Arsenal '09-'10: Passing

With the start of the 2010-2011 English Premier League season less than a week away, I will embark on a series of posts analyzing Arsenal's 2009-2010 campaign and see if anything can be learned from it. The Gunners finished a distant third to Manchester United and Chelsea in 2010.

All data used in this series comes from the Guardian's Premier League Chalkboards. As I was made aware of these chalkboards via a fellow Gooner in an email lamenting the high number of passes in a particular match from last season, I will dedicate my first post in this series to looking at Arsenal's passing. I will look at three statistics - number of passes, ratio of Arsenal passes to opponents' passes, and percentage of successful passes - and compare Arsenal's performance in each against their competition.

Total Number of Passes

One of the common comments (criticisms?) of Arsenal is that they pass so much more than their opponents. This can manifest itself in supporters' frustrations as it sometimes seems that the Gunners seem to enjoy passing for passing's sake and less about setting up a score. It can also be a sense of enjoyment, as such an emphasis on passing can sometimes open up opportunities not realized by other teams' strategies. The first task I set about was quantifying just how many more times Arsenal pass in a match vs. their opponents.

Looking at my transcription of the Guardian data, I took the differential in the total number of passes made by Arsenal and their opponent for each of the 38 matches in last year's Premier League season. This is noted by the variable PD. I then used Minitab's "Graphical Summary" module to look at the data's distribution. See Figure 1 for that module's output.

Figure 1: Graphical Summary of PD for Arsenal's 2009-2010 Premier League campaign

One can see that nearly all of Arsenal's matches end with them making a greater number of total passes than the opposition (note: a review of the raw data shows only two matches where this is not the case). The graphical summary also aids us in quantifying the statistical likelihood of the value of the average number of passes in excess of their competitor made by Arsenal. We can now use the one-sample t-test to quantify the range of that average as the p-value of the normality test is greater than 0.05.

Figure 2: One sample t-test of PD for Arsenal's 2009-2010 Premier League campaign

The one sample t-test shows us that Arsenal do indeed have a statistically significant mean difference in their number of passes. To be exact, the mean is 180.5 passes per match, with the 95% likelihood of the true average being 153.4 passes or greater.

One other question that could be asked is: Do Arsenal show a statistically significant gap in their increased number of passes against the competition at home vs. while on the road? Turning to my copy of the Guardian data, one can see that the data has been categorized as "home" vs. "away". Luckily, both sets of data, when grouped by venue, also test to be normal. This means a two-sample t-test can be applied to the data to determine if there is a statistically significant gap between home vs. away passing differential. See Figure 3 for the results.

Figure 3: Two sample t-test of PD for Arsenal's 2009-2010 Premier League campaign

To declare a statistically significant difference, we'd need to see a p-value of 0.05 or less. In this case, the p-value for the test statistic is 0.054, which is just outside the acceptance range. While we can't statistically declare there is a difference, it looks pretty much like there is one with Arsenal making 52 more passes vs. the opposition when at home as compared to on the road.

Ratio of Arsenal Passes to Opponent Passes

One other way to look at the passing gap is to look at the ratio of the number of Arsenal passes to those of the opposition (PR). When looking at this statistic, we get an even cleaner distribution. See Figure 4 for the graphical summary.

Figure 4: Graphical Summary of PR for Arsenal's 2009-2010 Premier League campaign

As the results indicated a normal distribution (p-value > 0.05), a one-sample t-test is used again to quantify the size of the ratio. See Figure 5 below.

Figure 5: One sample t-test of PR for Arsenal's 2009-2010 Premier League campaign

The result in Figure 5 shows that, on average, Arsenal makes 72% more passes than the opposition. It also shows the 95% likelihood of the true average being 59% or greater.

What does this look like within a match? One way to look at it is via the Guardian Chalkboards. Figure 6 shows the chalkboard for the passes made during the Arsenal/Sunderland match on February 20th, 2010.

Figure 6: Guardian Chalkboard for Arsenal/Sunderland Match (Feb 20, 2010)

This match represented the peak value for the ratio of Arsenal-to-opponent passes, a value of 2.67. The only two times Arsenal fell below 1.0 for this ratio was the August 29th, 2009 match at Manchester United and the September 26th, 2009 match at Fulham. Unfortunately, a two-sample t-test of the pass ratio data for home vs. away matches did not yield a statistically significant difference, therefore we cannot draw a conclusion of home vs. away passing ratio differences.

Passing Accuracy

A final way to look at Arsenal's passing is to look at their accuracy of passing vs. the competition. The statistic is the difference in passing accuracy (number of successful passes/total number of passes), and is expressed via the term P%D.

A graphical summary of P%D is found in Figure 7.

Figure 7:

Graphical Summary of P%D for Arsenal's 2009-2010 Premier League campaign

Figures 8 and 9 show the results of the one- and two-sample t-test that were performed on the normally distributed data set for P%D.

Figure 8:

One sample t-test of P%D for Arsenal's 2009-2010 Premier League campaign

Figure 9: Two sample t-test of P%D for Arsenal's 2009-2010 Premier League campaign

One can see that while Arsenal is 7.82% more accurate than their competition when it comes to passes, there is not a statistically significant difference in their passing accuracy at home vs. on the road.

Conclusions

During the 2009-2010 Premier League campaign, on average Arsenal:

• made 181 more passes per match than their opposition
• made 72% more passes per match than their opposition
• passing accuracy was 7.2% higher than their opposition

What did this translate to in terms of goals and wins? That's for a subsequent post!