Premier League Net Passing 2012-2013

In February, Dan wrote two excellent pieces explaining the net passing statistic and how the relationship between net passing and goal difference for an individual team can shine light on the importance that team places on dominating possession (we prefer using the net passing metric over possession percentage because it is more fine grained). Net passing is simply the number of passes a team completes over the course of a game less the number their opponent completes. If team B completes more passes in a game than than team A, team A's net passing for the game is negative.

For teams whose tactics are largely centered around ball retention and patient buildup play we expect a strong positive relationship between net passing and goal difference. In other words, as net passing increases for these teams we would expect to see goal difference increase positively.

For teams who prefer to play primarily on the counter, outpossessing the opponent is unimportant. Counterattacking teams want their opponent to have possession and to commit numbers forward so they can break quickly while the opposition is out of position. Counterattacks require fewer passes than slow buildup play from the back. Therefore, for primarily counterattacking teams, we expect no discernible relationship between net passing and goal difference.

Of course, many top level sides use both counterattacking and possession styles based on factors like the style of play of the opposition and whether the game is played at a club's home stadium or an away ground. For instance, we'd expect Manchester United to boss possession in a league game against Stoke at Old Trafford and have a positive net passing value (which they did last Saturday). However, in a Champions League game against Barcelona at the Nou Camp, we'd expect them to keep a compact defensive shape, allow Barca to have the bulk of possession and then look to quickly counter and therefore have a negative net passing value. For these sides, we'd expect a weaker relationship between net passing and goal difference.

Premier League Net Passing 2012-2013
The bar chart below shows the average net passing for each of the Premier League's 20 teams after eight games (Reading and Sunderland have played only seven games). Teams are listed from left to right according to their position in the league table (Chelsea currently sit atop the table while QPR are last). Manchester City, a side with very technical players capable of short intricate passes, have the highest net passing value. They are outpassing their opponents by an average of 231 passes per game. Stoke City, a team that focuses more on physical strength and territory than possession, have the lowest net passing value. They are being outpassed by an average of 226 passes per game.

Click to Enlarge

I'm also including this graph of passes completed per game for anyone interested.

Click to Enlarge

Net Passing and League Standing Relationship
While Dan looked at the relationship between net passing and points per game for individual teams, I wanted to look at the relationship between net passing and league standing for all 20 Premier League teams to determine the explanatory power of net passing on league position. If we believed that net passing was the only factor that determined whether a team won or lost a game, we'd expect the team with the highest net passing value to be in first place in the league and the team with the lowest net passing value to be in last. The bars in the net passing bar graph above would get progressively shorter as we moved right from the first place team to the last.

Clearly this is not the case. Manchester City have the highest net passing value yet they are only third in the league. QPR have a positive net passing value but are in last place. Liverpool have the fourth highest net passing value in the league but are still in the bottom half of the table while West Brom and West Ham are 6th and 7th respectively despite having substantial negative net passing values.

The graph shows what we're all well aware of- there are more factors that determine the winner of a soccer game than simply who passes the ball more. For example, in Manchester United's two defeats this season to Everton and Tottenham they outpassed their rivals by 818 passes. Arsenal completed 414 more passes than Norwich last Saturday but were beaten 1-0. Teams have to convert possession into goal scoring opportunities and then have to finish those opportunities. For a number of reasons, it often makes sense for certain teams to employ tactics that aren't focused on ball retention and allow the opposition to control the bulk of possession- it doesn't necessarily mean these teams will finish in the bottom of the league because they have a low net passing value.

The bar graph is interesting however in that it shows of the ten teams that have positive net passing values, seven of them are in the top half of the table. Of the ten with negative net passing values, seven are in the bottom half of the table. That there are more teams with positive net passing values in the top half of the league suggests there may be a relationship between net passing and league position.

To determine exactly what the explanatory power of net passing on league position is, I plotted league position versus net passing for each of the 20 Premier League teams below. Teams higher up on the y axis are in the bottom half of the league standings and teams further to the left on the x axis have higher negative net passing values. If we believe that higher net passing values improve a team's league standing, we'd expect our trend line to slope down and to the right (indicating that as net passing increases, league position gets closer to first place). Indeed, the trend line is negative. The r^2 value of 0.229 tells us that net passing explains about 23% of the variation in league standing. So although net passing clearly isn't the only factor that determines the winner of a game, it does seem to play a part in determining league position.

Click to Enlarge


The negative slope makes sense. Controlling possession and outpassing your opponent requires a team to have players that are technically gifted (good control and passing ability). Having a lot of technically gifted players also makes a team more likely to win games. Manchester United will always boss possession against a team like Stoke because their players are more technically talented and more often than not they'll beat Stoke because they have superior talent. Because of the superior talent required to play a possession game, it makes sense that top teams also generally have high net passing values.

The analysis however does not determine the subtle difference of whether top teams are top teams because they dominate passing or whether they dominate passing because they are top teams (for a team like Arsenal with a strong emphasis on ball retention regardless of the opponent my guess is the former, for a more tactically more flexible team like Manchester United I'd guess the latter).

Sample Size Issues
The significance of this analysis is limited by the small number of games played in the Premier League thus far. Eighteen teams have played only eight games and two have played only seven. Teams have also not played the same schedules as one another which will also influence net passing and league position. For example, West Ham has only played three games against teams currently in the top half of the table (and lost two) while seven of QPR's eight games have been against teams in the top half. Would QPR and West Ham's net passing and league position look different if their schedules had been swapped? More than likely they would. It would be interesting to do this analysis for the whole of last season. A project for the future perhaps.


More on net passing

Last week I discussed how we should interpret the net passing statistic and whether it's related to a team's performance. In that post, I analyzed the points per game earned by each of the big six clubs within various levels of net passing. Points per game is an important performance metric, as points ultimately determine each team's place in the table. In addition to points per game, another useful performance metric is goal difference. One advantage of goal difference is that it is more fine-grained than points per game. Whether a team wins 1-0 or 7-0 they still earn three points. Goal difference, however, allows for variation within wins and losses. The continuous nature of goal difference also lends itself to statistical analysis.

The figures below plot net passing against goal difference for each of the members of the big six (the bivariate regression is estimated using OLS). For clubs that rely on passing and possession to build goals, we would expect a positive relationship between net passing and goal difference. In other words, we would expect that as a team completes more and more passes than an opponent, the team's goal difference increases. For teams that rely more on a counter-attacking style of play, we would not expect a discernible relationship between net passing and goal difference.

(Click to enlarge)

For Manchester City and Liverpool, net passing explains less than 1 percent of the variation in goal difference (R2=0.005 and R2=0.002, respectively). Net passing explains about 5 percent of the variation in goal difference for Manchester United and Chelsea (R2=0.048 and R2=0.054, respectively). For the North London clubs, however, net passing explains considerably more of the variation in goal difference. Remarkably, about 35 percent of the variation in Tottenham's goal difference is explained by net passing  (R2=0.345), and nearly 40 percent of the variation in Arsenal's goal difference is explained by net passing  (R2=0.396). Not surprisingly, the coefficient on net passing is statistically significant only in the Arsenal and Tottenham models (p<0.01). For Arsenal, each additional 100 passes completed more than the opposing team is associated with nearly a 1 goal increase in goal difference. An increase of the same size in net passing for Tottenham is associated with a 0.6 goal increase in goal difference.

In sum, there is a strong relationship between net passing and goal difference for Arsenal and Tottenham, a weak relationship for Manchester United and Chelsea, and no apparent relationship for Manchester City and Liverpool.

Interpreting the net passing statistic

I use a statistic that I refer to as "net passing" quite often in my analysis on this blog. Net passing is simply the number of passes completed by a team net of the number of passes completed by their opponent. The purpose of such a statistic is to provide a simple description of one aspect of the game: passing. Depending on a team's style of play and tactical approach, net passing may or may not be predictive of actual match outcomes. For example, some teams depend on possession and passing to break down an opposing team's defense. On the other hand, some teams play deeper and generate scoring opportunities on the counter attack. Playing on the counter attack requires much fewer passes, and consequently, net passing is probably not very predictive of performance for these teams. The figures below show the average points per game for each of the big six clubs by the level of net passing (as of game week 23).

(Click on figures to enlarge)

The light blue bars in the figures above indicate that the points per game statistic is based on only one or two games. As a result of such a small sample, these statistics are not very reliable. Manchester City have a high points per game irrespective of their net passing. Somewhat surprisingly, they have not dropped any points in games in which their opponents have out-passed them, and they have collected the fewest points per game from games in which they completed at least 301 more passes than opposing teams. Manchester United have collected on average only 1.33 points per game from those in which they were out-passed, while they have been markedly more effective in games in which they have a positive value for net passing. Remarkably, they have out-passed opponents by a margin greater than 200 passes on only two occasions. Tottenham also have a much higher points per game when they out-pass opponents, and the pattern appears more pronounced than that of Manchester United. Chelsea have averaged a respectable 1.6 points per game in games in which they were out-passed or completed 100 or fewer passes more than opposing teams. The incredibly low 0.25 points per game for the net passing category of 101-200 serves as an important reminder that statistics can yield strange results, especially when estimated from small sample sizes. Net passing seems to be quite predictive of Arsenal's performance, which is perhaps not surprising given Arsenal's style of play. Finally, there is little variation in points per game across net passing levels for Liverpool (ignoring the >300 category, which is based on only two games).