<?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 Chris Carling on Medium]]></title>
        <description><![CDATA[Stories by Chris Carling on Medium]]></description>
        <link>https://medium.com/@ccarling_82663?source=rss-7663ba8e7a63------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*dmbNkD5D-u45r44go_cf0g.png</url>
            <title>Stories by Chris Carling on Medium</title>
            <link>https://medium.com/@ccarling_82663?source=rss-7663ba8e7a63------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 09:44:51 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ccarling_82663/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[FOOTBALL DATA & TECHNOLOGY FOR MATCH ANALYSIS]]></title>
            <link>https://medium.com/@ccarling_82663/football-data-technology-for-match-analysis-a19db1a788a4?source=rss-7663ba8e7a63------2</link>
            <guid isPermaLink="false">https://medium.com/p/a19db1a788a4</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[sport-analytics]]></category>
            <category><![CDATA[football]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Chris Carling]]></dc:creator>
            <pubDate>Thu, 08 Aug 2019 10:25:38 GMT</pubDate>
            <atom:updated>2019-08-08T10:25:38.985Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/910/1*5dmVlaymJ0jQdiGAa3LU6g.jpeg" /></figure><p><strong>1 Introduction</strong></p><p>Elite team sport companies invest substantial resources to analyze the own team’s performance as well as the performance of the opposing teams. Various aspects and several data sources are important descriptors for these performances. Depending on the available data, different analysis tasks can be executed. Analysts usually do not only want to have information about the what (e.g., “Team A won against Team B” or “Player X passed more often than player Y”) but instead want to investigate the why behind these facts. There is a need to understand why a certain movement happened and what the influencing factors were [20]. In football, tactical skills and cognitive processes underlying decision-making are considered essential requirements for excellence in sports performance. During a match several situations whose frequency, chronological order and complexity cannot be predicted emerge, demanding great capabilities of adaptation and immediate response from the players and teams based on the concepts of opposition that are present in each phase of the game [9] (1997). These situations of opposition are so evident in a football match that one can observe tactics through the spatial organization of the players in the field in view of the match circumstances with respect to the movements of the ball and the alternatives of action of both teammates and opponents [6] (2007). This form of understanding tactics gives special emphasis on movements and positioning within the field highlighting players’ ability to occupy and/or create free spaces according to the most suitable tactical principles for that given moment. Tactical principles are defined as a set of norms about the game that provide players with the possibility of rapidly achieve tactical solutions for the problems that arise of the situations they face [8] (1994).</p><p>Because of such features, tactical principles need to be understood, classified and presented within teams, lines and players’ behaviors during a match, so as their application facilitates achieving objectives that lead to score or avoid goals. Individually, the application of tactical principles highlights the player’s decision-making, his ability to cover the field, to pass and receive the ball, to apply and withstand pressure. About lines, it is fundamental to maintain the team balance in response to changes in the formations, coordinating the movements with teammates, on the other hand the ability to penetrate opponents playing between the lines and destabilize the organization (positional play). Collectively, the application of the tactical principles help the team to better control the game and the pitch coverage, to maintain ball possession, to perform variations in ball circulation to limit pressing, to change the pace of the game and materialize tactical actions aiming to break up opponents’ balance and, consequently, to more easily score a goal [24] (1980), [1] (2006). Therefore, the more adjusted and qualified the application of the tactical principles is, the better the tactical performance of the team or player will be in the match. Given the importance of the tactical principles for the organization and performance of the team within the field of play, [4] aims to contribute to the definition, in both the conceptual and operational level, of the tactical principles of the game of football, to propose the addition of two core tactical principles and highlight their practical applications in the offensive and defensive phases of play [4]. These principles can be traced back to analysis to study both teams and players performance.</p><p><strong>2 Match Data</strong></p><p>In the world of professional football, performance analytics about a player’s skill level and the overall tactics of a match are supportive for the success of a team. These analytics are possible thanks to the large amount of data that is increasingly available in the last decade. Elite football team typically have events data about the game on the one hand and positional data about the game on the other hand. Starting from these assumptions, the rise of big data technologies is shaping the future of performance analysis in elite football.</p><p>2.1 Events Data</p><p>Sport games can be described by an ordered sequence of events. We define events as actions being match relevant and happening during the match. Events typically are derived from movement data by automatic video analysis; also manual annotation is possible and done professionally by some data providers. This operation is time-consuming and error-prone. From a technical perspective, events are timestamped occurrences of previously known and defined categories optionally annotated with spatial coordinates or additional information as involved players. Most events will be directly ball related and corresponding to actions with the ball (for instance passes or dribbling). Other events may be time-dependent (e.g., start and end of a play period) or not directly dependent on the ball (e.g., a foul situation during a free kick). A multitude of methods to detect such events exist (in fact too many to list them here) ranging from manual annotation to fully automated systems [22], [23]. In practice, events might lack in accuracy, as they are usually annotated manually or as fully automatic recognition may produce false positive and negative events. The company Opta, for example, is commercially providing event data. As event data contains mostly information about players interacting with the ball, event data enables 2 us to conduct overall game statistics (e.g., passing networks, pass accuracy, or time between gaining the ball and shot on target) [20].</p><p>Opta [16] has a long-established, consistent list of ‘Event Definitions’ in football that are adhered to across all data collection centers and that is recognized as de facto standard. These data are collected in an organized way and this helps the creation of datasets useful for diverse research fields such datamining, sports analytics and continuous preference queries. The F24 data is a time coded feed that lists all player action events within the game with a player, team, event type, minute and second for each action. Each event has a series of qualifiers describing it.</p><p>2.2 Positional Data</p><p>The positional (tracking) data of the ball and players is tracked automatically by cameras or via sensors [17]. Locations are usually measured by local coordinate systems with reference to the game pitch. These measurements contain as a minimum the x- and y-coordinates and sometimes also the z-coordinate. There are different providers all over the world. Here we list that are most used. The STATS SportVU tracking system [18] delivers performance statistics by extracting and processing coordinates of players (X,Y) and the ball (X,Y,Z) through HD cameras as well as sophisticated software and statistical algorithms. STATS SportVU’s speed, distance, and possession data provide key insights and analysis points for professional football clubs. Unlike other tracking systems, STATS SportVU can track every player on the pitch, not just those on one team. ChyronHego [3] offers two player tracking solutions, optical and wearable, comprised of three different categories of data collection methods — Optical, GPS, and RF. ChyronHego is the only provider to offer all three types of data collection, and in a manner that uniquely enables a common, sharable dataset across our customer’s sports tracking ecosystem. The last innovation introduced in the market by ChyronHego is TRACAB Gen5. It uses a distributed system of cameras installed around the field of play and advanced image processing technology to capture and deliver real-time tracking data on the movements of each player, referees, and the ball. TRACAB Gen5 includes an all new distributed camera architecture in which cameras can be deployed on both sides of the field and behind each goal to capture action from four angles. With the flexibility of the system offering multiple camera views, Gen5 is able to track every object on the field at a much higher resolution. The accuracy of player identities is further enhanced through TRACAB Gen5’s AI capabilities, which enable the system to recognize and distinguish player numbers and jersey colors from any angle. From a typical football game, the positions of the 22 players and the ball can be recorded 25 times per second, resulting in approximately 135.000 datasets. Without computational assistance it is almost impossible to extract relevant information from the complete data [10].</p><p>3 Data Modeling &amp; Classification</p><p>In order to produce advanced analyses, the first task to accomplish is data modeling. Generally speaking, modeling is about giving structure to the problem of sport analysis, by prescribing which aspects are of importance to the analysis. Positional data is the starting point. These data can be eventually coupled and integrated together with events data, in particular to detect events like cards that are not recognizable from positional data only. We call elementary events something that happens, or is regarded as happening, describing the behavior of a player or group of players. It is possible to classify elementary events, using simple geometric rules or more advanced machine learning techniques based on neural networks. The first approach is more immediate to implement but more sensible to the tracking data provider: i.e. if the tracking system does not track the ball directly, the ball trajectory is linear during transfers then passes are identified simply; if the tracking system tracks both the ball and the players, the transfers are more difficult to detect since the trajectory is trembling. The second approach is more robust but it needs the positional data of thousands of matches, which are not available to the major part of the researchers. Both of these approaches allow to get a structured description of play in elementary events — such as conduction, passes, crosses, shots, goals, throw-ins, goal kicks, tackles — but also a description of movements of player or group of players. 3.1 Geometric Approach Tovinkere [22] shows an excellent example of the first approach. It presents a method for detecting semantic events which may happen in a football game. Their method requires the position information of the players and ball as input. Based on the tracking data, the method detects a wide range of player actions and game events using a set of heuristic rules which are derived from a hierarchical entity-relationship model representing the prior knowledge of football events. They have applied the proposed method as a key enabling module in a real-world sports entertainment application and the experimental results are given in the paper. The proposed event detection algorithm detects basic football actions and uses the detection result as the basis for detecting more complex football events. Once an event is detected and marked as valid, the algorithm will then invoke another set of heuristic rules to determine the specifics of the actions involved in the event. For the sake of convenience, they call semantically significant actions and consequences as events and simple physical motions as actions. We further differentiate the events as observed and interpreted ones. An observed event is an independent event which can be readily observed, e.g., a kick or a goal. An interpreted event is an annotation of an observed event, e.g., a shot-on-goal may be essentially a kick with scoring intended. The list of observed football events includes: ball-out of bounds, corner-kick, dead-ball-by-ref, deflection, drop ball, free-kick, foul, goal, goal-kick, hand-ball, injury, kick-from-penalty-mark, kick-off, offside, penalty-kick, reception, red-card, steal, substitution, throw-in, turnover, and yellow-card. The list of interpreted football events includes: assist, block, header, interception, save, and shot-on-goal. The list of football actions includes: collapse, defender movement (tackle), detection, dive, dribbling, drop-ball, fall, getting-up, hand-motions, holding-the-ball, hug, kick, jumping, lying-on-the-ground, place-ball-on-the ground, posture, reception, rolling, running, standing, sliding, stopping (halting), throw, and walking. This framework loads the domain specific knowledge at startup and the actual detection is performed in two phases. The primary goal of the first phase is to compute the derived information such as player motion and orientation that is required by the rule-based algorithm and identify all sections of the tracking data that contain player-ball interactions. This is accomplished by eliminating deflections that are associated with the ball bouncing of the ground or the goal post. The second phase determines which rules from the domain knowledge base are appropriate to be applied for each segment marked as a potential deflection by a player. These rules evaluate the game situation and execute rules that are relevant only to the current situation. These segments are then labeled as valid or invalid events depending on the outcome of the rules evaluated. This phase can include other methods of inference that act independently or in conjunction with the existing rule-based implementation.</p><p><strong>3.2 Machine Learning Approach</strong></p><p>Irnai [13] gives an example of the second approach. The method allows to recognize the labels of plays in football games based on trajectories of the ball and players. They use tracking data measured by a camera-based system from ChyronHego called TRACAB [3]. Their method aims to recognize ten types of plays with ball touch such as pass, trap, and shoot that are actually recorded in J. League games. For this purpose, there are several challenges for automatically recognizing plays. Firstly, camera-based tracking data includes some position errors due to inaccuracy of the image recognition. In particular, these errors occur frequently in situations where some players gather around the ball, the ball moves out of the pitch, and so on. They design a method that avoids such errors of tracking data by combining carefully designed features. Secondly, there are several plays whose definition are ambiguous among target classes. They tackle with such difficulty by applying machine learning for play recognition. First, they focus on the fact that the ball’s trajectory or speed changes greatly at the moment of the play occurrence. Based on this fact, they detect the occurrence of plays in the tracking data with a rule-based method. Second, they extract the features for each detected play from trajectories of the ball and each player and train a classifier of plays by machine learning. For machine learning, they compare three types of algorithms: Conditional Random Field (CRF), Support Vector Machine (SVM), and Random Forest (RF), finally they show and compare the results. Some papers introduce approaches to identify compounds events by analyzing the positional data of football matches, training machine learning algorithms such as Support Vector Machine, K-Nearest Neighbors and Random Forest, based on features, which are calculated on the basis of the positional data [17], [21]. Others contributions introduces a hierarchical architecture of artificial neural networks to find tactical patterns in those positional data. The results from the classification using the hierarchical setup were compared to the results gained by an expert manually classifying the different categories. Short and long game initiations can be detected with relative high accuracy leading to the conclusion that the hierarchical architecture is capable of recognizing different tactical patterns and variations in these patterns [10]. In view of all this, a mixed approach, using both geometric and machine learning techniques, is the less prone to errors and give the most complete overview of all the events occurring during a football match.</p><p><strong>4 Match Analysis Tools</strong></p><p>Match preparation is the preeminent key to winning. On average, it takes 25 hours over the course of three days using video from five previous matches. It is an overburdening, subjective process for even the best-equipped backroom staffs. The widespread adoption of information technology and the growing availability of data opens up myriad possibilities; we are only at the beginning of the era of match analysis supported by automatic tools. Different computer-based approaches try to extract and analyze tactical events and patterns in sport games. Many tools and software have been developed and commercialized to provide tactical insights to coaching, helping them in the decision-making process. These tools are usually automatic or semi-automatic and provide information to manage and monitor the performance of the players, the lines, then the entire team. The insights try to monitor as much variables as possible to describe the play in order to optimize the team effectiveness, finding relations among variables and understanding which ones are correlated with the strengths and weaknesses of the own team and the opponents (in statistics, correlation is a relationship between two variables such that each value of the first corresponds to a value of the second, following a certain regularity). The output of these tools is often summarized in digital reports that are available before and after the match. After the 2018 World Cup, following on from the International Football Association Board’s decision to approve hand-held technologies on the bench, world governing body FIFA is allowing to receive real-time support from an analyst throughout the duration of games. The software and infrastructure FIFA put in place for the 2018 World Cup allowed all teams a stable and secure connection between the match analysts located in the stands to the technical staff on the bench pitch side, as CNN reported [7]. By providing a dedicated platform, each team that chooses to use the system will have exactly the same capacity to provide visual and statistical data that can inform in-game decision making throughout the tournament. Sitting in the media section of stadiums, analysts will be able to track players’ movements from two optical cameras and send coaches annotated video clips on the team’s performance. This information would then be sent to coaching staff through a tablet device, allowing the two to communicate with each other. FIFA relies on a visual-tracking technology called Matrics to provide extensive data set in real time that makes up the on-site heat maps, passes completed and distance covered. The company behind Matrics is an Italian firm called Deltatre that uses several technologies and manual inputs from a large crew to deliver the real-time stats. This technology has been around for years, but it’s becoming more diverse, detailed and accurate [5]. OptaPro’s suite of products and services enable clubs, leagues and federations to make more informed decisions, applying data and analytics to enhance analysis and recruitment. ProVision allows to personalize metrics to analyze players, to have insights into playing styles, quality of chances and on-field decision-making. ProPortal is a live data application, providing minute-by-minute analysis to quickly identify trends to help inform in-game decision making through visualizations, graphics and tables. It helps for monitoring the team performance in an individual match and during the season, to support clubs identifying longer-term performance trends. Another important player in the market is STATS [18] with various products. In particular, STATS Edge is a revolutionary search and analytics application that integrates video and data in a simple and intuitive interface. By leveraging artificial intelligence, STATS Edge allows coaches and analysts to quickly find clips and analyze complex patterns within a game, streamlining the process of assessing a team’s strengths and weaknesses. In general, many descriptive, predictive, and prescriptive analytics are feasible over individual players or teams and there are a bunch of tools developed starting from academic publications. Tovinkere [22] developed an application delivering 3D graphical simulation and novel statistics of actual football games over the internet. The system first tracks the positions of the players and ball during the game based on either video input or microwave sensors. Then it detects the player actions and game events using the event detection algorithm with certain amount of human interference. The results are a) an XML representation describing all the events happened during the game and b) an animation sequence containing the position and action information for graphical rendering. The event and animation information is packaged and uploaded to servers from where an end user can search for interesting game. In the data mining field, datasets can help for the validation of various techniques aiming the detection of temporal patterns and key events. As the dataset has the player coordinates, we can also discover the special relations between these patterns and some specific field regions. Using additional information about the localization of the matches, it could be possible to find player patterns correlated to environmental variables. J. Borrie [2] presents an analysis approach based on a very general type of time pattern called a T-pattern. The corresponding detection algorithm for a T-pattern allows detection of repeated temporal and sequential structures in real-time behavior records. A generic observational software package called Theme has been developed specifically to operationalize T-pattern detection as an analysis process. Specialists are able to analyze the moves performed, the player positioning, pass distances and many others variables related to a specific player. In addition, the information of all team players can be combined to perform analysis over the team strategy and positioning. J. Gudmundsson [11] provide two tools for passing analysis; the first tool calculates all possible passing alternatives at each point in time and the second tool computes frequent pass sequences. J. Gudmundsson developed another tool that aims at analyzing a player’s, or players’, movement. Given a trajectory or a set of trajectories they look for subtrajectories clusters, which are movements of a player, or players, that are repeated often. StatsBomb [19] provides a tool for coaches and recruiters to analyze opponents game style, strengths and weaknesses, to understand player characteristics and strategies to attack/defend. Other contributes try to find new metrics and KPIs (key performance indicators) useful to evaluate football matches trends, starting from a probabilistic framework. D. Link [14], for example, uses a modelling procedure that starts from the understanding of football, then evaluating the key factors that are involved in attack actions. Their procedure determines a quantitative representation of the probability of a goal being scored for every point in time at which a player is in possession of the ball and they call it dangerousity. They take into account the position, so called zone of the ball, the degree of ball control, the pressure that is put on the player by the opponent and the density of opponent players in front of the goal.</p><p><strong>5 Conclusions</strong></p><p>The proliferation of optical and device tracking system in professional leagues in recent years have produced a large volume of data, and this has subsequently enabled a proliferation of research and efforts across a variety of research communities [12]. It seems relevant to find out concepts and methods allowing to assemble and to organize knowledge about game complexity and dynamic interaction properties of the teams. There has been great progress in research in this field, and it is now up to elite football teams to apply this research and these tools on the pitch. An important problem of doing analysis relates to modeling — what are the aspects in the sports events/matches which are of importance? How can we qualitatively and quantitatively assess individual player performance or the development of tactic and strategic capabilities of a team? Answers to these questions need to consider both data-driven approaches from Data Science, but also, models and concepts developed in Sport Science [20]. As summarized before, the starting point is an organized classification of elementary events at player, unit, team levels. This classification can be achieved using geometric rules and machine learning techniques, in order to get a complete overview of all the events occurring during a football match. Memmert &amp; Perl [15] refers that to evaluate performance data, normally qualitative and quantitative methods are used separately, and suggested the combination of net-based qualitative analyses and stochastic quantitative analyses to improve the information output significantly. This paper surveys the recent research into football sports analysis, and describes a framework in which the various research problems found an application in match analysis tools for both real-time and not real-time use. The paper should be useful as a concise introduction for match analysts and researchers who are new to the field.</p><p><strong>References</strong></p><p>[1] Aboutoihi, S. Football: Guide de L’éducateur Sportif, Paris: Editions ACTIO. 2006.</p><p>[2] Borrie, A., Jonsson, G.K., Magnusson, M. Temporal pattern analysis and its applicability in sport: an explanation and exemplar data Journal of Sport Sciences 20(10) 845–52, November 2002. 2002.</p><p>[3] ChyronHego. TRACAB Gen5, 2019. URL <a href="https://chyronhego.com/products/sportstracking/">https://chyronhego.com/products/sportstracking/</a>tracab-optical-tracking/ <a href="https://chyronhego.com/chyronhego-introducestracab-">https://chyronhego.com/chyronhego-introducestracab-</a>gen5/.</p><p>[4] da Costa, I.T., Garganta da Silva, J. M., Greco P.J., Mesquita, I. Tactical Principles of Soccer: concepts and application Tactical Principles of Soccer, Motriz. 2009.</p><p>[5] Deltatre. Matrics, 2018. URL <a href="https://www.deltatre.com/">https://www.deltatre.com/</a></p><p><a href="https://techinsight.com.vn/language/en/big-data-changing-world-football/">https://techinsight.com.vn/language/en/big-data-changing-world-football/</a></p><p>[6] Duprat, E. Enseigner Le Football en milieu scolaire (Collèges, Lycées) et au club,Paris: Editions ACTIO. 2007.</p><p>[7] FIFA. URL <a href="https://www.fifa.com/">https://www.fifa.com/</a> <a href="https://edition.cnn.com/2018/06/12/sport/fifaworld-">https://edition.cnn.com/2018/06/12/sport/fifaworld-</a>cup-2018-data-analyst-spt-intl/index.html</p><p>[8] Garganta, J., Pinto, J. O ensino do futebol. In: A. Graca e J. Oliveira (Ed.). O ensino dos jogos desportivos, Faculdade de Ci^encias do Desporto e de Educacao Fisica da Universidade do Porto: Rainho &amp; Neves Lda, v.1, p.95–136. 1994.</p><p>[9] Garganta, J. Modelacao tàctica do jogo de futebol — estudo da organizacao da fase ofensiva em equipas de alto rendimento, 312 p. (Doutorado). Faculdade de Ci^encias do Desporto e de Educacao Fisica da Universidade do Porto, Universidade do Porto, Porto, 312 p. 1997.</p><p>[10] Grunz, A., Memmert, D., Perl, J. Tactical pattern recognition in soccer games by means of special self-organizing maps, Hum Mov Sci. 2012 Apr;31(2):334–43. 2012.</p><p>[11] Gudmundsson, J. Football analysis using spatio-temporal tools, Computers, Environment and Urban Systems, Volume 47, September 2014, Pages 16–27. 2014.</p><p>[12] Gudmundsson, J., Horton, M. Spatio-Temporal Analysis of Team Sports-A Survey, ACM Computing Surveys (CSUR), Volume 50 Issue 2, June 2017 Article №22. 2017.</p><p>[13] Irnai, T., Uchiyama, A., Magome, T., Higashino, T. Play recognition using spatiotemporal relationship of football tracking data, 2017 Tenth International Conference on Mobile Computing and Ubiquitous Network (ICMU). 2017.</p><p>[14] Link, D., Lang, S., Seidenschwarz, P. Real Time Quantification of Dangerousity in Football using Spatiotemporal Tracking Data, PLoS ONE, 11(12). 2016.</p><p>[15] Memmert D., Perl J. Game Intelligence Analysis by Means of a Combination of Variance-Analysis and Neural Networks, International Journal of Computer Science in Sport 4(1): 29–39. 2006.</p><p>[16] Opta Sports Ltd. Opta Live Performance Data, 2015. URL</p><p><a href="https://www.optasports.com/news/opta-s-event-definitions/">https://www.optasports.com/news/opta-s-event-definitions/.</a></p><p>[17] Richly, K., Bothe, M., Rohloff, T., Schwarz, C. Recognizing Compound Events in Spatio-Temporal Football Data, International Conference on Internet of Things and Big Data. 2016.</p><p>[18] STATS LLC. SportVU Player Tracking, 2015. URL</p><p><a href="https://www.stats.com/sportvu-football/">https://www.stats.com/sportvu-football/</a> <a href="https://www.stats.com/football/">https://www.stats.com/football/.</a></p><p>[19] StatsBomb. StatsBomb IQ. URL <a href="https://statsbomb.com/">https://statsbomb.com/</a></p><p><a href="https://statsbomb.com/teams/">https://statsbomb.com/teams/.</a></p><p>[20] Stein, M., Janetzko, H., Seebacher, D., Jager, A., Nagel, M., Holsch, J., Kosub, S., Schreck, T., Keim, D.A., Grossniklaus, M. How to Make Sense of Team Sport Data: From Acquisition to Data Modeling and Research Aspects,</p><p>[21] Toshie, M., Masahide N., Masahide N., Mahito F., Mahito, F., Nobuyuki, Y. Soccer Event Classification Based on Formation Analysis of Player Positions, The Journal of The Institute of Image Information and Television Engineers 61(9):1367–1375. 2017.</p><p>[22] Tovinkere, V., Qian, R.J. Detecting Semantic Events in Soccer Games: Towards A Complete Solution Proceedings of the IEEE International Conference on Multimedia and Expo (ICME), Tokyo, Japan, 25 August 2001. 2001.</p><p>[23] Xu, C., Zhang, Y.F., Zhu, G., Rui, Y., Lu, H., Huang, Q. Using Webcast Text for Semantic Event Detection in Broadcast Sports Video, IEEE Trans. Multimed, 10, 1342–1355. 2008.</p><p>[24] Zerhouni, M. Principes de base du football contemporain, Fleury: Orges. 1980.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a19db1a788a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>