Scouting Tomorrow's Stars with Similarity Algorithms

Our data scouting is based on a complementary approach consisting of four different methods. One of these is the similarity approach. So today we're looking at similarity algorithms. A similarity algorithm, also known as a similarity measure or similarity function, is a method used to quantify the similarity or dissimilarity between two objects. These objects can be text documents, images, user profiles or even numbers or player data profiles.

Imagine if you could identify rising stars early on. Not by chance, but based on precise data analysis. That's exactly what smart similarity algorithms make possible.They compare player profiles, evaluate performance data and playing style and identify talents that closely match the profiles of the stars. With this approach, we can spot promising young players where others rarely look – in lower leagues or in emerging markets

Pioneers Bill James & Billy Beane

In the 1970s and 1980s, Bill James was the first to look at baseball with completely new eyes. Among other things, he developed similarity scores to compare players with each other.

Instead of relying on classic statistics such as batting average or subjective scouting reports, he developed new metrics – known as sabermetrics. The goal was to better capture the true value of a player. For example, rather than just counting how many hits a player gets, he looked at how often he actually gets on base, i.e., creates opportunities for runs. His work continues to have a significant impact on how baseball teams evaluate players, make strategic decisions, and analyze games.

Billy Beane, general manager of the Oakland Athletics, took up these ideas in the early 2000s. His club was clearly at a financial disadvantage, so he had to think differently. Using James' methods and data-driven scouting, he specifically sought out players who were underrated on the transfer market – perhaps because they had unconventional batting techniques or didn't fit the classic scouting mold, but were statistically extremely valuable.

The “Moneyball” principle

• Value comes from data – not from name, reputation or label.

• Misrated players can become the backbone of a team if you spot them early.

• Winning on a budget is possible by focusing on undervalued strengths.

• The method challenged traditional scouting and changed how clubs think about recruitment.

• Moneyball inspired a shift toward data-driven decision making across professional sports.

  
  

Today, these approaches can also be found in football: clubs use data to discover hidden talent, compare player profiles more objectively or predict how well someone will adapt to a new league.

The Method

There are a variety of different similarity algorithms. The best known are Euclidean, Manhattan, Cosine, Jaccard and K-Means Clustering. We keep to ourselves exactly which metrics we apply the algorithm to for which position profiles.

In very simple terms, the overall similarity of the distance to the average for all relevant metrics is compared for the players. This allows us to examine the data sets of the individual players for overall similarity and obtain a result ranging from -1 (opposite) to 1 (identical).

We apply the following use cases:

• Starting with a top player in a top league to identify promising young talents with similar characteristics in lower leagues. Especially interesting in emerging markets.

• Starting with a young talent to find top players with similar profiles in order to better understand and classify the talent's style and development potential.

• Use team and league comparisons to predict how well a player will fit into a particular team or league.

So we can start with a benchmark player and find similar players.

As a first example, let's take Kevin de Bruyne from Manchester City in his prime during the 22-23 season. He showed outstanding values in various dimensions. When we apply the algorithm to him, we find Iliass Bel Hassani from RKC Waalwijk in the Eredivisie in the 22-23 season, with a similarity of 93.21%. This means that he performs similarly with similar strengths and style, but in a different league.

A direct comparison of the data profiles illustrates the similarity in the percentile values of the individual metrics.

Jens Petter Hauge from Bodo-Glimt had a similarity of 95.17% to Florian Wirtz in 2024. No wonder he was signed by Eintracht Frankfurt at the end of 2024. Now back to Bodo, he scored twice in the Champions League game against Tottenham on last Tuesday.

Last season, we found Matt O'Riley from Celtic Glasgow. He also has a very high similarity with Florian Wirtz. After Celtic, he played for Brighton & Hove Albion and is now on loan at Olympique Marseille.

This does not mean that the players found are as good as De Bruyne or Wirtz. However, it does mean that they have similar strengths and weaknesses and are as similar as possible in terms of style and characteristics. Both players therefore achieve similar performances in their leagues using similar means.

Good players can be found in various ways using data. And different algorithms also find different players.

Ballon d'Or 2025 Ousmane Dembélé

To demonstrate the effectiveness of our approach, we apply our similarity algorithm to the current Ballon d’Or winner, Ousmane Dembélé. We then identify the most similar U24 and U17 players in Europe. And of course, the same method works for any player profile – whether it’s Messi, Rodri, Yamal, De Bruyne or others.

Finding players with a high degree of similarity does not automatically mean that they will one day win the Ballon d'Or. Of course, the strength of the league must also be taken into account. However, it shows that with the high Ousmane Dembélé score, they have very good prerequisites for further development, improvement and success at a higher level.

What all of the selected players have in common due to their high similarity to Dembélé is that their versatility allows them to play in multiple positions, giving their team valuable flexibility and unpredictability.

We used data from the last 24-25 seasons as our database, as the data from the current season is still too limited and unreliable. Some players have since moved to better leagues.

Data Profile Ousmane Dembélé (28, Winger, PSG)

This is what a Ballon d’Or profile looks like. As expected, Dembélé's 24-25 data profile is rich, complete and dominant across most areas of the game. Explosive output with consistency and efficiency. Impactful in creation, progression and finishing.

He is known for explosive pace, outstanding dribbling, two-footed technique and creativity in the final third. His profile combines chance creation, ball progression and 1v1 dominance, making him one of the most dynamic and dangerous wide players in world football.

U23-Player most similar to Ousmane Dembélé

Ivan Brnić (Celje > Istanbul Başakşehir) Similarity 93.20%

Ivan Brnić is a 24-year-old Croatian winger (primarily on the left) who currently plays for İstanbul Başakşehir. He is known for his adaptability (he can also play right wing or centre forward), has experience in the Süper Lig, and has moved through clubs such as Hajduk Split, Maribor, Olympiacos, NK Celje (on loan), before his current transfer.

Mason Greenwood (Olympique Marseille) Similarity 92.86%

Mason Greenwood (born 1 October 2001) is an English forward who, after rising through Manchester United’s academy, moved to Olympique Marseille in 2024.

He is known for his goal instincts, ambidexterity and versatility across attacking roles.

Yuito Suzuki (Brøndby > SC Freiburg) Similarity 92.13%

Yuito Suzuki is a versatile 23-year-old Japanese attacking midfielder who has played for Shimizu S-Pulse, RC Strasbourg, and Brøndby IF, and will join SC Freiburg in summer 2025. He is known for his technical ability, creativity, and strong goal instinct.

Marius Broholm (Rosenborg > LOSC Lille) Similarity 91.85%

Marius Sivertsen Broholm (born December 26, 2004) is a Norwegian winger / attacking midfielder who joined LOSC Lille from Rosenborg in summer 2025.

He is left-footed, has previously been on loan at Kristiansund, and had a standout 2024 season with Rosenborg where he scored 8 goals in 29 league appearances.

U17-Player most similar to Ousmane Dembélé

It gets exciting when we analyze significantly younger players. Our algorithm can, of course, be applied to any player, any league and any age group. Here are three examples from the 2024/25 U17 Bundesliga season.

Salvatore Mule (VfB Stuttgart U17) Similarity 90.46%

Salvatore Mule (born 24 January 2008) is a German youth player who operates primarily as a left winger, with the ability to play as an attacking midfielder or centre-forward.

He currently plays for VfB Stuttgart U19, having joined on 1 January 2025, with a contract running until 30 June 2028.

Zain Biazid (TSG Hoffenheim U17) Similarity 89.82%

Born 14 January 2008, he is a German youth player (of Syrian origin) who plays as a left winger. He currently is in the TSG Hoffenheim U19 squad (he joined that level on 1 July 2025) and has been part of Hoffenheim’s youth system since 2019.

Lennart Karl (FC Bayern München U17) Similarity 89.45%

Lennart Karl (born 22 February 2008) is a German attacking midfielder/winger playing for Bayern Munich’s youth and senior squads. He is 1.68 m tall, has performed impressively in Bayern’s U-teams, and has already made his first senior appearance (in the FIFA Club World Cup 2025) for Bayern Munich.

U19-Player most similar to Ousmane Dembélé

Deli Hajdini (VfB Stuttgart U19) Similarity 89.64%

Deli Hajdini (born 2 October 2006) is a Kosovan left winger currently playing for VfB Stuttgart II.

He stands 1.79 m tall and has represented Kosovo at U19 level.

Jayden Osei Addai (AZ II > Como) Similarity 89.38%

Jayden Osei Addai (born 26 August 2005) is a Dutch forward/winger who joined Como 1907 in July 2025. He is 1.77 m tall, mainly plays as a right winger (but can also play left), and will be under contract in Italy until June 2030.

Some most similar player per League:

League	Player	Club	Similarity
Challenge League	Salim Ben Seghir	Xamax	85.50%
Super League	Alvin Sanches	Lausanne → Young Boys	85.50%
CH U19 Elite	Melvin Hodza	Zürich U19	87.20%
Bundesliga	Michael Olise	Bayern München	87.30%
Serie A	Kenan Yıldız	Juventus	87.90%
Austrian Bundesliga	Dorgeles Nene	Salzburg	88.60%
U19 Bundesliga	Fritz Fleck	Borussia M'gladbach U19	88.70%
CH U17 Elite	Ivan Cossalter	Winterthur U17	88.20%

Reverse Similarity Path

We can also take the opposite approach and start with a young player. For selected talents, we then look for the most similar players from the top leagues. This analysis provides valuable insights into the talent's playing style and characteristics.

Here are a few examples from the 2023-24 season.

Brajan Gruda (19, Mainz 05 U19, RWF-AMF-RW)

Matthias Seidl (22, Rapid Wien, AMF)

Monju Momuluh (21, Regionalliga Hannover 96 II,RWB)

Cross-Gender Similarity Experiment

We took the player profile of Swiss talent Sydney Schertenleib (then 17, Grasshopper, 23/24 season – now at FC Barcelona) and found a striking match in Désiré Doué (19, PSG, Ligue 1) with 88.43 % similarity between their profiles. Both combine creativity, ball progression and passing quality. A fascinating example.

Similartiy Scores at League or Team Level

Our similarity algorithm can be applied across different layers: league, team and player.

At league level, similarity scores help estimate how quickly a player might adapt to a new competition or highlight how crucial adaptability will be for success.

At team level, they can reveal how well a player’s style matches the tactical identity and playing patterns of a club.

Like all data scouting methods, this similarity approach is about player identification - means player pre-selection. Data scouting with Similarity is about smartly discovering new, interesting and sometimes even surprising players — who are then further analyzed by scouts through video and live scouting.

Smart data scouting

The approach of finding players based on similarity is just one of the good approaches that can be used for data scouting. Use our smart Moneyball scouting algorithms to search relevant markets for players who fit your specific requirements. With worldwide coverage, including youth leagues and over 20 fine-tuned position profiles, we select a data-based scouting longlist from thousands of player profiles. Customers confirm that our algorithms identify players whose strengths and style of play perfectly match the tactical orientation and requirements of their team. 👉Moneyball: Exploiting potential with smart data scouting

Our unique selling point is the combination of in-depth tactical knowledge and expertise in analytics, innovation methods, design thinking and systemic analysis. We are happy to provide you with individual support in exploring and utilizing the full potential of your data. Following article shows our philosophy, methodology and how we think football.

👉Change the game—innovation on the pitch

Tailored Football Data Solutions At footballytics, we go beyond standard stats. We create custom analyses, algorithms and models that match your football philosophy. From player profiling and data-driven scouting to tactical insights – we turn complex data into actionable decisions.

👉 Discover how data can elevate your game

Learn analytics – explore our free podcasts and videos.

Discover how data, tactics and AI come together in football. Each Learncast is short, practical, and packed with real insights. 👉 Start your learning journey now

Did you like the quality of our post? Then reward our hard work with your “credits” and like and share this post within your social network. Thank you very much.

To make sure you don't miss any posts, you can subscribe to the blog.

footballytics – improve the game - change the game

We support clubs, coaches, agencies and players with analysis and consulting services on how to use and interpret data. This enables them to make better decisions in scouting, game analysis and on the field.

---

Blog from www.footballytics.ch 

improve the game - change the ǝɯɐƃ

Share this post

follow us on social media 🔗 Bluesky 🔗Linkedin