Authors: Bret R. Myers; Brian Q. Coughlin
Corresponding Author:
Bret R. Myers
204 Eagle Glen Drive
Coatesville, PA 19320
bret.myers@villanova.edu
804-357-5876
Bret Myers is an assistant professor of management and operations at Villanova University. He also works as an analytics consultant for Toronto FC of Major League Soccer. Bret’s research and consulting is at the intersection of core sporting knowledge and the leveraging of data analysis to improve decision making for competitive advantage.
Brian Coughlin is a senior data analyst at Decision Resources Group in Exton, PA. He also serves as director of lacrosse operations at Villanova University. His passion lies in the field of analytics with a specific interest in mining data, analyzing statistics, and offering strategic recommendations that help organizations make better decisions.
On the relationship between attacking third passes and success in the English Premier League
ABSTRACT
This research examined how changes in attacking third pass behavior can impact a team’s ability to maintain leads and secure wins based on data collected from the 2011-2012 English Premier League Season. A team’s attacking third behavior is measured by the number of attacking third passes completed per minute. The results of this paper suggest that while teams tend to complete less passes in the final third when they are ahead in a match vs. being behind, there is evidence to suggest that a drop in attacking third pass behavior when ahead in a match will reduce the likelihood of maintaining a lead and securing three points.
Keywords: Soccer Strategy, Coaching Strategy, Sports Analytics, Soccer Analytics, Protecting a Lead, Staying Aggressive throughout a Match
INTRODUCTION
Advances in technology have made way for more advanced data collection in soccer. Companies like Opta and Prozone capture data relating to approximately 1,600 to 2,000 events per match (1). Teams in the English Premier League average over 400 passes attempted per match (11). An entire season of EPL matches will make way to approximately 15,000 to 16,000 passes (assuming a 38 match season). Included in this data are the location of the origin and destination of each pass. Accordingly, it is common to break down the fields into thirds (Defensive, Middle, Attacking) and the corresponding completion rates. As intuition would suggest, the attacking third tends to be the most difficult area of the field to complete passes (3). As a team advances in their attack, the level of defensive pressure increases in order to deny the creation of scoring opportunities. Therefore, analysts have used attacking third completion rates as a key performance indicator, which is supported by evidence of an association between attacking third passes and goals scored (4).
While it is significant and relevant to link attacking third pass behavior with scoring goals, this research looked to use the metric as a descriptive measure of a team’s territorial presence. There are both controllable and uncontrollable factors when it comes to pass completion. Players are in control of the decisions they make with and without the ball. A player with the ball decides where to dribble, pass, or shoot, and players without the ball decide where they should move. The uncontrollable factors include the intersection of a player’s technical ability and the level of fatigue that impact successful execution of intended events, along with the level of defensive pressure the opposition applies. While acknowledging these uncontrollable factors, it is assumed that teams at least partially control their ability to complete passes in the final third. Thus, if it is observed that a team’s attacking third pass behavior is significantly different over an interval of time, it is assumed that decision-making plays a role. A team that is exhibiting a lower than normal rate of attacking third passes is likely to be going with a more defensive strategy while a team that exhibits a higher rate of attacking third passes than normal is likely choosing a more aggressive attacking strategy.
Research in the field of soccer strategy and in team sports in general is lacking (2). While there has not been much research done in the way of soccer strategy, there has been an increase in the amount and type of data that can be collected. Much of these data have been made available through video and the advanced analysis of it.
The analyses that have been done using this advanced data have been able to provide better models of both individual and team behaviors. For example, data mining is used to develop a specific decision rule for the timing of substitutions that increases a team’s ability to come from behind in a match (7). The proposed approach has benefits across top competitions worldwide. In another example, advanced video analysis from Sky Sports split screen PlayerCam broadcasts is used to identify a clear positive relationship between exploratory behavior frequency and pass completion (4). Finally, in a study involving OPTA ball tracking data, spatiotemporal analytics is used to infer team strategy in the English Premier League (EPL) and how it can vary when playing at home vs. away (5).
Using advanced data provided by Opta’s Four Four Two Stat’s Zone App (10), we looked to find a statistic that could mirror the strategy that a team was likely to have. Attacking third passing is a relatively new statistic that has been used by coaches and analysts to show the tendencies of a team (7). Attacking third possession has also been shown to have strong correlations with winning (6). An analysis of the 2011-12 EPL season showed that there was a strong, significant correlation between the variables of rate of attacking third passes completed and the number of points that a team accumulates over the course of a season (8). This statistic has an even stronger correlation with the rate of goals for a team and can thus be a good indication of a team’s strategy.
Based on this primary and secondary research, we sought to analyze the statistic of attacking third passes further. We aimed not only to analyze it in the context of an overall season but also study how it impacted a team’s likelihood of succeeding in a given game. To elaborate, is it beneficial to maintain a certain volume of attacking throughout a game or should teams adjust their strategies when they find themselves ahead or behind in a game, given that it can be advantageous for teams to switch parts of their strategy, such as when to substitute, given the game situation? Also, do teams exhibit different passing tendencies on their home pitch compared to an away match, and more importantly, should they?
METHODOLOGY
First, we looked to examine the relationship between score state and attacking third pass behavior. Second, we looked at the converse proposition of how changes in attacking third pass behavior can influence the ability to maintain a positive score state (and hence win the match).
In order to better understand strategy involving passing, data have been collected from 380 EPL matches during the 2011-12 season using the FourFourTwo Statszone application. The variables tracked are the total number of attacking third passes attempted and completed for the home and away team segmented by whether the team is tied, ahead or behind, the total number of attacking third passes attempted and completed after the latest lead was established by the home and away teams and whether that lead was maintained, and the amount of time during which a team held a lead, was behind, and tied within a match. These statistics were analyzed on a per-game basis as well as on a per team basis over the course of the 2011-12 season.
RESULTS
Intuition suggests that teams tend to attack more when they are behind in a game as opposed to when the score is tied or when they are ahead. An initial analysis of the data from the 2011-12 EPL season indicated that teams will attempt a greater volume of attacking third passes per minute (1.05) when down as compared to when ahead (.85). This is evidence that teams are likely to change their attacking strategy depending on the situation in a game. The next statistical test compares the mean attacking third pass completed per minute according to the score state of a match (ahead, tied, or behind). A one-way ANOVA test was used and the results are displayed in Table 1 below:
The test results reveaedl a significant difference in the mean attacking third passes completed per minute depending on the scoring margin. This is evidence that a team’s willingness to attack is dependent on the score state of a match. A team’s willingness to attack increased as a team’s score state drops to tied or behind.
Ultimately, a main objective of this research was to understand how playing strategies can influence success within a match. In the data set collected, instances were isolated to where a team was ahead in a match. The team ahead earned a “success” when they end up winning the match from the time they first go ahead until the end of the match. A team is charged with a “failure” if they end up tying or losing a match after initially being ahead. In order to examine the relationship between attacking third passes and success, we aggregated the attacking third passing rates for the success vs. failure intervals of time. We also included the home/away variable to see if there was an identical effect in each case. Table 2 displays the results of the t-tests between attacking third pass behavior of success vs. failure cases for home and away.
The analysis above shows that there is a statistically significant difference between the attacking third pass behaviors in cases where a lead was successfully held versus cases where a team failed to hold a lead. The results hold for both home and away cases, although the level of statistical significance is higher for away teams versus home teams. The results suggest teams that are more successful in maintaining leads tend to complete more passes in the attacking third in the process.
To further examine this concept, we considered the average attacking third passing rates of all 20 teams during the 2011-2012 Season. We also calculated the standard deviation. We then used the following equations to determine upper and lower bounds on normal statistical behavior with respect to attacking third pass behavior:
〖UB〗_i=x ̅_i+1.5s_i (1)
〖LB〗_i=x ̅_i+1.5s_i (2)
Where x ̅_i is the average attacking third passes completed per minute of team i and si is the standard deviation of passes completed per minute of team i. The selection of 1.5 standard deviations of control is an arbitrary selection which implies that teams come at it in an attacking third passing rate within the lower and upper bound with about 0.87 probability. Table 3 displays statistics for each team as well as the lower and upper bounds.
Given the lower and upper bounds for each team, we wanted to examine how each team’s attacking third pass behavior impacted their ability to have success in holding on to leads. For each qualifying match in the 2011-2012 season, successes and failure were charted, as well as the attacking third pass behavior during the interval of time between when a team went ahead and the end of the game.
In terms of attacking third pass behavior, there are three levels. Below normal attacking third pass behavior occurs when a team exhibits an attacking third pass completion rate below the lower control limit. Normal attacking third pass behavior occurs when a team exhibits an attacking third pass completion rate within the lower and upper bounds. Finally, above normal attacking third pass behavior is the case where a team exhibits an attacking third pass completion rate above the upper control limit. Table 4 shows the impact of attacking third pass behavior and success rates in maintaining a lead.
These results show that all 20 teams experienced a drop off in success rate when falling below the lower bound in terms of their attacking third pass behavior. In events where teams went above the upper bound, 13 out of 14 teams enjoyed a 100% success rate. In fact, there was only one failed case with Liverpool in passing above the upper bound of a team’s normal behavior. Figure 1 shows the collective success rates of all teams according to the attacking third pass behavior.
Table 5 shows a numerical breakdown of successes and failures based on whether the attacking third passing behavior was above or below the lower bound. Results are also broken down into home/away.
There is strong statistical evidence that likelihood of holding on to a lead drops when teams pass less than normal in the attacking third. The results are consistent for teams playing at home or away.
CONCLUSIONS
Overall, this research exemplified how advanced statistics can be used to explain, discuss, and affect playing strategy. The high correlations of attacking third pass behavior and success were evident in the 2011-2012 season, but most importantly, teams that maintained a consistent or better rate of attacking third pass behavior in the attacking third increased their likelihood of winning the match.
One limitation of this study is that there is only a focus on attacking third passes completed. This study does not incorporate the impact of other critical attacking third statistics such as shots and goals scored. In addition, the other two thirds of the field will be important, namely the defending third. However, it is valuable information to know that attacking third pass behavior is not only influential in trying to create chances, but that it can have defensive implications as well. The “Tiki-taka” style exhibited by both Barcelona and Spain has also gotten attention for the fact that the best defense is simply to keep the ball away from the opposition. Although not incorporated into this study, intuition would suggest that both Barcelona and Spain would not only exhibit high attacking third pass completion rates, but that in games where they go ahead, their attacking third pass behavior would remain normal or even better.
Another limitation of this study is that only the 2011-2012 EPL was examined. There was considerable labor to extract the statistics from each of the 380 matches completed. We felt like this was a sufficient sample size given the goals of our study and this was confirmed by the statistical significance obtained in the various tests.
The elevator pitch of this paper to managers is that teams should not change their attacking mentality when they get a lead in a match, especially, with a large amount of time remaining. This has implications on how the manager directs the team during the game, substitutions that are made, and any types of other tactical/formation changes.
APPLICATIONS IN SPORT
The biggest beneficiaries of the information provided by this study are coaches in soccer. Accordingly, the players should be encouraged by their coaches to continue with a consistent or even more positive approach in protecting a lead, which involves continuing to advance the play into the attacking third of the opposition. Based on experience as a player, assistant coach, and general observer of the sport, it appears that there is too much conservatism that sets in when a team goes ahead in a match. The empirical evidence coming from the EPL is compelling and although other competitions were not included in the study, it is likely the similar effects could be observed in other levels of play. As the old saying goes, “if it ain’t broke, don’t fix it.” Teams get ahead of their opposition normally for a greater prowess in attacking third play; therefore, it is important to continue to maintain or even improve upon that activity.
ACKNOWLEDGMENTS
None
REFERENCES
1. Bialik, C. (2014, June 10th). The people tracking every touch, pass, and tackle in the World Cup. Retrieved from: http://fivethirtyeight.com/features/the-people-tracking-every-touch-pass-and-tackle-in-the-world-cup/.
2. Chaudhuri, O. (2012, January 26th). Premier League passing trends 2011/2012. Retrieved from: http://eplindex.com/9035/premier-league-passing-trends-2011-12.html
3. Chiappori, P.A., Levitt, S., & Groseclose, T. (2002) Testing mixed-strategy equilibria when players are heterogeneous: The case of penalty kicks in soccer. American Economic Review, 92(4), 1138-1151.
4. Jordet, G., Bloomfield, J., & Heijmerikx, J. (2013). The hidden foundation of field vision in English Premier League (EPL) soccer players. MIT Sports Analytics Conference. Retrieved from: http://www.sloansportsconference.com/wp-content/uploads/2013/02/The-hidden-foundation-of-field-vision-in-English-Premier-LeagueEPL-soccer-players.pdf
5. Lucey, P., Oliver, D., Carr, P., Roth, J., & Matthews, I. (2013). Assessing team strategy using spatiotemporal data. Disney Research. Retrieved from: http://www.disneyresearch.com/wp-content/uploads/PROJECT_AssessingTeamStrategy_kdd2013_paper.pdf
6. Medeiros, J. (2014-January 10th). The winning formula: Data analytics has become the latest tool keeping football teams one step ahead. Retrieved from: http://www.wired.co.uk/magazine/archive/2014/01/features/the-winning-formula
7. Myers, B.R. (2012). A proposed decision rule for the timing of soccer substitutions. Journal of Quantitative Analysis in Sports, 8(1), 11.
8. Passing in the final third and goals – EPL 2011-12 #MCFCAnalytics (2012, August 30th) Retrieved from: http://analysefootball.com/2012/08/30/passing-in-the-final-third-and-goals-epl-2011-12-mcfcanalytics/
9. Pleuler, D. (2013, August 7th). Central winger: What separates the pass-happy attacks of Portland Timbers and Real Salt Lake? MLS Soccer. Retrieved from: http://www.mlssoccer.com/post/2013/08/07/central-winger-what-separates-pass-happy-attacks-portland-timbers-and-real-salt-lake
10. Stats Zone (n.d.). Retrieved from: http://www.fourfourtwo.com/statszone
11. Wiebe, A. (2013, March 8th). Opta Spotlight: Statistically speaking, how MLS compares to soccer’s biggest leagues. Retrieved from: http://m.mlssoccer.com/news/article/2013/03/08/opta-spotlight-major-league-soccer-keeping-pace-rest-world
