<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Eitan Zavorin on Medium]]></title>
        <description><![CDATA[Stories by Eitan Zavorin on Medium]]></description>
        <link>https://medium.com/@eitan.zavorin?source=rss-79350568bfe9------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*hwGNZLnMfoWzCwIM</url>
            <title>Stories by Eitan Zavorin on Medium</title>
            <link>https://medium.com/@eitan.zavorin?source=rss-79350568bfe9------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 04:28:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@eitan.zavorin/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Discovering the Doppelgängers of the World’s Greatest Soccer Players: An Analysis of Play Style and…]]></title>
            <link>https://medium.com/@eitan.zavorin/discovering-the-doppelg%C3%A4ngers-of-the-worlds-greatest-soccer-players-an-analysis-of-play-style-and-6a15e69db2c3?source=rss-79350568bfe9------2</link>
            <guid isPermaLink="false">https://medium.com/p/6a15e69db2c3</guid>
            <category><![CDATA[inst414spr24final]]></category>
            <dc:creator><![CDATA[Eitan Zavorin]]></dc:creator>
            <pubDate>Thu, 16 May 2024 03:32:01 GMT</pubDate>
            <atom:updated>2024-05-16T03:32:01.465Z</atom:updated>
            <content:encoded><![CDATA[<h3>Discovering the Doppelgängers of the World’s Greatest Soccer Players: An Analysis of Play Style and Recent Form</h3><p>In the rapidly evolving world of soccer, identifying player similarities is a crucial task that can provide valuable insights for various stakeholders, such as coaches, talent scouts, club executives, and fans. By carefully measuring similarities between soccer players based on their play styles, skill attributes, and recent form, we can answer very important and controversial questions about player comparisons and potential talent recruitment strategies. In this report, I will delve into the process of creating a model that pinpoints the most similar soccer players to three selected individuals. Utilizing cosine similarity calculations on a vast dataset of players and their attributes and then combining that with a separate clustering model on the players’ recent form, I aim to uncover hidden patterns and similarities that inform decision-making processes in the world of soccer. Knowing that the most controversial question among soccer fans around the world is whether Lionel Messi or Cristiano Ronaldo is the best player in the world, I’ve chosen the following research question: With big names like Messi, Ronaldo, and Wayne Rooney stealing the spotlight over the recent decades of soccer, who are the most comparable players to them that deserve more attention?</p><p>Before diving into my investigation, it’s crucial to understand my stakeholders so that I can bring the most value to them in creating my models and the discoveries I make along the way. One primary category of stakeholders invested in knowing player similarities is team recruitment staff and talent scouts. These professionals seek to proactively identify potential talent prospects for transfer or recruitment to improve team performance. Scouts can make educated decisions regarding transfer targets by accurately assessing player similarities, ensuring that new recruits complement pre-existing team dynamics. Additionally, coaches and team managers could benefit from insights into player similarities, utilizing them to plan training sessions, tactics, and formations based on player profiles and play styles. Lastly, fans will always be a stakeholder in any soccer-related investigation such as this one. Fans in soccer communities around the world could benefit from deepening their understanding of player performance and the game, enhancing their viewing experience, fostering engaging discussions about the sport, and even inspiring them to learn from the greats.</p><p>To answer our question, I would need an accurate, detailed collection of data on soccer players around the world, including elaborate data on the players’ play styles and skill attributes, as well as their recent form. Limiting ourselves to only skill attributes or recent form would not account for a huge part of what makes a transfer target valuable. It is only by combining these two realms of the game, that I can equip ourselves with the data I need to make insightful discoveries and provide the same for our various stakeholders. Fortunately, by leveraging two comprehensive datasets, I can get the best of both worlds and exactly what I need to begin my investigation.</p><p>The datasets used in this analysis comprise of the Player and Player_Attributes tables from Kaggle’s “European Soccer Database”, as well as a transfer market table from Kaggle’s “Soccer players values and their statistics” database. The dataset from the first database — which we will refer to as the Player Attributes table — contains detailed attribute-related information on soccer players, such as finishing, dribbling, ball control, sprint speed, passing, and more. Each record in the dataset represents a soccer player and includes various quantitative metrics that quantify their play styles and skill levels. To complement the first dataset, the table from the second database — which we will refer to as the Player Form table — includes different kinds of data on the same soccer players, including a player’s games played, goals, assists, pass completion, fouls, blocks, and over two hundred more columns on player stats from a chosen season. To ensure near-perfect homogeneity between the two datasets with respect to time, the Player Attributes table and the Player Form table are taken from seasons that are just two years apart. Using these advanced metrics on player attributes and form allows us to measure similarities between players based on multiple dimensions.</p><p>To collect the first dataset, I transformed the publicly available SQLite format of the Player Attributes table available on Kaggle into a CSV file. Using the pandas library, I extracted the data and transformed it to create a structured dataset suitable for analysis. I cleaned the data by removing duplicates, removing rows with significant missing values, and selecting only the relevant quantitative attributes for similarity measurement. Eventually, my final product of the Player Attributes table contained about 600 player rows, including only the relevant player attributes necessary to proceed with my investigation.</p><p>For my second dataset, I used similar steps for creating my Player Form table. After reading in the CSV file, I cleaned and organized the dataframe by removing irrelevant columns, removing duplicates, and selecting only quantitative attributes for k-means clustering. Eventually, I was left with a cleaned, organized, numerical dataframe of almost 2000 player rows, including almost 300 rows related to performance stats. With both of my datasets now cleaned up, I was ready to move forward with my analysis.</p><p><strong>Analysis, Part 1: Analyzing Player Attribute Similarities Using Cosine Similarity</strong></p><p>In the first half of my analysis, I created a model that finds the 10 most similar players to a target player based on player attributes. Considering all the concepts we learned in our course, I feel that the most appropriate way to compare players’ skill attributes would be through calculating cosine similarity. This would allow me to find players who might not be quite the same as the target players, but would still show very similar play styles. To do this, I measure the similarities between soccer players using cosine distance calculations, which also tell us the cosine similarity. Each player is represented as a vector of attributes, including finishing, dribbling, ball control, etc., and the cosine distance is computed between pairs of player vectors to quantify their similarities.</p><p>For three selected players — Lionel Messi, Cristiano Ronaldo, and Wayne Rooney — I identify the top 10 most similar players in the dataset based on cosine distance scores on all their relevant play style attributes. By discovering players who exhibit similar skill profiles and playing styles, we gain insights into some very influential but possibly less appreciated players in the game. Here are my findings:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*aXfwaBhW6Y1qnRl0.png" /></figure><p>The analysis revealed interesting patterns of player similarities across various dimensions. For example, Ronaldo, known for his exceptional finishing and long-shot abilities, demonstrates high similarity scores with players who excel in similar attributes. Similarly, Messi, renowned for his outstanding dribbling, ball control, passing, and vision, shares similarities with players who exhibit similar playmaking abilities. Thirdly, Wayne Rooney, whose strength, aggression, and shooting traits stand out amongst the competition, pair him closely with many players known for sharing the same characteristics. These findings partially answer our research question, as well as numerous valuable insights for talent scouts and team managers seeking to identify players who align with specific skill requirements that only some of the best players are known to have.</p><p><strong>Analysis, Part 2: Analyzing Player Form Similarities Using K-Means Clustering</strong></p><p>For the second half of my analysis, I created a k-means clustering model that would find the 10 most similar players to the same three target players, but this time using players’ recent form rather than play style attributes. I chose clustering because, just like calculating cosine similarities, this was another interesting and accurate way to find players who are similar to each other. The best part about the approach and the new dataset, is that I would be finding a mostly different set of players similar to our targets. This is not to say that the two sets of top 10 similar players to each target player would be wrong; it means that I’d be providing an even larger pool of similar players for stakeholders to look into for more informed decision-making regarding transfer market targets, team contracts, etc. To do this, I run my k-means clustering on the vast set of players, setting Wayne Rooney, Cristiano Ronaldo, and Lionel Messi as my preset cluster centers. This would categorize the rest of the players based on who they’re most similar to, giving me the following clusters:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/352/0*RwILD7Ul8Lom8RF1" /></figure><p>These gave me three very large clusters. The first, Cluster 0, outputs 623 players who were more similar to Wayne Rooney than the other two. The second, Cluster 1, outputs a whopping 1023 players who were most similar to Cristiano Ronaldo. Thirdly, Cluster 2 outputs 334 players who were most similar to Lionel Messi. Now, it’s obvious that if a stakeholder asked for players similar to one of these three, and were given a list of 1000, or even 300 players, it would be a pretty meaningless list. So, I did a bit of research, and found a way to locate the 10 closest points to each cluster center, giving me the top 10 most similar players to each target based on the clustering algorithm on player performance stats. Here’s an example of the final product:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/275/0*kz46vIOlJC4Srm8u" /></figure><p>Here, you can see the top 10 most similar players to Wayne Rooney based on his performance stats. If you compare this to the top 10 list we got from the cosine similarity analysis, you’ll see that the two lists are entirely different. However, I would argue that most well-versed soccer fans would agree that both lists are equally valid comparisons.</p><p>Despite the insights gained from my investigation, it’s crucial to acknowledge the limitations and biases inherent in my approach. One limitation is the complete dependence on quantitative attributes, which may not capture the full spectrum of player abilities and characteristics. Some qualitative measures, such as attacking and defensive work rates (assigned as low, medium, or high), are important to take into account when measuring the player profiles and players. Additionally, the model’s calculations may be biased towards players with skill profiles or playing styles prevalent in the dataset, possibly overlooking players with unique sets of attributes or unconventional play styles. Thirdly, the first dataset was last updated eight years ago, so it’s missing updates on many player profiles and completely misses any players that emerged onto the scene in the last eight years. It’s also important to mention that the second dataset was not only taken from a different source (both Kaggle, but from different parent databases). This could create limitations in the homogeneity of the data, especially considering there is a two-year gap between the update times of the two datasets I used. Sadly, there was nothing I could’ve done about this two-year gap between the datasets, but fortunately, in our context, two years doesn’t change much in soccer. At the end of the day, this model would work on up-to-date datasets, and the insights that our stakeholders can gather from it are still very much valid and valuable.</p><p>In conclusion, my analysis exhibits the power of data-driven approaches in uncovering player similarities and informing decision-making processes in various aspects of the world of soccer. By utilizing cosine distance calculations on a vast dataset of player attributes, and combining that with k-means clustering on a separate dataset of recent player performance, we can gain valuable insights into player comparisons and talent recruitment strategies. Moving forward, additional research and refinement of our model could improve its accuracy, relevance, and applicability, ultimately empowering stakeholders to make more educated decisions in the dynamic world of soccer.</p><p><a href="https://github.com/eitanzav/INST414FinalProject">GitHub - eitanzav/INST414FinalProject</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6a15e69db2c3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Predicting a Soccer Player’s Overall Skill Rating through Cosine Similarity Analysis]]></title>
            <link>https://medium.com/inst414-data-science-tech/predicting-a-soccer-players-overall-skill-rating-through-cosine-similarity-analysis-d49d448aec38?source=rss-79350568bfe9------2</link>
            <guid isPermaLink="false">https://medium.com/p/d49d448aec38</guid>
            <category><![CDATA[inst414spr24a06]]></category>
            <dc:creator><![CDATA[Eitan Zavorin]]></dc:creator>
            <pubDate>Thu, 02 May 2024 03:45:32 GMT</pubDate>
            <atom:updated>2024-05-16T15:47:33.639Z</atom:updated>
            <content:encoded><![CDATA[<p>In soccer analytics, understanding player performance and similarity is crucial for various stakeholders, including coaches, team managers, and scouts. One interesting question that I am choosing to research: How can we identify players with similar playing styles and skills to a target player and predict the target player’s overall rating based on their attributes? Answering this question could bring valuable insights to team managers and scouts when considering potential transfers or assessing the value of existing players. Therefore, this data exploration could inform decisions related to player recruitment, team composition, and strategic planning.</p><p>The dataset used in this analysis comprises of the Player and Player_Attributes tables from Kaggle’s public European Soccer Database. These tables contain detailed attribute-related information on soccer players, such as finishing, dribbling, ball control, sprint speed, passing, and more. Each record in the dataset represents a soccer player and includes various quantitative metrics that quantify their play styles and skill levels. Using these attributes for players in the database allows us to measure similarities based on multiple dimensions.</p><p>For my analysis, I chose to carry out a regression as the predictive modeling approach since I aim to predict the player’s overall rating, a quantitative, continuous variable. Regression models are well-suited for predicting numerical values, making them ideal for this scenario.</p><p>The attributes and features used for the supervised model include various player traits such as speed, passing accuracy, dribbling, shooting, and sprint speed. These attributes are crucial indicators of a player’s technical abilities and play style, which help for identifying player similarity and predicting overall rating.</p><p>By analyzing the collected data and applying the regression model, we can identify players with similar playing styles and skills to a target player and predict the target player’s overall rating.</p><p>Github Repository: <a href="https://github.com/eitanzav/INST414Module6Assignment">https://github.com/eitanzav/INST414Module6Assignment</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d49d448aec38" width="1" height="1" alt=""><hr><p><a href="https://medium.com/inst414-data-science-tech/predicting-a-soccer-players-overall-skill-rating-through-cosine-similarity-analysis-d49d448aec38">Predicting a Soccer Player’s Overall Skill Rating through Cosine Similarity Analysis</a> was originally published in <a href="https://medium.com/inst414-data-science-tech">INST414: Data Science Techniques</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Art of Soccer Strategy: Exploring Team Dynamics through Clustering]]></title>
            <link>https://medium.com/inst414-data-science-tech/the-art-of-soccer-strategy-exploring-team-dynamics-through-clustering-ec5a46518f40?source=rss-79350568bfe9------2</link>
            <guid isPermaLink="false">https://medium.com/p/ec5a46518f40</guid>
            <category><![CDATA[inst414spr24a04]]></category>
            <dc:creator><![CDATA[Eitan Zavorin]]></dc:creator>
            <pubDate>Thu, 02 May 2024 00:23:26 GMT</pubDate>
            <atom:updated>2024-05-03T14:48:42.637Z</atom:updated>
            <content:encoded><![CDATA[<p>In soccer, it is essential for team managers and coaches to understand the nuances of team dynamics and play styles in order to effectively strategize when facing opposing teams. Professional teams — especially the biggest ones — constantly employ large teams of analysts, tons of money, and countless hours of work to plan tactics for upcoming games. Each team has a unique approach to dissecting the vast data on other teams’ play styles and statistics to inform their coaching strategies and game tactics. In a thorough data analysis, I aim to provide team managers and coaches with a new look into how they can analyze data to prepare for upcoming matches. Using a dataset with roughly 300 teams in the world’s top soccer leagues, I will be investigating the following research question: By uncovering patterns of similarity among soccer teams based on their play styles, can I help soccer team managers and coaches make more accurate, informed decisions in their preparations for upcoming games?</p><p>The data I will use to answer this question comes from Kaggle’s public European Soccer Database, which includes vast information regarding about 300 teams and their play style attributes, such as build-up play speed, chance creation positioning, and defense aggression. Utilizing these attributes allows us to understand a quantifiable and comprehensive overview of a team’s playing style. By capturing the essence of how teams approach various facets of the game, we can gather insights into their strategic tendencies.</p><p>After carefully cleaning the data by removing irrelevant columns, non-quantitative attributes, and columns with mostly null values, my final dataset is ready to be clustered. In order to do this, I am using traditional K-means clustering, which clusters based on pairwise Euclidean distances from point to point. After testing many different k values, it became clear to me that k = 5 best splits the data while avoiding redundancy between the top attributes and play styles between different clusters. The k value was also determined using my own expertise in the subject matter, because I could tell when clusters were too similar based on the teams that were in them and the top attributes in each. Once I established a clustering model that seemed accurate based on my high-level understanding of the sport, I ended with these clusters:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/992/1*CD8b4x6Rl2LorsDfT36T_w.png" /></figure><p>These five clusters depict 5 different types of teams that distinguish from each other by their tactical approaches. Cluster 1 encompasses teams that emphasize defensive solidity and moderate tempo in build-up play, with some examples being Valencia CF and Club Brugge KV. Cluster 2 exhibits proactive attacking tactics, prioritizing chance creation and shooting, as seen in Eintracht Frankfurt and Fiorentina. Cluster 3 focuses on high-tempo, expansive strategy, such as Manchester United and West Ham United. Cluster 4 excels in defensive structure, positional discipline, and wide attacking play, as demonstrated by Olympique Lyonnais and Crystal Palace. Lastly, Cluster 5 represents more possessive teams focusing on controlling the game and moving the ball forward slowly, typified by Milan and Feyenoord. Judging by the accurate team samples, it’s clear that the teams that share clusters do, in fact, have many things in common that the average fan or team coach might not be aware of.</p><p>Upon analyzing the clusters I have laid out based on all of these teams’ play style attributes, we can not only discern distinct clusters representing different tactical approaches, but we can also gain a ton of insight on any given team based on its attributes and the teams that are similar in play style. Here’s a possible scenario in which a manager or coach could use the clusters I have found:</p><p>Suppose I am the manager of the English league team, Newcastle United. Every year, my team plays several seasonal games against Manchester United, but we rarely play against Athletic Club de Bilbao, which plays in the Spanish league. Suppose that in a European tournament, Newcastle United gets scheduled to play against Athletic Club de Bilbao. It may be intimidating to be put against a team that we rarely play since we know less of what to expect in their play style. However, if we take a look at Cluster 3 above, we can see that Athletic Club de Bilbao has a very similar play style to Manchester United — who we know how to play against. By knowing this connection in similarities and tactics Athletic Club de Bilbao is known to play with, we could gain some priceless insights on how to tactically prepare for our game against a team we otherwise wouldn’t be familiar with. This is just one hypothetical scenario of how my clustering model could bring valuable insights to guide managerial decision-making for team success on short-term and long-term scales. So, to answer my initial research question: Uncovering patterns of similarity among soccer teams based on their play styles can really help soccer team managers and coaches make more accurate, informed decisions in their preparations for upcoming games.</p><p>Before I can finish, it’s essential to acknowledge the limitations in my research. Firstly, the available play style attributes do not capture the full complexity of team dynamics; it’s a certainty that the model overlooks subtle nuances that aren’t present in the data. Additionally, the dataset was last updated eight years ago, and the attributes and team strategies could be somewhat different now than they were eight years ago. With a more detailed and up-to-date dataset, we could acquire much more accurate calculations and gather much more useful insights on the teams. With that said, this is a great start toward bringing a way for coaches to strategize and steer their teams to short-term and long-term success.</p><p>Link to GitHub Repository: <a href="https://github.com/eitanzav/INST414Module4Assignment">https://github.com/eitanzav/INST414Module4Assignment</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ec5a46518f40" width="1" height="1" alt=""><hr><p><a href="https://medium.com/inst414-data-science-tech/the-art-of-soccer-strategy-exploring-team-dynamics-through-clustering-ec5a46518f40">The Art of Soccer Strategy: Exploring Team Dynamics through Clustering</a> was originally published in <a href="https://medium.com/inst414-data-science-tech">INST414: Data Science Techniques</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Discovering the Play-Style Doppelgängers of our Favorite Soccer Players]]></title>
            <link>https://medium.com/inst414-data-science-tech/discovering-the-play-style-doppelg%C3%A4ngers-of-our-favorite-soccer-players-42e93c5b2c6c?source=rss-79350568bfe9------2</link>
            <guid isPermaLink="false">https://medium.com/p/42e93c5b2c6c</guid>
            <category><![CDATA[inst414spr24a03]]></category>
            <dc:creator><![CDATA[Eitan Zavorin]]></dc:creator>
            <pubDate>Wed, 01 May 2024 03:39:38 GMT</pubDate>
            <atom:updated>2024-05-03T14:47:40.698Z</atom:updated>
            <content:encoded><![CDATA[<h3>Discovering the Play-Style Doppelgängers of the World’s Greatest Soccer Players</h3><p>In the rapidly evolving world of soccer, identifying player similarities is a crucial task that can provide valuable insights for various stakeholders, such as coaches, talent scouts, club executives, and fans. By carefully measuring similarities between soccer players based on their play styles and skill attributes, we can answer very important and controversial questions about player comparisons and potential talent recruitment strategies. In this report, I will delve into the process of creating a model that pinpoints the most similar soccer players to three selected individuals. Utilizing cosine similarity calculations on a vast dataset of players and their attributes, I aim to uncover hidden patterns and similarities that inform decision-making processes in the world of soccer. Knowing that the most controversial question among soccer fans around the world is whether Lionel Messi or Cristiano Ronaldo is the best player in the world, I’ve chosen the following research question: With big names like Messi, Ronaldo, and Wayne Rooney stealing the spotlight over the recent decades of soccer, who are the most comparable players to them that deserve more attention?</p><p>Before diving into my investigation, it’s crucial to understand my stakeholders so that I can bring the most value to them in the creation of my model and the discoveries I make along the way. One primary category of stakeholders invested in knowing player similarities is team recruitment staff and talent scouts. These professionals seek to proactively identify potential talent prospects for transfer or recruitment to improve team performance. Scouts can make educated decisions regarding transfer targets by accurately assessing player similarities, ensuring that new recruits complement pre-existing team dynamics. Additionally, coaches and team managers could benefit from insights into player similarities, utilizing them to plan training sessions, tactics, and formations based on player profiles and play styles. Lastly, fans will always be a stakeholder in any soccer-related investigation such as this one. Fans in soccer communities around the world could benefit from deepening their understanding of player performance and the game, enhancing their viewing experience, fostering engaging discussions about the sport, and even inspiring them to learn from the greats.</p><p>The dataset used in this analysis comprises of the Player and Player_Attributes tables from Kaggle’s public European Soccer Database. These tables contain detailed attribute-related information on soccer players, such as finishing, dribbling, ball control, sprint speed, passing, and more. Each record in the dataset represents a soccer player and includes various quantitative metrics that quantify their play styles and skill levels. Using these attributes for players in the database allows us to measure similarities based on multiple dimensions.</p><p>To collect the dataset, we transformed the publicly available SQLite format available on Kaggle into a CSV file. Using the pandas library, I extracted the data and transformed it to create a structured dataset suitable for analysis. I cleaned the data by removing duplicates, removing rows with significant missing values, and selecting only the relevant quantitative attributes for similarity measurement. Eventually, I was left with a dataset of about 600 player rows, including only the relevant player attributes necessary to proceed with my investigation.</p><p>In my analysis, I measure the similarities between soccer players using cosine distance calculations, which also tell us the cosine similarity. Each player is represented as a vector of attributes including finishing, dribbling, ball control, etc., and the cosine distance is computed between pairs of player vectors to quantify their similarities.</p><p>For three selected players — Lionel Messi, Cristiano Ronaldo, and Wayne Rooney — I identify the top 10 most similar players in the dataset based on cosine distance scores on all their relevant play style attributes. By discovering players who exhibit similar skill profiles and playing styles, we gain insights into some very influential but possibly less appreciated players in the game. Here are my findings:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tLbr31TelZKLn2WO7FV7pw.png" /></figure><p>The analysis revealed interesting patterns of player similarities across various dimensions. For example, Ronaldo, known for his exceptional finishing and long shot abilities, demonstrates high similarity scores with players who excel in similar attributes. Similarly, Messi, renowned for his outstanding dribbling, ball control, passing and vision, shares similarities with players who exhibit similar playmaking abilities. Thirdly, Wayne Rooney, whose strength, aggression, and shooting traits stand out amongst the competition, pair him closely with many players known for sharing the same characteristics. These findings provide an answer to our research question, as well as numerous valuable insights for talent scouts and team managers seeking to identify players who align with specific skill requirements that only some of the best players are known to have.</p><p>Despite the insights gained from my investigation, it’s crucial to acknowledge the limitations and biases inherent in my approach. One limitation is the complete dependence on quantitative attributes, which may not capture the full spectrum of player abilities and characteristics. Some qualitative measures, such as attacking and defensive work rates (assigned as low, medium, or high), are important to take into account when measuring the player profiles and players. Additionally, the model’s calculations may be biased towards players with skill profiles or playing styles prevalent in the dataset, possibly overlooking players with unique sets of attributes or unconventional play styles. Lastly, this data was last updated eight years ago, so it’s missing updates on many player profiles and completely misses any players that emerged onto the scene in the last eight years. Regardless, this model would work on up-to-date datasets, and the insights that our stakeholders can gather from it are still very much valid and valuable.</p><p>In conclusion, my analysis exhibits the power of data-driven approaches in uncovering player similarities and informing decision-making processes in various aspects of the world of soccer. By utilizing cosine distance calculations on a vast dataset of player attributes, we can gain valuable insights into player comparisons and talent recruitment strategies. Moving forward, additional research and refinement of our model could improve its accuracy, relevance, and applicability, ultimately empowering stakeholders to make more educated decisions in the dynamic world of soccer.</p><p>GitHub Repository Link: <a href="https://github.com/eitanzav/INST414Module3Assignment">https://github.com/eitanzav/INST414Module3Assignment</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42e93c5b2c6c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/inst414-data-science-tech/discovering-the-play-style-doppelg%C3%A4ngers-of-our-favorite-soccer-players-42e93c5b2c6c">Discovering the Play-Style Doppelgängers of our Favorite Soccer Players</a> was originally published in <a href="https://medium.com/inst414-data-science-tech">INST414: Data Science Techniques</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Two Types of Soccer Players: Shaping vs. Revolutionizing the Game]]></title>
            <link>https://medium.com/inst414-data-science-tech/the-two-types-of-soccer-players-shaping-vs-revolutionizing-the-game-6ac26cd9d94a?source=rss-79350568bfe9------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ac26cd9d94a</guid>
            <category><![CDATA[inst414spr24a02]]></category>
            <dc:creator><![CDATA[Eitan Zavorin]]></dc:creator>
            <pubDate>Tue, 30 Apr 2024 23:19:02 GMT</pubDate>
            <atom:updated>2024-05-03T14:46:56.658Z</atom:updated>
            <content:encoded><![CDATA[<p>In the realm of sports analytics, understanding the complex relationships among players can reveal invaluable insights for coaches, team managers, and scouts. Leveraging network analysis techniques, I delve into the structure of a web-based network centered around soccer players who have bolstered the ranks of current and previous generations. Using my analysis, we embark on a journey to answer a fundamental question: Who are the key players shaping the dynamics of the soccer field?</p><p>The burning question in the minds of team managers, coaches, scouts, and fans revolves around identifying the most influential players whose play style and performance drive their team’s overall success. By deciphering a large network of soccer players, including vast data on players’ attributes, play styles, and skill ratings, I aim to pinpoint the key influencers whose strategic contributions steer their team toward victory.</p><p>In order to optimize the value of my investigation, it’s crucial to consider the stakeholders invested in the outcomes of the analysis. The primary stakeholders that the soccer world comprises of are team managers, coaches, scouts, club executives, and fans. For managers and coaches, insights into player centrality and influence within the network can inform critical decisions related to transfer targets, tactical strategies, starting lineups, and substitutions during matches. Scouts and talent evaluators depend on these analyses to identify talent prospects for transfers and recruitment. Additionally, club executives use these insights for decision guidance regarding player contracts, financial investment, recruitment strategies, and the long-term development of the team. Lastly and most definitely not least, the huge fanbases in the soccer community could benefit from a deeper understanding of player dynamics and performance, enhancing their watching experience, fostering engaging discussions about the sport, and even inspiring them to learn from the greats.</p><p>To determine which key players are shaping the dynamics of the soccer game, I will utilize and manipulate the public European Soccer Database dataset on Kaggle. This entire dataset includes seven tables — Country, League, Match, Player, Player_Attributes, Team, and Team_Attributes — containing a combined total of over 200,000 records, capturing several generations of soccer from numerous angles. For my investigation on soccer players, I will be using cleaned, trimmed, and merged versions of the Player and Player_Attributes table. After various steps of cleaning, the final dataset I am using contains roughly 600 rows. Each row represents a current or former player with a minimum overall skill rating of 82 (out of 100), and includes 32 columns of information regarding their play style and skill levels, such as attacking/defensive work rates, dribbling, passing, shooting, and more.</p><p>Before cleaning the data, I first had to prepare it by transforming it from the SQLite format it was in on Kaggle to CSV using DB Browser for SQLite. Once that was done, I could import it into my Jupyter Labs workspace to begin cleaning and organizing. First, I executed an inner join on the Player and Player_Attributes tables using each player’s unique player_ID, so that player names could be next to their attributes in one table. For relevance and downsizing my dataset to an interactable size, I removed all players with overall ratings under 82 (stakeholders would be less interested in worse players, anyway). Then, I dropped the player duplicates and null values from the table. The inner join created some duplicate columns, so I removed those along with several other columns that were irrelevant to the investigation — anything unrelated to play style or skill level. After all of these steps, my final data frame had just over 600 rows of players, including only the relevant attributes necessary to move forward with my investigation.</p><p>In the context of my network analysis, each player represents a node or vertex in the graph. These nodes encapsulate the individual soccer players, each characterized by a unique set of attributes depicting their skills and play style. An edge or relationship is created between two players when they share a certain number of equal attributes, which signifies similarities in their play style, skill level ratings, or other relevant characteristics. As for defining “importance” within the network, nodes with higher centrality degrees possess more similarities with other players. It’s crucial to understand here that both ends of the centrality degree spectrum are valuable nodes in the context of our investigation. Players with high degrees of centrality and “importance” in the network can provide insights into what characteristics are seen in most of the successful, high-rated players in the world. On the other side of the spectrum, outliers with low centrality degrees might not necessarily be “important” from a network centrality perspective, but they display what types of players, play styles, and skills stand out amongst the crowd. Thus, stakeholders could gain countless insights by diving deep into both sides. By identifying and analyzing both high centrality nodes and low centrality nodes, we gain valuable insights into the key players shaping the dynamics of the soccer field and driving team success. For some examples of the most “important” nodes in my graph, here are the top 10 players in centrality degree ratings:</p><ol><li>Iker Casillas 0.02463054187192118</li><li>Mario Gomez 0.02134646962233169</li><li>Mickael Landreau 0.02134646962233169</li><li>Carlos Kameni 0.019704433497536946</li><li>Sebastian Frey 0.019704433497536946</li><li>Florent Malouda 0.0180623973727422</li><li>Gabi 0.0180623973727422</li><li>Heiko Westermann 0.0180623973727422</li><li>Helton 0.0180623973727422</li><li>Jakub Blaszczykowski 0.0180623973727422</li></ol><p>These players demonstrate the highest number of similarities among the top-rated players. Looking into their attributes would give insight into what play styles could make so many players successful. Now, for more graphs, observations, and insights:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*L4EgM9qYDH9DwHBg" /></figure><p>Here is one articulation of the network in its entirety. At first glance, this visual provides little meaning, but if you hover over one of the nodes (shown below), it will show you all the nodes connected to it through similarities in play style and skill. This can lead to countless insights, such as the discovery of who plays similarly to a particular player or transfer target.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SSUfCmIrfp0WNrDI" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1019/0*A4pdOQ2BnIN2jB-o" /></figure><p>Here is another visual of the entire network, displaying both sides of the spectrum of centrality, as I mentioned earlier. In the eye of the circle, we can see the players that are most similar to each other, while the outside shows the players who stand out in regard to their skills and play styles. In truth, there is no simple answer to our research question of who the key players shaping the dynamics of soccer are. On the one hand, it can be argued that the players in the center, who all have high skill ratings and influence on their team, outline the common skills and play styles found in most successful and influential soccer players. By exploring this subset, we have found that some of the most influential players include Iker Casillas, Mario Gomez, and Samuel Eto’o. Conversely, it can also be argued that the players on the outside are revolutionizing the game, steering their teams to success with a very unique set of skills. By exploring this subset, we have found that some of the most revolutionary and successful players include Andrea Pirlo, Paul Scholes, and Zlatan Ibrahimovic. At the end of the day, both types of players shape the dynamics of the game in their own ways, and neither side can take all the credit.</p><p>But don’t be mistaken. That doesnt mean that these findings don’t give us very valuable insights. Exploring and interacting with the graphs above would guide managers and coaches in strategy, help scouts identify talents for recruitment, enable club executives to drive long-term developmental and organizational decisions, and give fans around the world a deeper understanding of the leaders in the game they love.</p><p>Finally, it’s important to note that this is far from the end of the conversation. This data analysis could be expanded in many different ways to address some of its limitations, the biggest being the lack of player performance statistics. Although player ratings in various characteristics are a good way of measuring a player’s play style and skills, this could be enhanced further by taking seasonal/all-time performance stats into account. Knowing the number of goal contributions, shot blocks, saves, etc., would allow us to gather even more insight from our data than we can now. Secondly, having data on each player’s position would allow us to dive deeper into each position, comparing players of the same position to each other. These future steps could be taken to expand our knowledge of the topic and grant even more value to the stakeholders interested.</p><p><a href="https://github.com/eitanzav/INST414Module2Assignment">https://github.com/eitanzav/INST414Module2Assignment</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ac26cd9d94a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/inst414-data-science-tech/the-two-types-of-soccer-players-shaping-vs-revolutionizing-the-game-6ac26cd9d94a">The Two Types of Soccer Players: Shaping vs. Revolutionizing the Game</a> was originally published in <a href="https://medium.com/inst414-data-science-tech">INST414: Data Science Techniques</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Highest-Selling First-Person Shooter Game and Series in the World]]></title>
            <link>https://medium.com/inst414-data-science-tech/the-highest-selling-first-person-shooter-game-and-series-in-the-world-63b2eadb1c80?source=rss-79350568bfe9------2</link>
            <guid isPermaLink="false">https://medium.com/p/63b2eadb1c80</guid>
            <category><![CDATA[inst414spr24a01]]></category>
            <dc:creator><![CDATA[Eitan Zavorin]]></dc:creator>
            <pubDate>Fri, 26 Apr 2024 00:03:26 GMT</pubDate>
            <atom:updated>2024-04-29T15:39:45.324Z</atom:updated>
            <content:encoded><![CDATA[<p>Since I was a kid, I’ve always enjoyed playing video games with my friends. My friends and I loved to play games like Fifa, Rocket League, Fortnite, and most of all, Call of Duty. For two decades, Call of Duty has been one of the world’s most popular video game franchises. It’s been a huge part of the lives of many first-person shooter game lovers, including myself. But with other first-person shooter games like Battlefield, Apex Legends, and Halo blowing up in recent decades, gamers have been constantly debating the same question: What is the number 1 FPS game/series of all time?</p><p>At the end of the day, the answers to this question will forever be subjective; every gamer’s answers to this question may be different due to their unique experiences playing video games. However, using data on global video game sales, I’m going to attempt to answer these two research questions from a fact-based, “objective” perspective.</p><p>By finding an answer to this research question, we could not only inform the massive fanbases of the answer from a purely fact-based, data-analytical perspective, but we can also potentially assist the developers of less popular franchises in understanding why they aren’t the leader of their industry. By analyzing the differences in sales metrics of different series and games, we can help first-person shooter developers set their sights and standards to the best of the best. By comparing their less-bought games to more popular games of the same genre, this group of stakeholders could gather countless insights on how to improve their games to reach — and even surpass — their competition. Another group of stakeholders that would be interested in my findings are investors and shareholders. Investors in game development companies would be interested in the financial performance and profitability of popular and unpopular video games. My findings could very much influence their investment decisions. Lastly, one major group of stakeholders in my findings is retailers and distributors of video games. Whether these companies are selling online or physical copies, they would benefit greatly from knowing which games are in high demand and which ones aren’t. Just like with investors, and the video game developers themselves, these findings could very much influence their decisions to achieve greater financial success.</p><p>To determine which first-person shooter game and series is the best of all time, I will be utilizing and manipulating the public Video Game Sales dataset on Kaggle. This entire dataset includes 16,600 records, each representing a video game and various attributes related to that game, such as its genre, ranking, platform, release date, developer, and of course, number of sales. It also includes global sales with a breakdown by location: Europe, North America, Japan, and Other. In order to clean and organize the data, I first dropped all the duplicate rows using the Name column, since I noticed there are some duplicate rows (multiple rows for one game). Then, I filtered the data frame to include only the games in the ‘Shooter’ genre since the other genres are not relevant to my research question. To further remove unnecessary data, I dropped the columns of sales by location (EU_Sales, NA_Sales, JP_Sales, and Other_Sales), since we are only interested in global sales. This could be noted as a potential step in future research: to research further how the sales rankings vary based on location. For now, we will set that idea aside and just find the global, overall sales rankings, since those findings bring a lot more meaning to the context of the problem and value to the stakeholders. Lastly, I dropped the Platform column because the first platform that the games came out on doesn’t matter in the context of this problem. With all these organizational changes made, the final data frame we are left with has the columns: ID, Rank, Name, Year, Genre, Publisher, and Global_Sales. With these columns left, we have appropriately extracted only the necessary and important data for the context of our research questions.</p><p>After utilizing Python’s matplotlib, numpy, and pandas packages to rearrange and visualize the data, I finished with some very insightful visuals. Before I get into the findings, I’m going to restate the research question so that we know exactly how to interpret the visuals and findings I am about to present.</p><p>Research question: Based on global video game sales, what is the number 1 FPS game/series of all time?</p><p>Below are two visualizations that give us the insight we need to answer our research question.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/989/0*vjjltPZ-vao7mWim" /></figure><p>This bar graph displays the top 10 highest-selling first-person shooter games, according to our data. From this, we can gather two large observations and takeaways in relation to our research question. Firstly, Nintendo’s Duck Hunt is clear to be the highest-selling game in its genre, almost doubling the game in second place. Secondly, if you take a closer look, you’ll see that 6 out of the 10 games (and 5 out of the top 6) belong to the Call of Duty video game franchise. These two observations leave us with a very interesting conclusion: when it comes down to global sales, the single most popular first-person shooter game in the world is Duck Hunt, but the highest-selling first-person shooter game series is Call of Duty. To take an even deeper look into this, I present my second visualization.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/670/0*KO6gpcm3_t797lS4" /></figure><p>This pie chart displays the five highest-selling publishers for first-person shooter games. As we can see, Activision leads the five, with Microsoft Game Studios in second, Electronic Arts in third, and Nintendo in fourth. This builds on our insights from the first visual, which is that Activision’s Call of Duty is, in fact, the most popular FPS series in the world, despite Nintendo’s Duck Hunt being the highest-selling single game in the genre. With this, we can conclude our investigation of our research question. Based on global sales alone, we know that the number one first-person shooter game in the world is Duck Hunt, and the number one series is Call of Duty.</p><p>An important limitation to note: This data was last updated eight years ago, meaning that our findings might still be somewhat accurate to this day but will not account for any video game sales within the first eight years. An argument can be made that this data being not fully up-to-date can reduce its validity, to which I respond with two thoughts. Firstly, even with some new FPS games emerging and bolstering the ranks over the last eight years, the highest sold games of all time from eight years ago will still be very high on the list. Secondly, the previously listed stakeholders could still gain a lot of this data; identifying which games were the most successful even eight years ago, will allow game developers to identify successful elements in those games, investors to have an idea of which developers succeed the most, and the gaming community to understand fact-based answers to their long debated question of which FPS game is the best.</p><p><a href="https://github.com/eitanzav/INST414Module1Assignment">https://github.com/eitanzav/INST414Module1Assignment</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=63b2eadb1c80" width="1" height="1" alt=""><hr><p><a href="https://medium.com/inst414-data-science-tech/the-highest-selling-first-person-shooter-game-and-series-in-the-world-63b2eadb1c80">The Highest-Selling First-Person Shooter Game and Series in the World</a> was originally published in <a href="https://medium.com/inst414-data-science-tech">INST414: Data Science Techniques</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>