Football's Invisible Heroes
Marcel Schmelzer & the Butterfly Effect in Football
Reality filtered by emotions
For 20 years I hold a season ticket for Borussia Dortmund. Of course, I stand on the famous Südtribüne. As you can imagine, watching a game from the yellow wall is quite emotional. Needless to say, the emotional part of the game implies for many people the beauty of the game. Being captured by feelings is what gives life meaning. However my analytical skills are mostly clouded by collective emotions. Instead of remaining a rational human being, I turn into a sub-part of the mass: hyper-angry, explosive and ecstatic.
Periodically players receive rage from fans, after apparent „bad“ actions. However, in hindsight, such actions often prove to be irrelevant. Yet we forget that instantly when the next apparently bad action happens. The same is true for positive movements, however, with the reverse effect on the judgment. A player who does spectacular things with the ball is perceived as “good”. In both moments, we are deluded by our narrow attention, which focuses mainly on actions in the here and now, and with the ball, or at least in its near proximity.
One player that was a frequent recipient of my personal frustration was Marcel Schmelzer. Yet, Schmelzer is a hero among fans because he played a lifetime for Borussia. He is a real Borusse, though, within a game, he wasn’t just a recipient of passes, but often the creator of confusion. He did not touch the ball particularly elegantly. He wasn’t very robust. He didn’t score a lot. And his passes didn’t seem quite accurate. At best, he appeared to be average for fans and the public
When a ball slipped through his foot again, I claimed (sometimes seriously), that I have better technical skills than him. Quite arrogant for someone who never played above an amateur football level. Even if true, that did not make him a bad player or myself a good one. Independent of my over-confidence, Schmelzer never seemed to be a „good“ player to the public. (Quite Contrary to the player’s judgment. Arjen Robben rated Schmelzer as his toughest opponent in the Bundesliga.)
But the problem here is not Schmelzer. The problem is the lack of a definition of what good means.
What do you mean by „good” player, exactly?
When we evaluate something, we have first to define what we mean by „good“ or „bad“. A shared definition is the precondition of a proper discussion. This is true for all categories and not just in rating a player. In Schmelzer’s case, maybe it is true that he was below average regarding his technical and offensive abilities. However, without evidence, we regress to mere subjective opinion and are stuck with the problem of definition.
Yet, such an opinion can even be supported by cherry-picking certain stats. Schmelzer for example only scored 3 and assisted 23 goals in 257 Bundesliga games. That means he only contributed every 809 minutes directly to a goal. Raphael Guerreiro his successor at Borussia, in contrast, shot 26 goals and assisted 29 goals and hence scored every 156 minutes. Does that mean that Schmelzer is roughly 5 times worse than Guerreiro?
Although such analysis and the conclusion are naive and too superficial, it makes the following point clear: From a reductionist statistical perspective, we can say that Schmelzer is worse than Guerreiro.
However, the critical question is: What does essentially matter in football? What makes a good player? How well do minutes per score or passing accuracy of a left back have to do with team success? How can stats hint at the true player quality? Can it ever?
Maybe.
We are aware that data is used to reveal meaningful patterns about a player‘s strengths and weaknesses. However, stats do not take the complexity of a football game into account. Because even if we find statistical correlations, that does not mean, that we have found universal truths about the game of football. Yes, big data is a valuable resource. But big data is also potentially misleading. Because correlation is not causation.
Regarding Schmelzer vs. Guerreiro discussion, I suspect almost everyone would prefer to have Guerreiro on his team. However, when we just look at the Goalimpact – the player’s impact on the goal difference, then Schmelzer Peak GI is significantly higher (175) than Guerreiro’s (162).
Why is that?
Football as a complex system
In order to understand why stats can be misleading, we have to make a short detour to complexity theory. Complex systems are composed of many interacting variables. Typical examples are biological organisms, ecosystems, or the financial system (although economists aspire hard to put it into categorizable boxes.) In such systems, small effects can have a tremendous impact. In weather forecasts, this refers to the “Butterfly“ Effect, which describes how a butterfly’s flap of wings in South America can lead to a tornado in North America.
What most people however misunderstand is the idea that any small effect must have these massive outcomes. More than the size of outcomes, the butterfly effect refers to the unknown and unpredictable consequences of effects. Because everything is intertwined, causal links are often hard to recognize. They fade away in the interdependent network. Therefore in a complex system, one must be skeptical of the notion of causality. The complex web of interdependent actions hides causality in multiple order effects. That doesn’t mean causal links aren’t there. It rather implies that cause and effect become opaque.
Thus, big data should always come with a prescription: Correlation is not causation, and the interpretation of disconnected stats lacks its complex element. Data is not useless but often accompanied by uncritical enthusiasm in our digital age. Even worse, focusing on the wrong stats can cause havoc.
My favorite foolishness example by probability expert Nassim Taleb goes as follows:
Assume you know that a river is on average four feet deep. Should you cross it?
No. Because the information about the average can be very misleading. It might turn out that the river is for the most part only 1 foot deep, but there might be one large hole of 100 feet into which you might fall. If you cannot swim, that could be quite dramatic. Of course, the example above is superficial. It lacks the complexity of interdependent things. The point yet is that data doesn’t necessarily lead to good decisions.
But let’s go back to Schmelzer.
What matters in football?
To win you have to score one more goal than your opponent.
- Johan Cruyff
With that detour, we come back to Marcel Schmelzer and the question “What is a good player?”. It might sound obvious, but a definition of good can be derived from the aim of a game itself. The purpose to play a football game is to win. And Cruyff wisely noted, that “to win a game you have to score one more goal than your opponent.”. A good player is thus one who improves the goal difference of your team systematically.
Such a top-down view is without narrative. It might be interesting to exhibit what reasons are behind a high Goalimpact and here statistical analysis of a player can reveal interesting patterns. However, some aspects of the game will be very hard to detect. That is why I refer to Schmelzer as an invisible hero. We do not know what he does good. But we know that he plays well, in terms of positive impact on the score.
(For many people not having reasons is unacceptable. Humans are by default reason seekers (even if they are false) and need to know why things are as they are.)
This is especially true in a complex game like football, where often invisible actions might produce cascading butterfly effects that positively impact your team. Indeed, soccer is complex enough that any reductionist approach must fail. And this is what we from Goalimpact truly believe. We do not value individual single-player actions. We value only the impact they have on that which matters: the goal difference.
Single-player actions and stats are at best indicative of their impact on the game. Mostly, however, they are deceiving. They do not necessarily tell me anything relevant about the game. They might be interesting for spectators (rightly so, because football is at first an entertaining business) but they will never reveal objective value.
Schmelzer is one of these players, that didn’t do anything spectacular. Quite contrary, when some technical fail occurred, you wondered how he became a professional. But remember: perception and stats are both subject to flaws and biases.
According to Goalimpact - Schmelzer is world-class.
If you ask me why, I can come up with reasons, but I have to accept that I’ll never know for sure. This fundamentally incomprehensible aspect of football is perhaps at the core of its enormous appeal.
Who knows?
You and me probably have better technique than Schmelle, so do a lot of amateur players i know. But it's all worthless because they don't have his athleticism and they don't have this technique when athletic players like him run towards them. What amateur players can't fathom is that their technique declines rapidly when someone athletic like Schmelle and other players in the highest league run towards you. You have less time, you panic more. The other thing is that doing technique things on the ball is a whole different level between doing them at 25km/h and doing them at 35km/h.
My technique is excellent vs children. They can't pressure me even in huge numbers. I'm completely "Pressing-Resistent". Now let Aubamayang pressure me with an acceleration, speed and agility that i have never experienced in real life and i will just stumble the ball.
I have seen this with Auba multiple times against team that played him for the first time, they knew he was fast, but adjusting for it on the field, stop doing things that are automatic, realizing you can't do things now in a certain range where he is, is a completely different thing. Apart from the fact that there is a boundary to adapt to someone like him.
If you think about that one big Kuba miss into the empty goal. People said i would have made it, but the difference is, you would not have been there in the first place because you don't run 36km/h. And having the technique and strength to hit a ball after a 36km/h sprint is something complete different to doing it at 25 km/h. Apart from the fact that the ball bumped shortly before he hit it from a bump in the lawn.
What Schmelle excelled in and was at a world class level was anticipating the game defensively. He was so good at anticipating lost balls and intercepting passes and players in a Gegenpressing manner. That was his main skill that made him elite and was a very good fit for BVB. He was very stiff at the Hips and wasn't a very good defender from a standstill or 1 vs 1 he had to do everything to prevent getting into these situations. Something that Hummels does nowadays to often.