avatarMarc Lamberts

Free AI web copilot to create summaries, insights and extended knowledge, download it at here

2379

Abstract

I want to look at the shooting profile. That means I’m going to look at the shot location data from Haaland.</p><figure id="a968"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*cttFJMkLUwRBVWWzwih2ig.png"><figcaption></figcaption></figure><p id="1b19">What we can see is that he 123 shots in the whole season with a corresponding xG of 30,27. That means he is overperforming his xG with + 5,73. We can also see that most of his shot comes from the central areas in the penalty area.</p><p id="35c5">If we want to look for a player that maybe doesn’t have the number of goals, but does come in the same positions with the same amount of xG/shot — we can use event data to calculate similarities.</p><p id="f73e">I will take all possible shot options and situations, and use Euclidean distance to calculate the 5 most similar players in terms of shooting to Haaland.</p><figure id="077b"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*vA6IllAdLzIXUz_KOTQOCQ.png"><figcaption></figcaption></figure><p id="550a">These are the 25 most similar players in terms of shooting profile with Haaland, but the similarities are all very close to each other. So the next step would be to first calculate z-scores (to look at the standard deviations) and then calculate similarities, to get a more steady idea of the similar players.</p><figure id="890e"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*Su3SmtjlwiiDTYSYVS7xYg.png"><figcaption></figcaption></figure><p id="c717">So what’s interesting is that the names don’t change and that the similarity doesn’t change that much. Now, these players don’t all look like players that might be most similar to Haaland in his shooting profile, so let’s look at their shot maps.</p><figure id="2a5a"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*OAh0Cs9juBwtFC9DcxS47Q.png"><figcaption></figcaption></figure><p id="29b8">Diogo Jota from Liverpool has the shot map above, and what strikes me as rather similar is the high xG shots in the central areas of the penalty area. 7 goals out of 6,93 xG, meaning he is slightly overperforming.</p><figure id="9d5b"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*in1qPSi20QdjBQ35deNZNQ.png"><figcaption></figcaption></figure><p id="e801">Caicedo has 1 goal out of 1,76 xG, but what’s interesting is that he is simi

Options

lar to Haaland. Not so much in shots in the box, but because the low xG shots are from the same location, just outside the penalty area.</p><figure id="e2d1"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*4UloWOxy-drh3s5VCq4YUg.png"><figcaption></figcaption></figure><p id="6537">For Cairney from Fulham, it’s almost the same as Caicedo — only Cairney has some more shots in the central areas in the penalty area. 2 goals from 1,88 xG.</p><figure id="eb91"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*6rdU-xxd9a4j9O0XZ_DmrQ.png"><figcaption></figcaption></figure><p id="4641">Dier shots mostly from inside the penalty area and when he does that, he does in the central areas and doesn’t move too wide. In that case he is similar to Haaland.</p><figure id="29ed"><img src="https://cdn-images-1.readmedium.com/v2/resize:fit:800/1*voXWTCAFTanHJqh6-wo1Uw.png"><figcaption></figcaption></figure><p id="8888">Jorginho is similar to Caicedo, as they are similar to Haaland in their shots from distance and the xG/shot that is lower from those positions. 2 goals from 2,11 xG.</p><h1 id="cbfc">Limitations</h1><p id="a1a5">The limitations of this approach are that I compared all players in the similarity models and didn’t think to filter for a position, minutes played, or where the teams are situated on the ranking. This gives me a skewed image and it would have been better to only select attackers if I were to look for a replacement for Haaland.</p><p id="7568">Similarities are calculated on the data at hand, but the context is crucial. Data without context is useless.</p><p id="25db">However, in this case, I wanted to see what kind of player is similar to Haaland and whether Haaland is an odd attacker or not.</p><h1 id="553f">Final thoughts</h1><p id="77ea">What’s interesting to me from looking at these profiles is not so much whether the similar players are as good in front of goal as Haaland, but that Haaland shoots many shots from inside the box, with high-quality chances. The types of chances could also be encountered by defenders and midfielders in set-piece situations or players that feel more comfortable shooting from a short distance. It could very well be that Haaland and defenders have the same pattern of shot locations, but evidently, the shot quality is different. Something to think about.</p></article></body>

Shooting profiles: using event data to find similar players

I’m rather fascinated by finding similar players with data or finding players that are closest to a particular role. This is working quite well and it’s about finetuning so the eye test matches my similarity models. But, of course, this is only limited to a particular kind of data and I want to take it one step further: event data.

In one of my previous articles I spoke about similar players to Erling Haaland and that’s something I will do now too. Only this time, I’m going to focus on event data and find a shooting profile that is similar.

Method

These similarity models were calculated on Wyscout general data and look at the qualitative and quantitative data of every player, but without the context op position and aim. In this article I want to look at event data. In specific, I want to find similar player to Erling Haaland, who have a similar shooting profile.

This means I will compare the shot locations of the players and also the xG generated from those locations. The data is from Premier League 2022–2023 and the data comes from Opta. All of the data is collected on July 18th, 2023. The expected goals numbers are generated with my own model and is considered slightly more generous than Opta, but not to the extent of Wyscout’s model.

Shotmap Erling Haaland

So, I want to look at the shooting profile. That means I’m going to look at the shot location data from Haaland.

What we can see is that he 123 shots in the whole season with a corresponding xG of 30,27. That means he is overperforming his xG with + 5,73. We can also see that most of his shot comes from the central areas in the penalty area.

If we want to look for a player that maybe doesn’t have the number of goals, but does come in the same positions with the same amount of xG/shot — we can use event data to calculate similarities.

I will take all possible shot options and situations, and use Euclidean distance to calculate the 5 most similar players in terms of shooting to Haaland.

These are the 25 most similar players in terms of shooting profile with Haaland, but the similarities are all very close to each other. So the next step would be to first calculate z-scores (to look at the standard deviations) and then calculate similarities, to get a more steady idea of the similar players.

So what’s interesting is that the names don’t change and that the similarity doesn’t change that much. Now, these players don’t all look like players that might be most similar to Haaland in his shooting profile, so let’s look at their shot maps.

Diogo Jota from Liverpool has the shot map above, and what strikes me as rather similar is the high xG shots in the central areas of the penalty area. 7 goals out of 6,93 xG, meaning he is slightly overperforming.

Caicedo has 1 goal out of 1,76 xG, but what’s interesting is that he is similar to Haaland. Not so much in shots in the box, but because the low xG shots are from the same location, just outside the penalty area.

For Cairney from Fulham, it’s almost the same as Caicedo — only Cairney has some more shots in the central areas in the penalty area. 2 goals from 1,88 xG.

Dier shots mostly from inside the penalty area and when he does that, he does in the central areas and doesn’t move too wide. In that case he is similar to Haaland.

Jorginho is similar to Caicedo, as they are similar to Haaland in their shots from distance and the xG/shot that is lower from those positions. 2 goals from 2,11 xG.

Limitations

The limitations of this approach are that I compared all players in the similarity models and didn’t think to filter for a position, minutes played, or where the teams are situated on the ranking. This gives me a skewed image and it would have been better to only select attackers if I were to look for a replacement for Haaland.

Similarities are calculated on the data at hand, but the context is crucial. Data without context is useless.

However, in this case, I wanted to see what kind of player is similar to Haaland and whether Haaland is an odd attacker or not.

Final thoughts

What’s interesting to me from looking at these profiles is not so much whether the similar players are as good in front of goal as Haaland, but that Haaland shoots many shots from inside the box, with high-quality chances. The types of chances could also be encountered by defenders and midfielders in set-piece situations or players that feel more comfortable shooting from a short distance. It could very well be that Haaland and defenders have the same pattern of shot locations, but evidently, the shot quality is different. Something to think about.

Data Analysis
Soccer
Football
Scouting
Premier League
Recommended from ReadMedium