subjective opinion fails, but it’s not that simple. However I do think player analytics in football can certainly offer clarity where goal and assist tallies alone cannot. If you want a great example, check out 11tegen11′s latest work. The author starts with simple goal tallies and then gradually adjusts the metrics to try to offer a clearer picture of underlying talent, finally producing a slightly amended, counter-intuitive list of the top scorers in the Eredivisie:
In this post, we’ve come from a traditional list of names and goals scored, to a sophisticated metric to judge goal scoring talent in its most honest way. It seems creating chances for yourself, or allowing team mates to do so, is a different skill from finishing those chances. Only the true top strikers blend these skills.
This is a brilliant and useful tool for player evaluation, but could it be used as an objective measure to rate the best player in the world? Maybe, but I would need some questions answered first (and please do in the comments if you have answers). Like, for example, the repeatability from game to game, season to season, year to year, of above average conversion rates for a particular player taking shots from certain areas of the pitch, a stat which presumably influences the Expected Goals (or ExpG) metric the author cites in his evaluation.
Because we should never simply look at a positive player metric and assume it reflects an innate, underlying talent. To give you a sense of what I mean, James Grayson tweeted the link out to this old blog posts on the now defunct hockey analytics site authored by Vic Ferrari, Irreverent Oiler Fans, and I think everyone should read it, on the concept of likelihood in sports. The basic idea here is that no one should assume a single positive player statistic is evenly distributed along a Gaussian curve, when some metrics like shot conversion rates in fact heavily swing one way or another each season, reflecting (seemingly) random variation.
You might argue in this instance repeatability in something like converting shots into goals doesn’t matter, because awards like the Ballon D’Or are a snap shot of who’s the best “at the moment.” Some of the variation in shot conversion for example might be attributed to “luck” in the sense of it being random (and therefore not within the player’s ability to control), but as Daniel Altman eloquently argued recently, some of it may be variables that we haven’t yet accounted for:
Of course, own goals and penalties are just two possible sources of luck. Injuries, weather, mascots mocking strikers with short tempers… these and many more might contribute as well. Indeed, to a given team, anything beyond its control that might affect the outcome of a game could look like luck. But to analysts, the majority of these factors are simply things we don’t measure. They may look random, but they are just variables omitted from our models.
Some of this variability may or may not be within a player’s ability to control. We just don’t know yet.
All this and we haven’t even discussed why goal-scorers are innately “better” players than defenders, or goal-keepers (for an interesting look at what makes defensive mids good, read this), and how to best evaluate other positions against each other.
So you can see the problem here. We may “believe” that some players are just better than others, but this is an intuitive guess, nothing more. Arguing for this guess as if your life depended on it is a waste of the short time we have on this planet.