Article Readability Stats: 3,268 words; Flesch-Kincaid Grade Level = 8.9
This is somewhat of a Part 2 in what’s turning out to be a series of posts analyzing various aspects of Tottenham’s players and their attacking data. See my first post discussing xG and xA of players over the past 5 seasons here. In this post, I’ll be looking at a few players’ shot statistics, focusing mainly on shot location and other variables like which foot they use and their respective xG.
This is a long read, so here’s a TL;DR: Son is truly two-footed, and that is probably what makes him such a deadly finisher. Dele creates/has much better chances on his left than right foot but takes few shots with his left and should use his left more. Kane loves March 5th, but August seriously hates him.
I used a python script to gather a host of shot data from understat.com, a website that compiles a ton of information on leagues, teams, and players throughout the seasons. To save space, please refer to my original post about how I gathered data. I use the same method here. The data includes variables for every shot a player has taken in the leagues that Understat covers (Premier League, LaLiga, Bundesliga, Serie A, Ligue 1, Russian Premier League). I gathered shots for Harry Kane, Son Heung-Min, Erik Lamela, Lucas Moura, Steven Bergwijn, Dele Alli, Giovanni Lo Celso, Eric Dier, and Toby Alderweireld. I gathered the main shot-takers as well as Dier and Alderweireld, 2 of our defenders who join the attack and come up for corners. I only pulled the shots for players both here last season at minimum and still here this season, so players like Eriksen and Fernando Llorente’s Hip have been excluded.
The dataset only includes Premier League matches, so no Champions league (… or EuropaL…) shots are included. Finally, after pulling the data, I added in what team the player was on for each shot, as the raw data only includes home and away teams, without specifying exactly what team the shot is for/against. While Kane’s only club covered in the sample is Tottenham, players like Lo Celso, Son, and Lucas have played for several teams. I then filtered out other teams to only include shots taken while playing for Spurs. This is partially because I’m a huge Spurs supporter and am curious about our shot statistics, and also because it’s good practice in statistics to control for any possible outside variables, like teammates on different teams that only a single player in the sample was receiving passes from.
Once I cleaned up the data, I imported it into Tableau for easy visualization and storytelling. I used JMP to perform some t-tests for analysis later on, which is basically checking to see if 2 means are statistically different from each other or if there’s not really a true difference between them.
The key variables in this analysis are the X and Y coordinates of the shots. Understat has assigned an X and Y number to every shot, which is where the player took the shot from on the pitch. Visualizing a football pitch, with the team attacking from left to right, with the attacking team’s goal line at x = 0, and the opposing team’s goal line at x = 1. The right-side touchline of the pitch is at y = 0, and the left-side touchline is at y = 1. So, the center spot is located at (0.5, 0.5), and the center of the goal you’re shooting on is at (1, 0.5). I hope that makes sense.
One limitation of the data is that the coordinates are on a square, but pitches are rectangles. This isn’t a major issue, but it means that a 0.1 change in X is not the same distance, in meters, as a 0.1 change in Y. Also important to note is the fact that almost all Premier League pitches have different dimensions, even if by just a few yards. To best reconcile this, I averaged out the distance from goal-line to goal-line (X axis) and from touchline to touchline (Y axis) of the 2019-20 grounds. I next calculated at what coordinates to add the 18-yard box and 6-yard box lines to. Since all grounds have the same dimensions for these boxes, that isn’t a major problem. In the visualizations, the 18-yard box is solid gray, and the 6-yard box is dashed gray. I limit the views to the attacking half of the pitch for easy comparison. Only a couple of Kane’s and one of Lucas’ shots are from their own half (all misses).
I validated shot locations by watching a number of Son, Kane, and Dele goals on footballia.net. I watched 5 each of their goals, noted the locations they shot from, and then located them in my sample. They were very accurate, so I have confidence in the (x,y) coordinates coding.
Please follow this link here to play around with this data yourself on Tableau Public. WordPress does not yet support embedded Java code.
The image below shows the full sample—all Spurs shots for all players I downloaded. Naturally, this view doesn’t show much at all. We can see that there are many goals right in front of the net, and they all have high xG. And we can see that the further away from the goal you get, the fewer shots there are. One thing I like about this Viz is how few shots Spurs take on the right or left side of the 18-yard box. But one of these shots is also my favorite-ever goal, which I’ll get to later on.
From here, we can filter down by various variables, such as player. First, let’s look at a few players’ shot maps. Again, head over to Tableau Public to see all of these graphics and ones I didn’t screenshot—they’re interactive too.
I don’t know about you, but I love looking at these maps. I’ve added a lot of information to each data point on Tableau, such as date; the team it was against; whether it was from a free kick, open play, corner, etc.; and the player who delivered the last pass before the shot. You can also highlight only the Result of the shot.
Of course, just looking at a shot map blindly gives little insight. We need to dig a little deeper into the data. First, I looked at average shot location, starting with the y-axis. This will show us which side of the pitch a player tends to take shots from. Kane’s average shot location on the Y-axis (left/right side of the pitch) is 0.5052—almost perfectly in the middle, so he doesn’t favor one side or the other. And his median location is exactly 0.5000, which is legitimately creepy. Since Kane’s the only striker in the group, naturally Son, Dele, and Bergwijn shoot more from the left, while Lo Celso, Lamela, and Lucas shoot more from the right. Dier and Alderweireld shoot more from the right, which makes sense since Dier has was used a bit as a right-sided midfielder and occasionally as a right back when he joined, and Toby almost exclusively plays as the right-sided CB. Toby has an average shot closer to the right touchline than Dier, which could show how Dier has played a fair bit at left-sided CB and as a central holding midfielder.
Next, I dug into each player’s shot data based on the foot they shot with. I looked at the average xG of every shot taken, the goal conversion rate (goals/total shots), and the total number of shots taken with each foot. This data, shown in the table below—coupled with shot locations—is the meat of my analysis.
The Average xG for each shot table below is where I first saw something major in the data:
Dele needs to use his left foot much more often. His average xG for all left-foot shots is 0.2644, much greater than 0.1208 average with his right. In fact, this is statistically significant at the 95% level (t-test p-value of 0.0006), which means that there’s less than a 5% chance that this big of a difference is random. Basically, it’s almost a given that Dele will have a higher xG with left-foot shots than right (on average).
Further, Dele has a 14.71% conversion rate with his left, and 16.46% with his right. This is an insignificant difference (p-value of 0.7915) from 34 left-footed shots and 237 right, so we can conclude that Dele should take more shots with his left foot. It is probable that defenders know he favors his right, so his high left-foot xG could be due to him putting defenders off by shifting onto his left, giving himself a much better look at goal (I should note that Lamela, who is notorious for only using his left foot, does not exhibit this with right-foot shots. Dele’s left-foot xG are in fact crazy). Last, it is important to note that no other player in the sample had a statistically significant difference between left- and right-foot average xG—the quality of their chances is the same on their right- or left foot. This is something unique to Dele in this sample.
Next, let’s look at Son. Son is known for having no weak foot—and his shot data backs that up. Son’s average xG on left-foot shots is 0.1148, and it’s 0.1137 for right-foot. An insignificant difference of only 0.011! He also converts 16.37% of left-foot and 17.62% of right-foot shots, again insignificant. He uses whichever foot would be the best for the situation, and his shot map backs that up. Outside the box, he prefers his left foot on the right side of the pitch, and his right foot on the left side. This is good practice (unless you’re trivela-master Quaresma or Le Cut Inside Man), so you can bend it into the net out of the keeper’s reach. Inside the box, Son tends to use his right foot on the right side, and left foot on the left side. Again, this is good practice because you can typically get a much better shot off when you’re running at the goal by using the foot of the side of net you’re on (right foot for right side of goal, and vice versa).
Since we discussed Dele’s right vs left xG and how Son chooses the best foot for the situation he’s in, let’s dive further into Dele, Son, Kane, and Lamela’s use of their feet. These are—or in Dele’s current form, were—some of Spurs’ main attackers of the past 5 seasons or so (who are still here). We know Son uses the best foot for the situation. We can further see that that is in fact the case when we add a trend line to his right- and left-foot shot maps. On his right-foot shot map (figure below), we see a “negative” trend (slope = -0.69 nice), which is what we want to see, given that outside the box you typically want to use your right foot on the left side of the pitch, and closer inside the box you want to use your right foot on the right side of the pitch. Lamela has a very similar trend (slope = -0.91). Kane has a less pronounced trend than Son and Lamela, but still a distinct shape (slope = -0.38). Dele’s line, however, is relatively flat (slope = -0.12). So, while he exhibits this trend, you can see how he’s using his right foot in many scenarios where he might be better served using his left.
Looking now at left-foot shot maps (figure below), we want to see a “positive” trend line, a mirror of the right-foot map. Son, as we would expect from the ambipedal maestro, exhibits this perfectly (slope = 1.18). Kane’s is again a solid trend (slope = 0.55), yet not as pronounced as Son’s; same with Lamela (slope = 0.29). Now look at Dele’s… It is actually the exact opposite from what we might “want” to see. The slope here (slope = -0.35) is strangely greater than his right-foot trend line. I want to walk you through my thought process regarding this data:
First, it appeared that Dele should take many of these shots with his right foot, and I figured maybe this is the reason he’s been underperforming of late. However, he is a fundamentally different player than Kane and Son, so his shots from close to the net on the right aren’t usually coming from him running at the keeper/defenders, but from being cute and technical from a stopped position in a somewhat-crowded box. From watching him the last few years, he tends to try a few movements in the box to shift his body or open his hips for a pass or shot. Finally, I realized that maybe this explains his abnormal xG with left-foot vs right-foot shots…
No matter the scenario, being relatively close to the goal on the right side of the pitch would lead a defender to expect you to either recycle possession or use your right foot to shoot/cross across the face of goal. If Dele shifts it onto his left foot to have a shot (which the defender may not expect since Dele favors his right foot), this could lead to a higher xG shot than a right-foot shot, where the defender could close the angle easily. I’ve noticed that Dele likes to open his hips for a quick pass/shot fairly often, so this would be a perfect opportunity to do so—even if he’s running at the keeper he likes to open up his hips and try to slot it far post.
I welcome any comments/discussion on this aspect of Dele’s analysis. I found it very interesting that his left-foot shot map goes against what we “should” see. For further validation, we can compare Lamela’s trend lines (since he and Dele play somewhat similar roles and are both very one-footed). As we saw above, Lamela’s trends follow Son and Kane, so Dele truly is an anomaly with his left foot. Dele if you’re reading this… take more shots with your left foot!!! Somehow your left foot shots are crazy better than your right. Get back to your best, you’ve got this! One last note I want to make about Dele is that his left foot is not necessarily weaker than his right—there’s no data to support that claim. Thus my conclusion that he should start shooting more with his left as opposed to shooting so much with his right.
Now let’s look at Kane in August (figure below). Until August 18, 2018, Kane hadn’t scored in the month. Before that day, he had taken 46 shots in August for a combined total of 4.52 xG (table below). And looking at his shot map for these 46 shots, it’s not that he’s taking bad strikes. He has several very high xG shots that were saved or hit the post. Kane truly does not perform well in August. The following table shows all players’ goals, total xG, and number of shots taken in August before Kane scored. Excluding Lucas, who only had one shot, every other player in the sample had scored. In fact, they were all outperforming their xG. August is Kane’s Bane.
Let’s move on past attackers to defenders now. Alderweireld has taken 19 shots on the right side of the pitch behind the 18-yard box; more than anywhere else except the middle of the 18-Yard Box around the penalty spot (which is expected since he comes up for corners). This shows how he pushes up from right-sided CB and has a crack from distance fairly regularly. Not a new insight, but cool to see. Dier has a similar spread of shots on both the right and left side of the pitch, but most of his shots from the right side outside the box are from 2014 to 2016, when he typically played as a right back, right-sided CB, and right-sided mid.
Lo Celso, Lucas, and Bergwijn have nothing extremely interesting to call out in detail, but here are a few things I found interesting. First, even though Lucas plays mainly on the right or as main striker, his shot map is almost symmetric. This shows how much he buzzes around the pitch, even when coming in from the right. Second, I thought it was interesting that Lo Celso has only taken two shots with his right foot so far. Both were low xG, and the keeper saved both. Those Argentinians legit have a thing against right feet. Finally, I was surprised to see how few shots Bergwijn has taken in the Prem. I must have rose-tinted glasses on, because I could’ve sworn he had taken over 11 shots. However, he has scored 3/11, a remarkable 27% goal-shot percent. That number is even better when looking at shots on target: 3 goals from 8 shots on target. I bet we see that W celebration more.
And finally, all (pseudo-) research includes future research suggestions. I would love to add the entire squad and look at more temporal view of the team—analyzing shot maps across seasons. I also want to look at the last year of Poch’s reign and the first year of Mourinho’s. Since a lot of the key players are still on the team for that span, we can look at the differences between the two managers. However, we should probably look at a random 1-year span of each manager, since managers tend to get sacked for a reason. Last, we could dive further into individual players across time, to see how they develop, and maybe even compare in-form shot maps to out-of-form ones to see if they are unlucky given the xG of their shots or maybe if they’re taking worse shots from bad positions.
In conclusion, Spurs players took shots, and I put them in Tableau to find fun stuff.
INTERESTING TIDBITS AND GOALS
I’ve added a few sheets at the end of the Tableau Viz that highlight some interesting/fun/banger goals.
Harry Kane has only ever played on March 5th twice, in 2016 and 2017. But he’s scored absolute bangers in both games. One beautiful goal against Arsenal from a crazy angle where he celebrated by sprinting faster than ever and ripping off his facemask (still brings me tears of joy… so much passion), and one brilliant show of strength against Everton, where he holds off defenders before belting the ball from deep past the keeper. The next time March 5th falls on a weekend is 2022 and again in 2023. I’ll be calling up the FA to make sure they schedule Spurs to play Saturday 3/5/2022 and Sunday 3/5/2023.
I also call out Kane’s infamous “claimed” goal vs Stoke, where he appealed to the League to have an Eriksen free kick goal be awarded to him, because his hair (yes, probably singular) touched the ball when he went up for a header. This had a high xG of 0.4587, so if it was an Eriksen goal, it would’ve had a much lower xG (based on the location of the free kick).