Introducing PassSonars for Scottish Football

Thanks to our data deal with ORTEC Sports we can now bring you an exciting array of new data analyses and visualizations for the Scottish Premiership. Some of this new stuff has already been presented in recent articles on the site (for example this by Jamie Kilday and this by Matt Rhein). And it’s good to see some Premiership players taking notice.

In this article, I will introduce a new way of visualizing player passing patterns called PassSonar. This chart was first introduced by Eliot McKinley of the excellent website American Soccer Analysis. Eliot was kind enough to share the R code behind his PassSonar visualization, and I adapted his code for use with ORTEC’s event data, which provides the distance and angle (relative to opposition goal) of every pass attempt.

A PassSonar shows the frequency distribution of different passing angles for a given player, as well as the typical length of pass attempts. In other words, it shows what direction a player’s pass attempts tend to go in, and whether his passes tend to be relatively short or long.

So, without further ado, here is what PassSonars look like.

Figure 1. PassSonars by position.

Figure 1. PassSonars by position.

As you can see in the figure above, I’ve created PassSonars for different positions to show variation in how the chart looks depending on what the player’s role is on the pitch. These data are for Rounds 1-33 of the current season of the Scottish Premiership.

One way to think about a PassSonar is as an ordered bar chart that has been folded in on itself to create a circle. The height of each bar represents the relative frequency of pass attempts within a given range of angles. There are 24, 15-degree angle intervals plotted in ascending order (i.e., 0-15, 16-30, etc.) in a clockwise fashion.

So for example, goalkeepers necessarily emphasize forward passing. Left backs tend to pass to the right. Right backs tend to pass to the left. And center fowards have the most diverse range of passing angles.

The color of each bar represents the median length of the pass attempt in that interval. So, as you would expect, goalkeepers and defenders tend to make long foward passes. Whereas central midfielders and center forwards tend to focus on shorter passes.

One question you might be asking yourself is why not have the height of the bar represent median pass length? Wouldn’t that be more intuitive? It’s a fair point, but see this Twitter thread for a discussion of the rationale for not doing it this way. It might take a little getting used to, but trust us, it’s more insightful to have bar height represent the frequency of pass attempts in a given angle range rather than median pass length.

Anyway, you get the picture (literally). The PassSonars in Figure 1 show what we already know to be true about typical passing patterns based on a player’s role.

PassSonars really start to get interesting when you compare teams and players to investigate differences in playing style.

For example, looking at defensive PassSonars can tell you a lot about whether a team tends to play out of the back or opt for the “route one” approach. To see what I mean, compare the two sets of PassSonars for Celtic and Livingston below. 

Figure 2. Celtic defense PassSonars.

Figure 2. Celtic defense PassSonars.

Figure 3. Livingston defense PassSonars.

Figure 3. Livingston defense PassSonars.

Compared with Livingston’s Liam Kelly, Celtic keeper Scott Bain attempts shorter passes more frequently, and uses a wider range of angles. Similarly, unlike Livi’s two main defenders who typically attempt long forward passes, Celtic’s defenders tend to attempt shorter, lateral passes indicative of their role in “recycling” possession as they play out of the back.

One limitation of PassSonars is that they do not give you any direct information on pass locations. This is generally fine for comparing central defenders. But when comparing players in roles where positioning is more flexible, it would help to be able to visualize typical passing locations along with a PassSonar. Again, thanks to ORTEC’s event data, which includes xy coordinates of each pass attempt, we can pair a PassSonar with a pass location heatmap to get a more complete picture of a player’s passing profile.

For example, the figures below show PassSonars and heatmaps for Celtic’s two main right backs this season, Mikael Lustig and Jeremy Toljan (I’ve used Ben Torvaney’s ggsoccer R package to draw the pitch).

Figure 4. Mikael Lustig (Celtic) PassSonar and pass location heatmap.

Figure 4. Mikael Lustig (Celtic) PassSonar and pass location heatmap.

Figure 5. Jeremy Toljan (Celtic) PassSonar and pass location heatmap.

Figure 5. Jeremy Toljan (Celtic) PassSonar and pass location heatmap.

Looking at the pass location heatmaps, we can see that Lustig does plenty of passing from his own half and the opposition half. In contrast, Toljan tends to do most of his passing in the opposition half, closer to or within the final third. Futhermore, the PassSonars show that Toljan tends to make use of a wider range of passing angles, and his pass attempts tend to be shorter than Lustig’s overall. All of this provides visual proof of Toljan’s more attacking style compared to Lustig (or Lustig’s more disciplined defending).

So hopefully, if you’ve made it this far, you’ll agree that PassSonars are not just eye candy for football stat nerds. These advanced data visualizations can provide insights about how teams and individual players operate on the pitch. This type of information can be useful for performance analysis, as well as opposition scouting. Look forward to seeing plenty more PassSonars as the season winds down and even more so in 2019-20.

Thanks for reading!