What's in the Game? : An Exploratory Design Study about Football Statistics

(1)

Spring term 2020 | LIU-IDA/KOGVET-A--20/008--SE

i

What’s in the Game?

An Exploratory Design Study about Football

Statistics

Johan Karlsson

Supervisor: Katerina Vrotsou Examinator: Arne Jönsson

(2)

ii Upphovsrätt

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare – under 25 år från publiceringsdatum under förutsättning att inga extraordinära omständigheter uppstår.

Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och tillgängligheten finns lösningar av teknisk och administrativ art.

Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsmannens litterära eller konstnärliga anseende eller egenart.

För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/.

Copyright

The publishers will keep this document online on the Internet – or its possible replacement – for a period of 25 years starting from the date of publication barring exceptional circumstances.

The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.

According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.

For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.

(3)

iii Abstract

The interest in football around the world is ever increasing with more than half of the world’s population watching the last World Cup. The amount and the granularity of information available about football is endless. In recent years the advent of new technology has led to possibilities to track and log all aspects of the game, making advanced statistics available to everyone. The focus of the research community of football statistics today is on professionals as the end recipient. The knowledge, requirements and goals of enthusiasts and experts differ which means that the design of statistics targeted at the different groups should differ.

This design study took an exploratory mixed methods approach to study how to design football statistics for enthusiasts. A variety of methods were used including, expert interview, survey and interviews with 8 football enthusiasts led to design suggestions of statistics which were evaluated in two phases. The iterative approach led to general guidelines of how to design football statistics, and 12 design suggestions of visualizations of statistics. The visualizations function as instantiations of the knowledge gained throughout the study. The design suggestions communicate information about how the conclusions drawn in the study can be practically implemented in a smartphone application about football statistics. The results of the study can be used to guide the design of football statistics for enthusiasts.

Keywords:

(4)

iv Acknowledgement

I would like to thank my supervisor Katerina Vrotsou for her feedback, guidance and reassurance during the course of this project. I would also like to thank Twelve Football for their hospitality and support, and for giving me the opportunity to do my master thesis about a subject that is dear to my heart. Additionally, I would like to thank all the participants, without whom this project would not have been possible. Finally, a special thanks to Razmus, Anna and Joel for their help and support during the process.

(5)

v

5.6 Player ... 40 5.7 Team ... 45 5.8 League ... 48 5.9 Match ... 50 6 Discussion ... 54 6.1 Expert Interviews ... 54 6.2 Survey ... 55 6.3 Interviews ... 55 6.4 Evaluation ... 57 6.5 Design Suggestions ... 58 7 Conclusion ... 61 8 References ... 62 Appendix 1 ... 67 Appendix 2 ... 68 Appendix 3 ... 70 Appendix 4 ... 71 Appendix 5 ... 72 Appendix 6 ... 84

(7)

1

1 Introduction

Football is the biggest sport in the world with 265 million people practicing it in 2007 (the last world-wide count of FIFA) (Kunz, 2007). The interest for the sport is even bigger from a consumer standpoint with more than half of the world’s population watching the FIFA world cup in 2018 (FIFA, 2018). This makes football one of the most valuable businesses in the world with the European market alone being worth 25 billion dollars (Deloitte UK, 2018).

The interest seems to be ever increasing and the technological evolution during the last decades has created new opportunities and new ways for people to communicate, bet, play, theorize and in other ways consume the sport. Today there is opportunity to watch live football every hour of the day, make fantasy leagues with teams consisting of players who might perform well the coming weekend and watch experts talk about who they think are good. New technology has also made it possible to collect match data on a new level of detail which means that every action and movement of the players can be tracked and logged. In short, there are endless possibilities for the interested to dissect every aspect of their favorite team or player.

Meanwhile statistics are becoming a natural part of our lives. It is used everywhere to predict outcomes and behavior and to make our everyday lives easier or supply compelling information. Weather forecasts, for example, are based on statistics, something that we take for granted and have been and natural part of our lives and planning of them for decades. A more recent example is how big internet corporations use statistics in the collection of data to streamline the lives of their users in every way and to provide engaging content. As mentioned above this trend can be seen in football as well with an increased collection of data and statistics. Being the biggest sport in the world and a major part of culture and society across the world the interest of football statistics can be found in a variety of stakeholders. The focus of the research community on sports statistics takes the perspective of the practitioners and teams. While this is an interesting aspect, most people interested in football are not the people practicing it but the enthusiasts. As the two groups do not necessarily have the same agenda and background knowledge, there should be a difference between the design of statistics and analytics for the two groups. This thesis therefore aims to study how statistics and analytics should be presented to people that are interested in the topic but not considered experts.

(8)

2

2 Background

Sports data have been collected for decades but with the advance of technology the amount and granularity of the collected data have seen a big increase lately. This has led to opportunities for sports clubs to extract valuable information from the data collected in order to help inform decisions and improve their performance (Shahzeb, 2017). During the early 2000s the Oakland Athletics and their general manager Billy Beane pioneered the use of data in the organization of sports clubs. The baseball club was at the time struggling with bad performance on the pitch and financial difficulties off it. To solve this problem, Beane used a data-driven approach to recruit overlooked players for a fraction of the price of the major stars. By using sabermetrics to evaluate and scout players The Oakland Athletics achieved a record breaking 20 consecutive wins despite their restricted budget (Lewis, 2003). The story has been told in the book and film

Moneyball which popularized the term sports analytics to the mainstream.

The statistics collected by the teams, leagues and third parties are now used in sport analytics to explain and predict performance. Today most of the big sports clubs in the world have data analysts working for them. Sports analytics are usually divided into two different categories, on-field analytics and off-field analytics (proemsports.com, 2017). On-field analytics, as the name suggests, focus on performance aspects during the games such as player fitness, game-strategy and shot statistics. Teams use models and simulations to analyze the data and earn valuable knowledge about how to maximize their strategy and make use of the weaknesses of the opponents. Off-the field analytics on the other hand deals more with the business side of running a sports club. It includes on-field statistics but also deals with other parameters that are associated with running a successful sports club, including ticket sales, management and marketing. This study will however focus primarily on the on-field analytics and how to spark interest and engage supporters with them in football.

Data provider firms like Opta have been collecting and categorizing match data from different sports for the last decades (optasports.com n.d.). Today the collection process takes place in real-time during live games. Every single action on the field is tracked, categorized and logged. This produces a vast amount of data during each game. The access to data has created opportunities for new knowledge and insights to be gained. To do that however the raw data itself is not the answer, instead it is just a commodity from which to extract valuable information. The sheer amount of data does however pose a problem in it of itself, by creating difficulties extracting the relevant information. This is referred to as the information overload

(9)

3

task at hand, it being processed or presented in an inappropriate way, making the information useless. The media landscape of today has made us, the consumers use to receive information that is packaged and designed for easy comprehension.

As a way to make sense of data and thereby solve the information overload problem machine learning and algorithms are used (Bzdok, Altman, & Krzywinski, 2018). Machine learning as a field of study has gained a lot of traction the last decade and is increasingly used in different domains. Machine learning takes a different approach from traditional programming and is inspired by the human way of learning. Instead of relying on predetermined rules and outputs programmed into the system, machine learning systems are based on patterns and inference learned from experience. Machine learning systems uses training data to learn these patterns and inferences that can then be used to explain the data, and to predict and model future possible outcomes (Flach, 2012).

Machine learning is taking advantage of the processing power of computers. Computers are vastly better at processing large amounts of data than humans. A regular laptop can take in millions of data points in a fraction of the time of a human. In sports analytics this makes it possible to evaluate thousands of games in a short period of time (Sumpter, 2016). By applying the right methods to this data, conclusions of interesting phenomena in the game can be identified. This in turn can lead to an understanding of player patterns and behaviors that result in successful and unsuccessful outcomes. Which can be used to make informed decisions and strategy changes in matches.

The output from machine learning systems is however only a step in the right direction from making raw data into valuable information. How the data is represented influences the interpretation of it by the user. To gain knowledge from the output of the system it must be adapted to the user. Nielsen (1993) points out that when an informational exchange between the human and the computer have not been optimized, the whole process may become highly time-consuming, inefficient and frustrating for the user.

Football analytics is used by a variety of stakeholders such as professional teams to make informed decisions, media outlets to provide engaging content for their costumers and betting companies to decide probabilities. It is a big field with many potential stakeholders from companies to private persons. Depending on who your target group is, however, the statistics and optimal way to visualize them differs. People with expertise in an area, (Sternberg, Sternberg, & Mio, 2012) such the staff of a football club generally has a broader understanding

(10)

4

of the game than a casual supporter. This leads to different understandings of the statistics and different requirements in the presentation of them.

Studies have previously shown that a majority of enthusiasts are unsatisfied with the live broadcasts and streaming services provided today. According to Deutsch, Harwood, Teller, & Deweese (n.d.) 61 % of enthusiasts say they are not satisfied with the broadcast and streaming available today. In the competitive entertainment climate of today, this is a threat to the sports industry. According to the enthusiasts asked in the survey some measures can be taken to improve the viewing experience. More than 40 % of enthusiasts for example, indicated that real-time statistics presented on the screen would increase their likelihood to watch. Instead many enthusiasts today say they seek live updates and statistics from other sources. 58 % use social media and other channels to get live updates and statistics from their favorite teams and players during games.

To the knowledge of the author, very little has been done within the field of football analytics from the perspective of the supporter as the end recipient. With the points brought up above as background there seems to be an opportunity to create a better overall fan experience through well-designed football analytics centered around the enthusiasts. The intent of this thesis is therefore to study how to present football analytics in a way that is appealing, engaging and valuable for the supporters.

2.1 Twelve Football

The study presented here was conducted in collaboration with Twelve Football. Twelve Football has since its founding been working with football statistics and analytics. Twelve Football is working towards clubs, enthusiasts and sponsors to create a deeper understanding of football.

The company is now in the process of creating a service which can provide match data to football enthusiasts in a compelling way. The vision of Twelve Football is to create a virtual commentator in the shape of a smartphone application which bases its output on facts, not opinion. The core of the service is the Twelve point system, a football expert algorithm formed by years of research that is using statistics and machine learning to assign a score to every single event that happens on the pitch. This is then used to rank the performance live during matches. The algorithm and ranking system are based on hundreds of thousands of historical match events which enables an objective evaluation of the players performances during the games. Twelve points system is based on expected goals, a term which has been popularized in recent

(11)

5

years. Expected goals is a statistical measure of the probability of scoring in any given situation. Twelve’s model uses match events from thousands of games as a statistical basis to calculate how each player’s action, increases or decreases its team’s probability of scoring, or refusing the other team to score. Each player action in a game is valued on a scale from -1000 to +1000, using the model. Each action is classified into one of four categories, defending, attacking, shots or off the ball. A player action that increases the team’s probability of scoring with 10 % is awarded 100 attacking points, while an action that increases the others team’s probability of scoring by 10 % is worth -100 defending points. The point system makes it possible to objectively rank players performances in games. The score of each category can be visualized, allowing people to see and analyze every event a player was involved in and understand how a player provides value to their team, and how their ranking was determined.

2.2 Purpose and Research Question

Twelves vision is to make advanced football analysis available to “regular football enthusiasts”. To do that it is important to present the data in an appealing and understandable way. The purpose of this project is therefore to first to learn about the target group, i.e. the football enthusiasts, by mapping out their requirements and goals regarding a potential application for football statistics. Secondly, based on that the goal is to create guidelines for how to design, present and visualize football statistics for enthusiasts. In addition, design suggestions will be created to illustrate how the guidelines can be practically implemented in a smartphone application. The research questions therefore are;

• How are enthusiasts consuming football statistics today and what are the factors that make it interesting and important?

• How should football statistics be designed, presented and visualized for football

enthusiasts?

• How should football statistics be implemented in a smartphone application for football

enthusiasts?

2.3 Delimitations

The project is limited to only include Swedish participants who consider themselves as football enthusiasts and experienced mobile users. There may therefore be limitations of the result which makes it ungeneralizable outside of this group. The goal of the study is to produce design suggestions based on goals and requirements of potential users. No implementation of those design suggestions will be made.

(12)

6

3 Theory

Below is a theoretical framework going explaining the study why it is important and putting the study in a scientific context. Firstly, two fields of research that has seen a rise in interest in recent years, research through design and information design are introduced, with the purpose of giving a background of the methodological approach of the study. Subsequently, different aspects of data visualization are presented with the purpose of providing an overview of the field. Finally, a brief explanation of considerations when designing for mobile platforms if provided.

3.1 Research through Design

Research through design (RtD) is a relatively new field and as all research methods its goal is to create new knowledge. The main difference between RtD and traditional research methods is that the objective is to explain how something could be rather than how something is (Löwgren, 2015).

Löwgren (2015) explains a typical RtD process as consisting of three parts, pre-studies, design practice and evaluation. The aim of the pre-studies is to assess the situation of interest today and to define a problem. The process often includes data collection and contact with the intended user in order to get an understanding of the context in which the product will be used. The pre-studies lead into the designing phase. The design practice and pre-study phases should however not be too separated, instead they guide each other forward and lead to new insights together. Löwgren divides the design phase into two steps, explorative and determinative. The objective with the explorative step is to create different possible solutions and to discover different directions which the design can take. In the determinative step the goal is to narrow down the possible solutions by eliminating the ones that do not solve the problem. During the evaluation phase the objective is to the test the design suggestions that have been created. As part of this process a prototype, meaning a physical representation of the product, is usually created. Prototyping and its result is an important part of the design process and it serves to represent, communicate around and test the product (Houde & Hill, 1997). Key insights can be gained in the implementation and testing of the product meaning that the final design can change even after this phase.

For the results to be reliable and of value to others, Löwgren (2015) highlights the importance of being transparent in the method and design process. There also needs to be a clear connection between the pre-studies and design practice with explanations of the design suggestions and

(13)

7

motivations behind them. This lets others to follow the development and the reasoning behind each design decision and ensures a well-founded end design.

Critique has been aimed at RtD because of the lack of standardized methods and trouble with reproducing the results. Gaver (2012) addressed this critique by referring to the user-centered approach that RtD studies inevitably have. Holtzblatt & Beyer (2017) describes user-centered design as a key factor to ensure the value of a product to the end user. A user-centered design approach is an iterative process that puts the needs and requriements of the user in focus at all stages of the process. In the beginning of a design project the end result is never known (Arvola, 2014). A good design process is often characterized by the exploration of a big solution space, meaning the creation and evaluation of different design suggestions. At the start of a project the potential design solutions are unlimited and the solution space is wide. As the ending of the project approaches, the needs and requirements of the user become more apparent and potential solutions become fewer as the “wrong” ones are excluded. Gaver (2012) states that the user-centered approach with data-driven creation and evaluation of sketches and design suggestions in some cases produces more knowledge than the use of classical analytical methods.

3.2 Information Design

During the last decades with the advance of the digital revolution, data has become an important and valuable asset. The words information and data are sometimes used interchangeably, they are however not the same thing. Information does not arise until the receiver perceives and understands the data (Pettersson, 2012). Data is ambiguous, the same message can be constructed, delivered and convey different information in a myriad of different ways, even though the raw data is the same. To be able to communicate a message in a way that conveys the intended information requires practice and expertise. This is precisely what information design is about. Information design is an interdisciplinary field which, according to Pettersson (2012), is about clarity of communication. It comprises a variety of different fields including language, art & aesthetic, cognition and communication. Pettersson states that in order to convey information in a good way it must be well designed, produced and distributed. To fulfill this Redish (2000) points out the importance of keeping the receiver in the center of the design process and that the final product is working for its user. According to Redish (2000), for the information to be working for its users they must be able to:

• Find what they need • Understand what they find

(14)

8 • And use what they understand appropriately

To increase the chance of fulfilling these three requirements Redish proposes a design process (figure 1). The process is divided into five steps but is at its core iterative meaning that there is no linear way of executing an information design project. The first step is however learning to know the intended receiver and information to communicate. This leads to planning of the project which requires you to have some general understanding from the first step. Then comes selecting, designing and testing the content, before producing the final product. All steps can be performed multiple times and returning to previous ones is recommended as new information and insights are gained. Even when the final product is produced improvement can be achieved by iterating the process again.

(15)

9

3.3 Quantitative Data Visualization

Visualizing data is the use of images to illustrate, comprehend and communicate information. There are a plenty of different methods to visualize quantitative data and the same data can be illustrated in a myriad of ways. The goal of visualizing data is to make it more accessible and easier to understand, if it is used in the wrong way it can lead to the opposite (Few, 2007). Which visualization method to use depends on the context and the receiver. There are, however, general rules and guidelines how to visualize data in a good way. In his pioneering work during the late 20th century Edward Tufte (2001) states the goal with the use of data visualization and guidelines to consider in the creation of them, these are presented in more detail below.

The objective of data visualization is according to Tufte (2001), to show the relevant data and to do it in a way so the reader comprehends the intended information and not focus on the method, graphical design or any other irrelevant aspect. Achieving this means presenting a lot of coherent data in a small space. Which encourages the reader to compare different parts of the data on different levels of granularity. It is also important that the visualization fulfills a clear purpose in its context which means being integrated with other statistical information and written description of it.

To achieve this Tufte highlights what he calls the “data ratio”. Generally higher the ink-data ratio the better the visualization. High ink-ink-data ratio is attained by simplifying the information in all variables as much as possible without distorting the data and exclude everything that does not communicate relevant information. Above all however, Tufte highlights the importance of review, edit and test the visualizations. There is no one correct answer to creating a good data visualization. Reviewing, editing and testing is, according to him, the key to convey much information using visualizations.

While Tufte makes it seem like his approach is a universal method to creating the best information visualization, Lankow, Crooks, & Ritchie (2012) argues that the format of the visualization depends on the objective of it. Tufte comes from an academic background which implies conveying information in an unbiased way. If the goal with the visualization however is to convey a certain narrative this can be reflected in the graphic by illustrating that to the reader. This sometimes means guiding the reader with the use of what Tufte would call unnecessary ink to convey the intended information. Spence (2014) brings up another purpose stating that a good visualization does not only provide an understanding of the information but can also lead to raised questions during the investigation of it. Depending on the goal of the

(16)

10

user and designer a visualization can be made not only to answer questions but also a sparked interest to further explore the information.

Lankow et al. (2012) explains three different objectives of visualization of information; appeal, comprehension and retention. Appeal is defined by catching and maintaining the attention of the reader. This can be done in different ways and often require making the information not only visual, but also visually interesting. Comprehension is, as the name suggest, the amount of information that the reader understands and is able to apply as knowledge. There is no optimal way to make information comprehensive, it depends on the type of information, the user and the context. In general, however, images are better at conveying spatial structures, location and detail while words are better for procedural information, logical conditions and abstract verbal concepts. Retention is the amount of information that is transferred to the long-term memory and therefore can be seen as new knowledge. For the retention aspect some studies show that despite the theories of Tufte, graphical elements that does not themselves convey information can serve to help the reader in the retention process (Bateman et al., 2010). Additionally text can be used to help the viewer in guiding their attention in the interpretation process of visualizations (Alhadad, 2018).

3.4 Interaction in Visualization

Static representations are one way of showing information which is in general more effective than presenting raw data. But every static visualization has a limitation of space making it impossible to show more than a certain amount of information. Adding interaction to a visualization creates new possibilities which can enhance the power of information (Spence, 2014). One example of interaction resulting in completely new information is when a user can through interactions make their way to a new website on the internet. It can also be restructuring or reordering of available information as filtering or sorting a list. Providing the user with the possibility to interact with the data can deliver new perspectives and insights that would never have emerged otherwise. While some interactions can result in a more powerful representation, others can distort the interesting aspects of the data. Which interaction technique is best suited depends on the task at hand and is a question that the designer must answer (Mazza, 2009).

According to Shneiderman (1996) interactions for visualizations can be broken down into three parts, first overview, then zooming and filtering, finally details on demand. The user should first be presented with an overview of the entire collection of data to get an understanding of the dataset. This leads to a shallow understanding of the data and can raise questions that need

(17)

11

further detail to be answered. Then the visualization should provide possibilities to zoom in on specific areas of interest and filter out information that is not of interest. Both of which can ease the cognitive load on the user by minimizing the information presented (Quiroga, Crosby, & Iding, 2004). A deeper examination of the data can lead to more questions and further desire for information. The user should therefore have the opportunity to explore relevant details should they require them.

Well-designed interaction in a visualization can be a means to overcome the information overload problem by letting the user decide how to display information in a relevant way (Mazza, 2009). More freedom is however not always a positive thing, it can make it harder to highlight a feature or distort relevant aspects within the data. There is therefore a need to test any interaction possibilities in the same way as all visualizations to determine whether the interaction possibilities and the data manipulated by them are of use to the viewer.

3.5 Cognitive and Perceptual Heuristics and Biases in Visualizations

People are exposed to a vast amount of information and decisions every day. To be able to manage all that information, we use cognitive heuristics to process it in a fast and efficient manner (Sternberg et al., 2012). A heuristic is a non-conscious cognitive short-cut or rule of thumb that use a practical method to faster process information in problem solving or decision-making. Heuristics has been developed to make the information processing in everyday life more efficient and less cognitively demanding. In most cases heuristics work well and alleviate cognitive load to make faster decisions. In some cases, however, heuristics can lead to biases. A bias is a misinterpretation of information which leads to a non-valid conclusion. Pohl (2016) defines a bias by five criteria:

1) reliably deviates from reality, 2) occurs systematically,

3) occurs involuntarily,

4) is difficult or impossible to avoid, and

5) appears rather distinct from the normal course of information processing.

Dimara, Franconeri, Plaisant, Bezerianos, & Dragicevic (2018) mapped 154 cognitive biases associated with data visualization. In information science and visualizations being aware of cognitive and perceptual heuristics is important to convey information in an efficient way. To be able to manage and make use of heuristics as opposed to fall in their pitfalls requires being aware what they are and considering them in the design and evaluation of the visualization.

(18)

12

3.6 Designing for Mobile

There are differences in designing for mobile and designing for a computer. The different platforms come with different characteristics such as different screen sizes and interaction possibilities. Mobile phones generally have smaller screens and touch-based interfaces as oppose to the pointer interface in traditional computers.

Mobile generally has less input options, missing a physical keyboard with all its possibilities and mouse actions such as hover and double tap being impossible or unusual. This brings different requirements when designing for mobile compared to computers. Because there is no way to explore which elements are interactive and get a guiding indication to what happens as by hovering in a pointer interface, Perea & Giner (2017) states that interactive elements on mobile devices need to be more obvious and the result of the interaction need to be more intuitive. This can be achieved in different ways. There are general design patterns to indicate an element is interactive, text can be underlined, and buttons use a distinct design. These have become ways to help the user in their exploration of a new interface or system. By using these patterns, the designer can take advantage of the prior knowledge of the user and shorten the learning time of the user (Perea & Giner, 2017). When using a system however people adapt to that system which means a designer can introduce local design patterns. This offers new possibilities for the designer to introduce methods to communicate information through the design of elements. This can be done by using affordances. Affordances are inherent perceived action possibilities that an element has to a user (Davis & Hunt, 2017). This can mean an element’s actual perceived possibility to interact with it but also how that interaction happens and what outcome is expected from it. By using affordances, intuitive information can help the user understand the meaning of elements through the design of them. This can be done for example by using metaphors from the real world and taking advantage of the prior knowledge of the user to symbolize information in an intuitive way.

Using icons is one way to make efficient use of the screen real estate. Icons usually take up less space than text. At the same time, a good or well-learned icon can fulfill the same purpose even quicker than text by taking advantage of the pre-attentive image processing of the human perceptual system (Spence, 2014).

Another aspect that is important when designing for touch-based platforms is the location of the interactive elements. Because the whole screen is a possible interaction point, it is a big area to interact with, especially for one-handed use. This means that consideration should be put into

(19)

13

the placement of interactive elements (Hoober & Berkman, 2011). In a touch interface the input device is the fingers of the user. By putting the interactive elements easy to reach it will minimize the travel distance making it easier and faster to press which can make for a better user experience. When designing the formfactor of today’s phones this means it is generally good practice to put the interactive elements on the bottom half of the screen so that it is easy to reach for both left and right one-handed use (Hoober & Berkman, 2011).

(20)

14

4 Method

The study took a mixed methods approach collecting data with interviews, a survey and evaluation of visualizations. Below the methods used to investigate the research question are explained. The process and result of the methods used are presented in the next chapter, Process

and Results.

4.1 Interviews

Pettersson (2012) points out the importance of knowing the receiver of the information to be able to communicate it accurately and effectively. Generally, a wider audience means more individual characteristics and less uniformity which affects the content of the message. The more that is known about the target group the more specific and streamlined the message can be. In order to shape the message after the group an understanding of it must be gained. Firstly, the target group needs to be defined and data, including age, culture and other factors must be collected. Everyone will interpret a message differently and these factors can provide an understanding of the commonalities within a group which can be of assistance in the design process. Another factor which should be considered is any previous interaction with, and feedback from, the target group.

The context in which a message is produced and received also influences how it is interpreted (Pettersson, 2012). This means that the sender needs to have an idea of both internal and external factors of how the message will be received. The external factors refer to the context in which the message will be received which can include social context, through what platform and in what kind of situation. The internal factors refer to the state of mind of the receiver. This is in most cases impossible to know but knowing the audience can provide guidance and clues to this.

According to Blandford (2013) interviewing is a good method to get a deep understanding of personal opinions and experiences. The opportunity to ask the same questions to different people offers a chance to display similarities and differences across a group. The method was therefore used to get an understanding of the users of football statistics today and the context of use.

An interview can take a structured, semi-structured or unstructured approach (Blandford, 2013). Blandford emphasizes the importance of preparations before any interview. In structured interviews an interview guide should be created to keep all interviews to a standard and not forgetting any important aspects. During the interview the questions should be formulated as

(21)

15

open as possible to not influence the answers of the interviewee which according to Brinkmann (2014) is important to get a diverse and detailed understanding of peoples experiences. Brinkmann also emphasizes the importance of being able to steer the interview to the topics of interest. Having an open approach to the interview and the topics covered in it can lead to the discovery of completely new areas information and knowledge. Follow-up questions, not included in the interview guide, are therefore an important tool to find discover new information. According to Lapan, Quartaroli, & Riemer (2012) this is an efficient way to cover topics that were not planned in advance or for some reason not covered in the interview guide.

4.2 Survey

A survey was conducted to examine which football statistics are the most interesting to enthusiasts. Djamba & Neumans (2002) explains that a survey is an effective way to collect data about specific questions from a specific group of people. It can for example be an effective way to learn about the need and demand of a group of people. When conducting a survey, one must first define the purpose and objective and what sample to use. After which you have to decide which data collection method to use. According to Showkat & Parveen (2017) questionnaires and interviews are the most common methods for collecting data. Questionnaires do typically not provide a rich data set since the questions are mostly close-ended or narrowly framed. Interviews generally provide richer data since there is opportunity to ask open-ended and follow-up questions. Which method to use depends on the purpose of the survey.

4.3 Evaluation

The main purpose of visualizing information is to bring new insight to the intended audience. A visualization can however not only be assessed by its ability to bring insight to the user, in practice, it must do so in a timely and satisfying manner. In order to determine whether a visualization fulfills these requirements it must be evaluated (Santos, 2005). Mazza (2009) identifies five different criteria that can be evaluated in an information visualization.

Functionality: does the visualization provide all the functionalities requested by the users. Effectiveness: does the visualization provide better knowledge than the raw data and sufficiently good knowledge to fulfill the users’ goals.

Efficiency: Does the visualization provide the user with knowledge in a timely manner. Usability: Is the interaction in the visualization intuitive and simple enough to fulfill the users’ goals.

(22)

16

Usefulness: How and when does the information visualization help fulfill the goals of the user. Depending on the goal of the study or visualization one or several of these aspects can be of focus in an evaluation.

The field of human-computer interaction has developed a lot of evaluation methods which can be used in design studies and in the evaluation of visual representations. Evaluating products intended for a specific user is done with usability testing both in the industry and academia. There are a variety of methods both for collecting quantitative and qualitative data and which one to use depends on the research question at hand (Rubin & Chisnell, 2008). Generally, quantitative data are used to get concrete answers to concrete questions, for example determining which out of two designs is the best. Qualitative data offer more insight about how a product helps fulfill the goals of a user, what problems there are in the design and possible improvements.

This is not a pure information visualization study neither a usability study of a product, but a lot of methodological inspiration for the evaluation process come from these fields. The goal with evaluating the design suggestions in this study is to get an understanding of which visualizations are effective, satisfying and interesting to the users. Based on that the evaluation took an exploratory approach inspired from usability testing. Exploratory studies are conducted early in the product cycle to reveal general insights about the impression and opinions of a product and if it fulfills the goals of the user (Rubin & Chisnell, 2008). According to Ellis & Dix (2006) evaluations that do not seek for specific knowledge can lead more comprehensive and effective results and generate more generalizable knowledge. An exploratory approach was therefore used since the primary purpose of the evaluation was not to achieve a seal of approval or to improve the design suggestions themselves, but to gain knowledge and insight of how to design football statistics.

Qualitative data collection methods like interviews and observations are commonly used in an exploratory evaluation. It gives the researcher opportunity to ask the participants questions to gain knowledge about how they interpret the design and if they understand the information presented (Rubin & Chisnell, 2008). It also gives the opportunity to explicitly ask the user if and how a visualization provides value to them (Mazza, 2009).

4.4 Thematic Analysis

The data collected in the interviews was analyzed using thematic analysis. The purpose of thematic analysis is to find and explore themes in a subject connected to a research question.

(23)

17

Themes can be problems, opportunities, phenomena or other patterns which can be observed in a normal situation (Blandford, 2013). There are different explanations and descriptions of thematic analysis and the popularity of the method in qualitative research has increased in recent years. According to Ryan & Bernard (2003) there have been many words for themes but all refer to the same thing. A theme has been identified if you can answer the question “What is

this an example of?”. They continue by stating that themes can take many different shapes,

some are wide and generalizable while others are more specific and answer binary questions.

The analysis in this study followed Braun & Clarke’s (2006) approach which consists of six phases.

1. Familiarizing with the data. The first step is to get familiarized with the data. This is done during the collection itself, in the transcription process and during the initial analysis. The process involves going through the data multiple times to be familiar with the depth and breadth of it. During this phase preliminary notes should be taken which will then be revisited and revised during the rest of the analysis process.

2. Generating initial codes. During this phase aspects of the data that seem interesting to the analyst are highlighted. This can be features that seem meaningful in any way to the research question and this phase usually results in a vast amount of codes. Some of the codes of this phase will be novel, others reoccurring and some contradictory, this is normal and desirable to get a complete picture of the dataset.

3. Searching for themes. When the entire dataset has been coded the search for wider themes begins. The codes generated in the previous phases are sorted into potential themes which gives an overview of the data. This is done analyzing and sorting the codes into overarching themes that tell a clear story of the dataset. The process can be simplified by using physical extracts of the data to give a visual representation of the data. In this phase codes may go into main themes and sub-themes. The codes that does not seem to fit into any theme but are still judged to be of interest to the research question are put into a separate category. The phase results in a set of themes, and sub-themes where all codes are included.

4. Reviewing themes. Once the dataset has been coded and sorted into themes, the themes must be reviewed. During this phase each theme is assessed to see if they are relevant to the research question and backed by enough data to be valid. The data within a theme should be coherent and yet clearly distinguishable from other themes. After this phase there is a clear picture of themes, how they fit together and the story they tell.

(24)

18

5. Defining and naming themes. After having a clear picture of the themes, they need to be clearly refined and defined. This is done by once again going over the data in each theme to determine the essence of them and naming them appropriately to convey the data in them.

6. Producing the report. In this phase the themes are structured in writing with the goal to concisely, coherently and interestingly convey the data to address the research question at hand. The goal is to in a convincing way, explain the conclusions that have been drawn, how they have been reached and how they are relevant to the study.

(25)

19

5 Process and Results

The process of this project followed an iterative approach with 5 different data collection phases. Firstly, interviews were conducted with two experts who had worked with football statistics in media for several years. Secondly, based on the insights gained in these interviews a survey was conducted to investigate which football statistics were of most interest to enthusiasts. Thirdly, eight interviews with consumers of football statistics were conducted to understand the needs and requirements of the users. This laid the foundation for 24 design suggestions which finally were evaluated in two phases with the same 8 participants that had been interviewed earlier in the study. Figure 2 is an overview of the data collection process and each phase is explained in more in detail below.

Figure 2 - The data collection process.

5.1 Expert Interview

The interview with the two experts took place at their place of work. The author of the study and both participants were present. The interview was conducted using an interview guide that had been created in advance (see appendix 1). The purpose of the conversation was to take advantage of their experiences and knowledge within the field. The goal was to get an understanding of in what direction the commercial use of football statistics has evolved and will continue to change in the coming years. Additionally, the goal was to learn what content enthusiasts are interested in about football statistics and in what way it should be presented to convey the information correctly.

5.1.1 Participants

Two experts who was working with football statistics were recruited for an interview. Both participants had extensive experience working with statistics targeted at football enthusiasts within TV and some experience with web-based content. The participants were recruited based on their experience with TV and were regarded as experts within the field of football statistics targeted at enthusiasts. The participants had no experience with working with interactive

(26)

20

content. Both participants gave their consent to take part of the study and for the interview to be recorded.

5.1.2 Results

The analysis of the interview was done using thematic analysis. The methodology followed Braun & Clarke’s (2006) approach explained in more detail above. The themes identified are presented with a brief explanation, how they are relevant to the research question and extracts from the interview that demonstrate instances of the theme in the data. Four themes of general learnings that the participants have gained during their work in the field were identified in the interviews. The themes serve as an introduction to the field for the author of the study and as a foundation for a survey with the purpose to determine which statistics people are most interested in.

Be clear about what the statistics are based on and transparent with the limitations of the method

One thing that the interviewees mentioned several times was the importance of transparency of the data towards the viewers. To make people understand and accept the statistics they must understand what raw data the they are based on, what valid conclusions that can be drawn and the flaws of the method. This means description of the raw data and use of it should be included in the visualization or at least somehow available to the consumer.

- “you need to be trustworthy and in our experience the best way to be that is to be

transparent about everything, if we question something the viewers will as well”

Do not overestimate the knowledge of the viewers

Even after having broadcasted over 50 episodes of a program about football statistics the interviewees said they do not take any prior knowledge from the viewers for granted when presenting new statistics. They had noticed that people need a detailed explanation to be interested in and to understand the statistics.

- “We don’t take any prior knowledge for granted with our content”

After accepting a concept, the viewers are willing to accept more complex concepts Even though no prior knowledge should be assumed, people that have accepted some statistical concepts are more prone to accept more complex ones. Indicating the statistics should be presented in a hierarchal order with less complex concepts first leading into more complex ones.

(27)

21

- “Now we can do more complex things than we did in the beginning… because the

viewers have gotten use to the concept”

Content can be seen from four different perspectives, player, team, league, match.

The interviewees repeated multiple times that they always organize statistics after one or multiple of four categories, player, team, league, match. This provides a clear angle and simplifies the communication of the statistics.

- “We structure everything in one of the four categories to have a clear structure and

objective with showing the stats”

5.2 Survey

Based on the insights gained in the expert interviews and statistics traditionally used in football, a survey was conducted to see what statistics enthusiasts are interested in. The main purpose of the survey was to examine what type of data football enthusiasts are interested in. Furthermore, the purpose of the questionnaire was to decide which data that was going to be visualized and evaluated. In this way the survey laid the foundation for the design suggestions and subsequent evaluations. The questionnaire focused on questions about which statistical measures enthusiasts find interesting in four different categories; player, team, league and match. For each category the participants were given 9-13 alternatives of statistical measures which they were asked if they find interesting or not. All survey questions are presented in appendix 2.

The survey was conducted using Google forms. A total of 50 people answered the questionnaire. The participants were collected using social media platforms where the survey was posted with an explanation stating that anyone interested in football was welcome to take part. The survey was restricted to be answered only once per person.

5.2.1 Results

Four bar charts with the result in each category were created and are presented below in figures 3-6. The top three most frequent answers in each category have been highlighted to indicate which data was chosen to be visualized in the respective categories. Each chart is presented below with a short explanation of which aspects were most frequently answered in each category. Finally, a summary of all charts with general conclusion follow below.

(28)

22

Figure 3 – Results of what kind of statistics football enthusiasts are interested in connected to individual players.

The results from the question about what people are interested in about football statistics about individual players can be seen in figure 3. The two most frequent answers were “How important

the player is for the team” and “How many goals a player has been involved in” with 35

answers each. “How a player performs compared to others who play in the same position” was the third most frequent answer with 32.

(29)

23

The results from the question about what people are interested in about football statistics when it comes to teams can be seen in figure 4. The far most popular answer was “The

strength/weaknesses of the team” with 44 answers. The second most frequent answer was “the performance of the team in each game” closely followed by “What type of players/which player the team should buy” with 31 and 30 answers respectively.

Figure 5 - Results of what kind of statistics football enthusiasts are interested in connected to a league.

The results from the question about what people are interested in about football statistics about teams can be seen in figure 5. The most frequent answer was “Which player is the best in

different categories” with 39 responses. The second most popular aspect answered 33 times

was “In what categories your team is performing better compared to others”. “What players

in the league produces the most goal chances” was the third most popular answer, given 29

(30)

24

Figure 6 - Results of what kind of statistics football enthusiasts are interested in before/during/after a match.

The answers of what kind of what statistics are interesting connected to a match can be seen in figure 6. Four aspects stuck out from the rest, “Ball possession” got the most answers with 31, closely followed by “How did the game play out” with 30. On a tied third place was “Who was

the best/worst player of the game” and “Prediction of the game” with 29 answers each. After

having assessed the data available a decision was made to exclude prediction of the game and include best/worst player for visualization and further evaluation. This was done because the data available for best/worst player was regarded as comprehensive and to keep the statistics to three per category, and decisions was based on the judgement by the author that the data available for best/worst player was more comprehensive and because of that would provide more value to the users.

The purpose of the survey was mainly to indicate which aspects of each category enthusiasts are most interested in and to decide which visualizations to create. The results of the survey do however offer some more general insights. Unsurprisingly, the answers indicate that the interest of the participants is skewed towards information about the teams they are supporting. Furthermore, the results indicate that the participants want to know positive information more than negative, meaning they rather know who are performing well than who is not. This can be seen in the chart about teams and the questions regarding “which player performs the best/worst

in different categories” where the positive version of the same question has 29 answers and the

(31)

25

skills is apparent. This can be seen in league and player charts where one of the top three answers are about goalscoring (“How many goals has the player been involved in” and ”Which

player has produces the most goal chances”).

As stated, the main purpose of the survey was to decide which information was most interesting for football enthusiasts and to serve as an indication and motivation for which visualizations that were going to be created and evaluated with potential users. The top three answers within each category, player, team, league and match answers were chosen to be visualized and be incorporated in a prototype of a design of a smartphone application.

5.3 Interviews

The goal with the interviews was to establish the requirements users have on an app about football statistics. The interviews were conducted one on one with eight participants. Each interview followed an interview guide that had been created in advance (appendix 3). Each interview lasted about an hour. Below follows a thorough description of the participants and sampling process followed by the results. The interviews were conducted in Swedish, the extract from the data presented in the result section have been translated into English.

5.3.1 Participants and sampling

For the interviews with the potential users a total of 8 participants were recruited using a purposive sample. The sampling of the participants was based on a mapping of users following @ChampionsLeague on twitter (figure 7) (Cipolla, 2017). Analysis of the audience showed that 86 % of the followers were men and 14 % women. Regarding age, almost 47 % were between the ages of 25-34.

Figure 7 – Demographics of twitter users following @championsLeague

Participants were recruited based on these demographics, with the assumption that the potential users of a mobile application based on football statistics will be football enthusiasts and frequent

(32)

26

users of technology. The Champions League following on twitter was therefore seen as a good indicator of the target group and thus of a good sampling for interviews.

A total of 8 participants were interviewed. According to Corbin & Strauss (2012) there is no exact number of participants which will always fulfill the goal of a qualitative data collection. The number of participants was therefore decided while the study was conducted and limited when the author experienced that a saturation in the data had been reached in accordance with the recommendations of Carpendale (2008).

5.3.2 Results

The thematic analysis of the first interview phase is presented below. The analysis was done using Braun & Clarke (2006) approach. A total of four themes were identified and divided into requirements and goals (table 3). A requirement is defined as a need or desire that the participants expressed from a service about football statistics and therefore something that should be considered in the design. A goal is defined as a task or an objective that were identified in the data and something that a service about football statistics should provide a solution for. A total of four main themes were identified. Within each theme a few subthemes were identified. The themes requirements and goals are presented below with a short explanation of the theme, how they are relevant to the research question and extracts from the data that demonstrate instances of the theme in the data. which serves as more concrete explanations of the respective theme and what should be considered in a design implementation.

Requirement Goal

R1: To have a good understanding of what the analysis means and what data it is based on

G1: To make objective conclusions about football supported by facts

R2: Focus on interesting and relevant aspects of the data

G2: To keep track of a player or a team

Table 1 – The themes identified in the interviews. 5.3.2.1 Requirements

Below follow the two requirements identified in the data that were categorized as requirements that the participants had about football statistics a long with a number of subthemes within each. A requirement in this context is defined as a quality that the design should have to provide a satisfying user experience.

(33)

27

R1: To have a good understanding of what the analysis means and what data it is based on

Something that was apparent in all interviews was the participants’ desire to understand the analysis that has been made to come up with a statistical measure. The purpose of this is to be able to assess how reliable and relevant the data is. This also means that they need to have an idea of what data the analyses are based on. Furthermore, one point that a few of the participants brought as an important factor in understanding statistics was the access to comparison measures. All of these points have been identified as a general theme about the understandability of the statistics but are explained in more detail in subthemes below.

R 1.1: Insight and understanding of what is behind the numbers

Almost all participants raised their interest of knowing the statistics and analyses behind the numbers. Many of them accept statistics even if they do not know how it is calculated. They did however request a better understanding of it to be able to evaluate how reliable and relevant they are. Not a single participant had knowledge about how the data is collected but they did express interest in knowing it. Regarding the statistical measurements typically used in football they had ideas about how they were calculated but trusted them even if they did not. Concerning more innovative statistics however, the participants were more skeptical about the data and interested in the numbers behind it.

“I do trust and take for granted the possession numbers because it has always been there. When it comes to expected goals on the other hand I’m more skeptical about what it actually says.”

R1.2: Transparency with the data, what it means and problems with it

Several of the participants are skeptically minded towards some statistical measures thinking that they do not reflect reality and fail to capture the context of a game. This does not mean they are uninterested in statistics. It does however mean they ask for insight about which parameters that are used and how they are weight against each other. The purpose of this is to get an understanding of the underlying data and what conclusions can be drawn from it. There is therefore a need for more detailed knowledge about the data and transparency about its strength and weaknesses.

“With expected goals for example, I accept it and trust it but I could not explain it to someone else and would like to know how it is actually calculated”

(34)

28

Another point connected to the data and visualizations which several of the participants highlighted was the importance of comparisons in the statistics. To understand what the numbers mean there is a need for comparisons. The numbers rarely say much in them of themselves but need to be put in a context for it to convey interesting information.

“I love to compare stats of teams between different seasons, just yesterday I looked how Liverpool’s season is comparing to the invincibles of 03/04… That’s something I can do for hours”

“…It is not until you see everyone else’s stats that you realize how good Messi actually is”

R2: Focus on the interesting and relevant aspects of the data

In the analysis of the data it became obvious that some aspects of statistics are of more interest than others. Of course, not every statistic is regarded as equally interesting and there are also ways to present the information in that makes it more engaging and interesting.

R2.1: Adapt the information presented after the player/position

In general people want to compare different players and teams to each other. At the same time however they pointed out how different aspects of the game are interesting for different positions. A defender should not be evaluated by their goalscoring in the same way as the number of interceptions does not necessarily say anything about how good an attacker is. Different players and positions have different purposes meaning they also should be evaluated based on these purposes.

Another example of this is how many of the participants said they are most interested in the facts and numbers that stick out of the ordinary. having some sort of standard for what to compare is therefore desirable, at the same time however the most interesting and engaging content can be to highlight the most extraordinary aspects of a player’s game.

“…this is especially hard with goalkeepers I mean it is like they play a completely different sport… But in some cases I feel like it is almost the problem to compare defenders and attackers”

R2.2: Focus on the positive aspects

In the interviews there was a bias of interest towards the positive aspects of performance of players and teams. Several of the participants said that they are more interested in positive information than negative, meaning they want to know who has performed well as oppose to

(35)

29

who has not. This is also indicated in the answers of the survey questions about the teams and the questions regarding which player performs the best/worst in different categories” where the positive version has 29 answers and the negative only has 16. Which further indicates that positive aspects should be emphasized when choosing and designing football statistics.

“I always want to see who is leading the scoring and assists leagues”

“I usually look up the best players in the worse teams to see who is the next star”

R2.3: Preference for attacking and offensive aspects of the game

Something else that became apparent in the interviews was the participants’ preference for offensive aspects as oppose to defensive ones. Several of the participants were openly more interested in goals and assists than the defensive counterparts. In general, there was more focus on the offensive parts of the game even though the interviewer did not steer the conversation in that direction.

“But it is not so weird that they [attackers] are more praised than defenders. Honesty it is much more interesting and easy to track goals and assists than tackles and interceptions because they are so concrete”

5.3.2.2 Goals

G1: To make objective conclusions about football supported by facts

During the interviews it became clear that everyone uses statistics differently, however, one goal with using statistics was the same across all participants. Everyone used it and thought it was interesting to provide a more nuanced and objective understanding of football.

G1.1: Predicting or explaining a performance or result

One goal with using statistics was common among all participants. Everyone expressed a wish to predict a result prior to a game, either because of pure interest, to facilitate discussions or to assist in betting. If the participants do not have chance to watch a game of their interest, they use statistics to get an explanation of the result and an idea of how the match played out. Several of the participants also expressed a demand for more detailed, objective and reliable statistics to fulfill this goal. This goal is not limited to single matches but also relevant to get an understanding of the performance of players and teams over time.

“That is why statistics is interesting, to get an idea of how something played out. Either before or after the game as a predication or an explanation for the outcome”

What&apos;s in the Game? : An Exploratory Design Study about Football Statistics