Blog: Monday, 19 November 2018
Dr Otto Koppius is Assistant Professor of Business Analytics at the Department of Technology & Operations Management at Rotterdam School of Management, Erasmus University (RSM). His primary interest is in sports analytics and more generally, the practical implementation of predictive analytics in organisations for data-driven decision-making. In this blog, he, RSM graduate Jesse de Bruin and thesis coach Matthijs Wolters, detail the findings of advanced data analytics methods to answer the question if football players return to their old form after a torn cruciate ligament.
Poor Ryan Thomas. Last summer, the New Zealand football player had just completed a lucrative transfer from Dutch football club PEC Zwolle to competitor PSV. But before he'd even set foot on the pitch he suffered a serious injury during a training session – a torn cruciate ligament. And he's not the only one: during the last season, Dutch international footballers Joël Veltman, Rick Karsdorp and Tessel Middag all suffered torn ligaments. This is one of the most serious injuries in football, which according to normal protocol involves six to nine months' recuperation after surgery, although recovery times of 10-12 months are sadly far from unusual.
We needed to look at more than just a few anecdotes of injuries and returns to play that have gone well or less well, so to get more – and particularly more systematic – insight into this, we delved into several football databases to collect detailed statistics about the performance of players. First, we identified 110 players in Dutch teams from the Netherlands’ Premier League, Primera Division, Bundesliga, Serie A and Ligue 1 who suffered an anterior cruciate ligament injury between 2012-2013 and 2016-2017i. For each of these players, we took the last 10 matches before the injury as our baseline, collecting data from football statistics website www.WhoScored.com about the average number of minutes played and a player's average ratingii. We then compared this data with the player’s performance in the 10 matches after return to play, after the player had officially played in a match.
The first finding comes as no real surprise: on average, players play fewer minutes per match after they return from injury. Before the injury, they played on average 74 minutes per match in the 10 matches we took as our baseline. In the 10 matches after their return, they played for an average of 63 minutes per match, i.e. 11 minutes less.
Even more interesting, the difference in average is not only because the player is carefully 'managed' during the first few matches after their return from injury, for example, players play for only 51 minutes during the first match after their return, and it takes nearly half of players more than three matches before they play a whole match. There is always a significant difference in the later games than before the injury: in the 10th match after their return, a player averages 70 minutes on the pitch, still four minutes below their average before the injury. Even though the difference becomes smaller, 10 matches after their return, the player is still substituted earlier or brought on later than before. Although we can't deduce from our data whether or not this is a result of an assessment by the coach, medical staff or the player, it does indicate that after more than two months, the player has still not returned to their previous form.
When we compare the WhoScored.com goal ratings before and after the injury, we see a similar pattern: ratings are lower after return to play. Before the injury, the average goal rating was 6.93. Ten matches after the injury, the average rating was still nearly a quarter of a point lower at 6.69. Although a quarter of a point might not seem much, we must remember that the WhoScored.com ratings are quite compressed: the majority of players score between 6 and 7.5 (exceptions above and below are rare), so a quarter of a point is quite a significant change. When we look at how the average rating develops during the ten matches after a player’s return (see Figure 1), we see that despite some progress, two months following their return, the player has still not returned to their previous form.
The performance drop of players is quite widely distributed (Fig. 2): nearly half of the players go down around half a point in their rating, a handful of players go down a point or more, but there is also a substantial percentage who are the same or even better after their return.
To find out which factors can explain these differences, we performed a regression analysisiii using various characteristics of the player and the competition as explanatory variables. It is interesting to see what factors did not have a direct influence: younger players did not regress more or less than older players. It made no difference whether they were a defender, midfielder or forward either, and there were no significant differences between the different leagues in terms of performance drop.
What does appear to have a big influence is the average performance before the injury: the better the player, the greater the performance drop. An average player has an average rating of 6.93 before injury and 6.69 after the injuryiv. A good player who scores 7.5 before injury will score on average 6.94 after the injury, i.e. a significant performance drop.
So, what about the players whose average rating is better after their injury? Are these players simply not that good? Based on the data analysed here, strictly speaking that would have to be your conclusion, but that would be too simplistic. From previous studies, we know for example that psychosocial factors like setting goals, self-confidence and therapy compliance play a very important role in successful rehabilitation, so that will undoubtedly be important, except that we simply don't have that kind of data for these footballers.
In addition, the decrease in the number of minutes played has a significant impact. The average decrease in number of minutes played after an injury (11 minutes) ensures a fall of 0.11 points in the WhoScored ratings. This is again consistent with the conclusion that most players have still not returned to their previous form more than two months after their return to play.
Most players seem to regress after their injury, regardless of their age, position or league, but what do they do differently on the pitch to cause this? In recent years there has been more detailed event data about passes, shots and duels, so we can now zoom in to what a player actually does on the pitchv.
Kinesiophobia is a well-documented phenomenon in the rehabilitation process for several types of injuries: the fear of moving or putting stress on the injured body part. In cruciate ligament injuries, this is mainly expressed in movements that require dynamic stabilisation of the knee. Translated into football terms, this occurs in the relatively uncontrolled movements in duels (e.g. a slide or landing after a head duel): do players tend to avoid duels after a cruciate ligament injury?
Unfortunately, the data is unable to provide a full conclusion: because the 'event data' for the older seasons was not recorded at the time, we can only compare before and after data for 49 players. Although there appear to be certain patterns in the event data, the data is not always accurate (it is updated by human annotators), so the random test is too small to draw hard conclusions. We are now waiting for sophisticated 3D tracking camera systems like BallJames from SciSports to get more precise data about player actions on the pitch. This will give us a better view of what and whether players really act differently on the field on their return to play after an injury.
Do footballers return to their previous form after a torn cruciate ligament? Many do not, unfortunately: more than two months after first returning to the pitch, the majority are still below their previous form. Whether this performance drop is permanent or whether it simply takes longer – up to 20 matches? Or up to 30? – is difficult to say with the current data. Nevertheless, it is important for coaches and players to realise that returning to previous form requires patience: ‘return to play’ does not mean return to performance. At the same time, there is a significant group who do return to the same form or even higher after the injury, so there's definitely hope.
So, what now for Ryan Thomas? Perhaps he can draw hope from a fellow PSV player who tore his cruciate ligament at the age of 23, but who was included in the Dutch website Wall of Fame of FC Cruciate Ligament, after successfully returning to match play after his injury and later becoming one of the best forwards in the world at Manchester United: Ruud van Nistelrooy.
[i] Players who had suffered a cruciate ligament injury in the 2017-2018 season were not included in the study. This was because they had either not yet returned to match play or because of a lack of data relating to their return. We did use data from the 2017-2018 season for the performance of players who had returned that season from a cruciate ligament injury which they had suffered in the previous season.
[ii] WhoScored.com Ratings are based on a unique, comprehensive statistical algorithm, calculated live during the game. There are over 200 raw statistics included in the calculation of a player'’s/team’'s rating, weighted according to their influence within the game. Every event of importance is taken into account, with a positive or negative effect on ratings weighted in relation to its area on the pitch and its outcome. (https://www.whoscored.com/Explanations)
[iii] We also repeated the following analyses with machine learning methods like gradient boosted regression trees and random forests. The exact numbers change slightly, depending on the method used, but the conclusions remain the same: the average performance before injury and the number of minutes played are by far the most important factors. Because the regression is the easiest to explain, we use these results in the text.
[iv] Again averaged over the 10 matches after injury
[v] With thanks to SciSports for making this data available
Rotterdam School of Management (RSM)
RSM offers Executive Education and Master programmes in various business areas for any stage of your career. For instance:
Corporate Communications & PR Manager
Science Communication and Media Officer