The role of analytics in football analysis

in Other

analytics word cloud

Football has been frenetically overwhelmed in the last few years by the need to gather more and more data about games and coaching sessions. From the software (and hardware) that analyse the number of passes, distances covered or average speeds to the behaviours displayed by a coach when delivering a coaching session or in game-context, the sports market is literally flooded with devices and programs that will gather this information within a blink of an eye. With this information in hand, the role of analytics within the football analysis context is therefore to identify patterns within the data and to predict future outcomes. In my perspective, this line of thought may be flawed because numbers are very useful for identifying patterns in football, but their use for explaining and analysing a football game is limited. Let´s look at some basic examples of these limitations:

1- Mascherano holds a better passing percentage than Piqué

Looking at the WhoScored statistics available online, Mascherano roughly played the same time in La Liga as Piqué (857min vs 855min), but holds a considerable difference in average of passes per game (73.3 passes vs 63 passes) and success rate of these passes (91.1% vs 89.7%). Obviously I am interested in analysing the last two items. These numbers show that a player is able to complete more passes with more success than another. We all agree on this. However do these numbers explain why the passes made by Mascherano are more effective than Piqué’s? Probably not. And why is that? In my opinion, this is because one can analyse the number of passes and the success rate, but one cannot  determine whether each pass was the best decision possible.

After years of making descriptive analysis of Barcelona games, I am able to conclude that Piqué risks the vertical pass more often than Mascherano, who most of the time plays the safest pass (backwards or sideways).

2- They should have bought Darren Bent

This is the title of the third chapter of a book called The numbers game: Why everything you know about football is wrong, by Chris Anderson and David Sally. I am usually highly skeptical of books with this kind of title, as they give the impression that one is about to witness the power of absolute truth once one understands what the authors are trying to prove. In this chapter the authors make an obvious point from statistical analysis: not every goal has the same value, as some goals when scored secure points and others don´t (i.e. 0-0 to 1-0 = 3 points; 1-0 to 1-1 = 1 point; 5-0 to 5-1 = 0 points).

Starting from this premise, the authors tried to prove that Darren Bent would be the best option available (for Liverpool to replace Torres, instead of Andy Carroll), taking into account his record of marginal points, or in simple terms the number of points that his team achieved as a direct result of the goals scored by Darren Bent. In fact Darren Bent throughout the 2009/2010 and 2010/2011 campaigns ranked second in marginal points table behind Rooney (2009/2010) and Tevez (2010/2011), who respectively had 20.02 points (scored 24 goals) and 15.01 points (scored 17 goals). During these two seasons, Bent played for two different clubs, Sunderland and Aston Villa.

In my perspective this argument based on the use of marginal points as a predictor of whether a new signing is best for a squad is fallacious. In first place it could be argued that the reason why Darren Bent scored so many marginal points is due to the fact that he played for mid/low-table teams, with a typical British tactical approach. Secondly and perhaps more importantly, is the consideration that the analytical method loses sight of the importance of the game model of each team. In simple terms, what is the point of buying a player who is unable to understand or does not have the characteristics to play in a specific way?

Two good examples of this can be presently seen in Spain: Diego Costa playing for the Spanish National Team and Suárez playing for FC Barcelona. Both players are unable to fit in a game model based on possession, mobility, playing between lines, use of all corridors, simple combinations, etc, simply because that model requires players who have the ability to pass and read the game quickly and especially are able to take the best decision possible.

Without numbers, I daresay that most of the times for Costa and Suárez the best decision is always to fight hard against the defenders, run like there is no tomorrow, fight for every ball as if the world is about to collapse and try to score no matter what. Also without numbers, I daresay that last season Suárez must have contributed with plenty of marginal points to Liverpool´s campaign.

However this season at FC Barcelona, it is clear that Suárez has nothing in common with his teammates in terms of how they perceive the game or in terms of decision-making skills. Therefore going back to the main question here, my opinion is that neither Carroll or Bent were the best options to replace Torres, but someone who would have similar characteristics to Torres (i.e. Lewandowski, Benzema or Jackson Martinez).

In closing, I think that it is quite clear that the analytics as a method per se is quite limited, as demonstrated with these two simple examples, but this is not to say that it is completely useless. I maintain that when used as a complement of descriptive analysis it could be a strong tool to enhance the Scouting Division of any club. Nonetheless whoever is responsible for the analysis should have a deep knowledge of the game, especially to identify the process of decision making, instead of just looking for patterns of play or statistics that only show a superficial perspective of the game.

by Pedro Castro

pedro castro bnw