Site icon R-bloggers

Visualizing the Premier League So Far, Part 2: Stats from Open Play and Set Pieces

[This article was first published on R by R(yo), and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

In Part 2, I will be looking into various statistics from different game situations in the English Premier League, specifically from open play, set pieces, and corner kicks. As with stats taken from “special” situations, do keep in mind that some teams may not have taken many shots from these situations. When talking about the numbers below I’ll make sure to qualify any statements with the total amount of shots taken or any other circumstance that I know of. As in Part 1 all the data comes from understat.com. Although I am looking a bit deeper in regards to stats from specific situations this is still more of a broad overview of different teams. I am hoping that writing this mini-series, along with all the data cleaning and viz work that I’ve already done, will be a springboard for more specific topics in the future.

Just a reminder that I am only using the data up to Matchday 12 (Nov. 10th).

You can follow along with the code: here

Let’s get started!

Open Play

As you might expect the team far-and-away at the top in terms of xG (30.77) and actual goals (27) scored is Manchester City (they lead the second best team, Chelsea, by nearly 8 xG). In terms of shot quantity at either end City are at the top with most shots taken and least conceded from open play (198 For, 64 against). However, in terms of shot quality things are drastically different as although City’s xG per shot is still the best with 0.155, their 0.178 xGA per shot from open play is the worst in the entire league! As you’ll see later on, their defensive stats from other situations are generally good to OK so the problem really comes from the fact that in open play, City become terribly exposed once teams are able to get past their intense press. You can check out a deeper analysis on this topic (from early October) from Grace Robertson on StatsBomb.

What might surprise a few is that the team in the 85th percentile for xG and 80th percentile for xG per shot is Brighton! The Seagulls (highlighted in white) are only behind City, Chelsea, and Liverpool for these stats, yet the difference is that their goals scored is only in the 50th percentile with only 10 open play goals from 15.07 xG (their games against West Ham and Burnley come to mind for missed opportunities). If they can start converting their chances they could be much higher than their current 12th place position although they might still need to improve a bit on their league-average defense. They have also been lucky defending set pieces as they have the league-worst for xGA per shot (0.45 from 30 shots) from these situations but have only given up 2 goals from 5.06 total xGA.

In Unai Emery’s second full season with Arsenal, the Gunners are hovering around the 50th percentile for all the offensive metrics shown which is clearly not good enough for a team aspiring to get back into the Champions League places. They are not shooting well enough (both in quantity and quality) and they’re letting teams take a lot of shots on their goal (fourth worst in the league with 143). On the other hand, their xGA per shot is the best in the league at 0.08.

A match that provided a great example for these stats was Arsenal’s 2-2 draw against Watford back in September where they allowed a jaw-dropping 31 shots (although only 10 were on target). Even with Watford’s poor shot quality their expected points (xPts) from this game was 2.56. Meanwhile, Arsenal still managed to do get 7 shots off (4 on target) but it was quite astonishing watching that second half and seeing Arsenal escape with a point (especially when simulations had them good for only 0.31 xPts)!

Set Pieces (Free Kicks and Corners excluding Direct FKs)

There isn’t a very large sample of shots from just free kicks so for this section I will combine free kicks and corners (while still excluding direct free kicks). It would have been very informative if the total number of set piece attempts and coordinates for where these set pieces took place were recorded but unfortunately understat doesn’t have that info on hand. Maybe in the future I’ll try to combine it with attempts data from another source.

Liverpool have only given up 24 (of which only 5 are from free kicks) which is tied fourth best alongside Manchester United. In general, Liverpool do well on both ends of the pitch concerning corners and set pieces although they don’t seem to create many high quality chances as they’re just about league average (median) for expected goals, 2.65 xG from 39 shots. Results-wise they are tied first with Bournemouth with 5 goals scored from these situations.

Chelsea are one of the worst teams defending set pieces with all of their defensive metrics except shots conceded being below the 25th percentile. Looking at the raw numbers they’ve conceded 2 goals (both against Liverpool back in September) from 2.31 xGA, from a total of 13 shots. They have also been conceding from set pieces in the Champions League (not shown in the data) as well which has prompted Frank Lampard to switch from a pure zonal marking system to a mixed system from October.

Some special notice must be given to Bournemouth (who I’ve highlighted in purple) who appear in the top three of every offensive metric shown here! They are tied with Liverpool with 5 goals from 5.75 xG and posting an impressive 0.33 xG per shot from these situations, only Manchester United are better with 0.39 xG per shot (these numbers come from 42 and 35 total shots respectively). The Cherries are heavily reliant on set pieces as they are below the 25th percentile for shots, goals, and xG from open play. They lead the league in proportion of total xG from set pieces (corners and free kicks, excluding direct free kicks) with 35.1%. I’ll go into a specific example of one of their attacking routines that have proved quite effective in the next section.

However, this prowess in attack doesn’t translate on the defensive side of things where Bournemouth are around the 50th percentile (shots against: 33, xGA: 3.255) and downright bad in terms of xGA per shot (0.35 xGA per shot) and goals conceded (4). Even with these stats only 17.5% of their total xGA come from their inability to defend set pieces, 8th best in the league!

Corner Kicks

Manchester City have gotten 45 shots off from corner kicks followed closely by Everton with 41. Everton lead the league in total xG from corners with 4.3 while City are close by in 5th (3.29). However, the similarities stop there as Everton have conceded the most goals from corners (tied at 4 with Aston Villa and West Ham) from a total xGA of 2.68 while City have conceded 2 from 0.5 xGA. This difference mainly comes from the fact that Everton have the worst xGA per Shot from corners in the league at 0.244 while City have the fourth best with 0.05 xGA per shot. Everton’s stat comes from from 11 shots, second best in the league just behind Manchester City (10 shots allowed from corners). These stats might suggest that although Everton are good at preventing shots from corners, when opponents do breach the defense, the resulting shot is usually of higher quality.

Spurs haven’t created a lot of chances from corner kicks (in quality and quantity) yet they’ve still managed to score 2 goals, one of them being Lucas Moura’s header (valued at just 0.04 xG) in that controversial game against City back in August! On the other hand they’ve been very good at preventing quality chances from corners having conceded 0 goals so far from just 0.97 xGA (tied with Liverpool and Leicester).

Chelsea has the worst record in xGA, not just among their fellow “top” teams but the second worst in the league with 0.177 xGA per shot. On the flip side, Leicester is probably the best team defending corner kicks with none of their stats below the 75th percentile, they lead the league in total xGA (0.38 from just 12 shots allowed) and xGA per shot (0.032).

In terms of xG per shot from corners, Burnley lead with 0.15 followed by West Ham and Sheffield United. Burnley have scored 4 goals from a total of 4.27 xG (so just around what understats’ model expected) but for all of Everton’s efforts, the aforementioned 41 shots and with 4.3 xG in total, they’ve only managed to score 2 from corner kicks. Meanwhile, Crystal Palace and Watford are the worst teams from corners as they are the bottom two teams in every metric besides shots.

Arsenal are joint top (with Burnley and Bournemouth) with 4 goals from corners but their other metrics don’t reflect this good haul as the Gunners have created chances amounting to only 2.133 xG from 26 shots (Burnley: 4.27 xG, Bournemouth: 3.28 xG). The Gunners have only accumulated 17.7% of their non-penalty xG from corner kicks while in comparison Burnley and Bournemouth have 24.7% and 20% respectively.

Bournemouth showed up as one of the top teams in the previous section and they seem to put a lot of hours on the training ground on various maneuvers. A good example would be Harry Wilson’s goal vs. Newcastle United, check out their routine here:

Conclusion

In a game of fine margins having that extra edge in set pieces or corners can really be the difference between relegation vs. safety, Europe vs. mid-table, or even the title. The employment of specialist coaches/consultants in this area in recent years, like Stuart Reid, while the fact that clubs like Brentford and FC Midtjylland also employ their own specialists (or even for throw-ins, see Thomas Gronnemark for Liverpool!) are in part due to the growing influence of analytics in football focusing on this aspect of the game.

Below are some articles regarding set-pieces if you’re interested:

This was a very broad strokes overview on where and how much teams are creating and conceding chances. Still, just looking at the xG data from understat can give you some leads on what teams you might want to look at more carefully and for that I’ll probably need to include data from other sources as well. That will be a challenge to tackle in a future part. In the more immediately future, the next part will look at how stats look for teams across every 15-minute intervals of games.

To leave a comment for the author, please follow the link and comment on their blog: R by R(yo).

R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.