Full Download Performance Analysis in Game Sports Concepts and Methods Concepts and Methods Martin Lames PDF
Full Download Performance Analysis in Game Sports Concepts and Methods Concepts and Methods Martin Lames PDF
Full Download Performance Analysis in Game Sports Concepts and Methods Concepts and Methods Martin Lames PDF
com
https://ebookmeta.com/product/performance-
analysis-in-game-sports-concepts-and-methods-
concepts-and-methods-martin-lames/
OR CLICK BUTTON
DOWLOAD EBOOK
https://ebookmeta.com/product/research-methods-the-key-concepts-
michael-hammond/
https://ebookmeta.com/product/family-therapy-concepts-and-
methods-12th-edition-michael-p-nichols/
https://ebookmeta.com/product/understandable-statistics-concepts-
and-methods-13th-edition-charles-henry-brase/
https://ebookmeta.com/product/decision-making-concepts-methods-
and-techniques-1st-edition-shyama-prasad-mukherjee/
Urban Mobility and Social Equity in Latin America:
Evidence, Concepts, Methods 1st Edition Daniel Oviedo
https://ebookmeta.com/product/urban-mobility-and-social-equity-
in-latin-america-evidence-concepts-methods-1st-edition-daniel-
oviedo/
https://ebookmeta.com/product/fluorescence-in-situ-hybridization-
fish-for-microbial-cells-methods-and-concepts-nuno-f-azevedo-
carina-almeida/
https://ebookmeta.com/product/the-essentials-of-financial-
modeling-in-excel-a-concise-guide-to-concepts-and-methods-
michael-rees/
https://ebookmeta.com/product/advances-in-modeling-and-
simulation-in-textile-engineering-new-concepts-methods-and-
applications-1st-edition-nicholus-tayari-akankwasa-editor/
https://ebookmeta.com/product/research-methods-the-key-
concepts-2nd-edition-michael-hammond-jerry-wellington/
Performance Analysis
in Game Sports:
Concepts and Methods
Martin Lames
123
Performance Analysis in Game Sports:
Concepts and Methods
Martin Lames
Performance Analysis in
Game Sports: Concepts
and Methods
Martin Lames
Faculty of Sports and Health Sciences
Technical University Munich
Munich, Germany
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This book is based on many years of teaching the subject of performance analysis
as a master’s course at TU München, Germany. Even more influential, it is based on
three decades of research in this area covering theoretical as well as practical
aspects. Theoretical activities include the search for conceptual and methodological
achievements in the analysis of game sports (a term for net games plus invasion
games), including mathematical and stochastic modelling. Practical activities com-
prise the introduction of technological innovations, giving support in match analysis
to top-level, mostly national teams, and deriving a conceptual framework for work-
ing in the practice of performance analysis based on these experiences.
The scientific roots of this work lie in the discipline of sports science one could
call “training and exercise science” as closest translation of German
“Trainingswissenschaft”. The aim of this discipline of sports science is to provide
scientific foundation for practical action in training and competition. Although it
seems to be a quite narrow and merely applied perspective, a closer look reveals that
this is by no means the case. To give a scientific foundation for practice it is, for
example, necessary to understand the structure of the respective sports discipline.
This, in turn, requires investigations of the type of basic research trying to establish
general findings that for example explain success in competition. Also, it is neces-
sary to identify properties of athletes as determinants of performance and to estab-
lish the relationships between them, what will altogether be called “theoretical
performance analysis”. Nevertheless, scientifically founded support for practice
remains the ultimate task that may only be solved by applying special and different
methods and concepts: “practical performance analysis”.
The book has six major chapters, starting with basic concepts, continuing with
the two most important methods of data collection in performance analysis, action
detection and position detection. Finally, concepts and methods of theoretical and
practical performance analysis are presented.
Chapter 1: Basics explains the underlying concepts and functions of performance
analysis in the broader framework of training and exercise science. Special attention
is given to—compared to other sports—the unique structure of game sports. The
concept introduced to explain the nature of game sports is a dynamic interaction
process with emerging behaviour. This concept will be substantiated in detail
because it is the reference point of the whole book. Another basic concept is the
distinction between theoretical and practical performance analysis that is introduced
v
vi Preface
methods and concepts for giving scientific support to practice are addressed in a
systematic, comprehensive manner, whereas giving a review on existing studies in
PA is not the priority. Readers of this textbook are scientists and students interested
in a comprehensive, concept-driven overview of the scientific discipline of perfor-
mance analysis, but also practitioners working in sports practice with an interest in
conceptual backgrounds and a critical reflection of their daily work.
1 Basics���������������������������������������������������������������������������������������������������������� 1
1.1 Definitions and Concepts������������������������������������������������������������������ 1
1.1.1 Performance Analysis ������������������������������������������������������������ 2
1.1.2 Competition, Training, and Athletes’ Abilities ���������������������� 2
1.1.3 Theoretical and Practical Performance Analysis�������������������� 4
1.1.4 General Model of Sports Performance Structure�������������������� 7
1.2 Performance Analysis in Game Sports �������������������������������������������� 9
1.2.1 The Nature of Game Sports���������������������������������������������������� 9
1.2.2 Basic Problems of Performance Analysis
in Game Sports������������������������������������������������������������������������ 12
1.3 Approaches in Performance Analysis ���������������������������������������������� 14
1.3.1 Classical Performance Analysis���������������������������������������������� 15
1.3.2 Notational Analysis���������������������������������������������������������������� 19
1.3.3 Sports Analytics���������������������������������������������������������������������� 21
2 Action Detection ���������������������������������������������������������������������������������������� 23
2.1 Assessment of Behaviour������������������������������������������������������������������ 23
2.2 Design of Observational Systems ���������������������������������������������������� 25
2.2.1 Type of Observational Systems���������������������������������������������� 27
2.2.2 Elements of Observational Systems���������������������������������������� 31
2.2.3 Complex Observational Systems�������������������������������������������� 37
2.3 Validation of Observational Systems������������������������������������������������ 37
2.3.1 General Framework of Validation ������������������������������������������ 38
2.3.2 Role of Observer �������������������������������������������������������������������� 40
2.3.3 Methods and Statistics for Testing Observer
Agreement������������������������������������������������������������������������������ 43
2.4 Examples for Studies Using Action Detection���������������������������������� 53
2.4.1 Event Profiling������������������������������������������������������������������������ 53
2.4.2 Detailed Event Observation���������������������������������������������������� 54
2.4.3 Hierarchical Categorial System���������������������������������������������� 56
ix
x Contents
3 Position Detection�������������������������������������������������������������������������������������� 59
3.1 Functioning of Position Tracking����������������������������������������������������� 60
3.1.1 Position Detection Methods���������������������������������������������������� 60
3.1.2 Signal Processing�������������������������������������������������������������������� 66
3.2 Validation of Tracking Systems�������������������������������������������������������� 70
3.2.1 Gold Standards for Position Tracking in Sports �������������������� 72
3.2.2 Design of Validation Studies�������������������������������������������������� 74
3.2.3 Accuracy of Position Tracking in Sports�������������������������������� 79
4 Theoretical Performance Analysis������������������������������������������������������������ 83
4.1 Statistical Approaches of TPA���������������������������������������������������������� 84
4.1.1 Performance Profiles�������������������������������������������������������������� 85
4.1.2 Impact of Influencing Factors ������������������������������������������������ 88
4.1.3 Criticism of Statistical Approaches���������������������������������������� 92
4.2 Modelling Approaches���������������������������������������������������������������������� 93
4.2.1 Methodological Aspects of Modelling Approaches���������������� 94
4.2.2 Direct Modelling of Game Behaviour������������������������������������ 97
4.2.3 Importing Models to PA���������������������������������������������������������� 115
4.3 Dynamical Systems Theory Approaches������������������������������������������ 133
4.3.1 Dynamical Systems Theories�������������������������������������������������� 133
4.3.2 Complex Systems, Synergetics, and Relative Phase�������������� 136
4.3.3 Ecological Psychology������������������������������������������������������������ 144
4.3.4 Applications of DST in PA ���������������������������������������������������� 148
4.3.5 Outlook DST in PA ���������������������������������������������������������������� 167
5 Practical Performance Analysis���������������������������������������������������������������� 177
5.1 Introduction�������������������������������������������������������������������������������������� 177
5.2 Concepts of PPA ������������������������������������������������������������������������������ 178
5.2.1 Definition, Aims, and Research Strategies������������������������������ 178
5.2.2 Informational Coupling of Competition and Training������������ 181
5.2.3 Comprehensive Performance Analysis ���������������������������������� 188
5.3 Methods of PPA�������������������������������������������������������������������������������� 190
5.3.1 Qualitative Game Analysis������������������������������������������������������ 190
5.3.2 Development of Match Strategies������������������������������������������ 201
5.3.3 Video-Based Tactics Training (VTT)�������������������������������������� 204
5.4 Game Analysts in Professional Training Systems���������������������������� 214
5.4.1 Applications of Game Analysis���������������������������������������������� 215
5.4.2 The Role of Game Analysts���������������������������������������������������� 218
5.4.3 Game Analysis Software�������������������������������������������������������� 220
5.4.4 Club Information Systems������������������������������������������������������ 223
6 Outlook������������������������������������������������������������������������������������������������������� 227
6.1 Outlook on the Core Topics of PA���������������������������������������������������� 227
6.1.1 Basic Concepts������������������������������������������������������������������������ 227
6.1.2 Action Detection �������������������������������������������������������������������� 228
Contents xi
Index�������������������������������������������������������������������������������������������������������������������� 255
Abbreviations
AI Artificial intelligence
DMA Double moving average
DST Dynamical systems theory
EPTS Electronic performance tracking system
FIFA Fédération Internationale de Football Association
GDR German Democratic Republic
GNSS Global navigation satellite system
GPS Global positioning system
IFAB International Football Association Board (rule commission of FIFA)
LPS Local positioning system
MLS Minimum least squares
NBA National basketball association (highest US basketball league)
NTSC National Television Standards Committee (video norm)
PA Performance analysis
PAL Phase alternation line (video norm)
PPA Practical performance analysis
QGA Qualitative game analysis
RFID Radio-frequency identification
SNA Social network analysis
TPA Theoretical performance analysis
TU Technical University
VBT Video-based tracking
VR Virtual reality
VTT Video-based tactics training
xiii
List of Figures
xv
xvi List of Figures
Figure 4.8 ACF of distance covered for the mean of each 5-min interval and mean
ACF the lags of all matches
Figure 4.9 Z-values for playing time, total distance covered, and distance covered
standardized with playing time for the 5-min interval with maximum
intensity (max) and for max+1
Figure 4.10 Interaction graphs and minimum spanning trees for basketball, football,
and handball
Figure 4.11 Two plays, their network, and their adjacency matrix
Figure 4.12 State-transition model for tennis
Figure 4.13 State-transition model for table tennis
Figure 4.14 Transition matrix of a tennis match
Figure 4.15 Transition matrix of a table tennis match
Figure 4.16 Relevance of tactical behaviours in tennis
Figure 4.17 System dynamics of a damped and a driven pendulum
Figure 4.18 A tennis rally of Justine Henin and Serena Williams with phase space
trajectories of each player
Figure 4.19 Illustration of Relative Phase for in-phase, anti-phase, and the general
phase relation between two objects
Figure 4.20 Coordination patterns in finger-waggling; positions of the two finger
tips over time in Kelso’s finger waggling experiment; potential land-
scapes for coupled oscillators
Figure 4.21 Conceptual model of a phase space in football
Figure 4.22 Left: Basic nonlinearities in team and net sports Right: Nonlinearities
in football matches
Figure 4.23 Left: Goals shot in Bundesliga season 2019/20 plotted against shots at
goal; Right: Plotted against goal attempts
Figure 4.24 Proportion of “chance variables” and chance goals of all goals
Figure 4.25 The rate of chance goals in scored and conceded goals by the first and
last team and the four first and last teams of Bundesliga 2011–12
Figure 4.26 Distribution of results in a football match assuming two independent
negative binomial distributions
Figure 4.27 Team centroids of Italy and France in the World Championship final
2006 with Relative Phase
Figure 4.28 Illustration of a rally in a net game as dynamical system
Figure 4.29 Perturbation profile for two matches of Nadal and Federer, French
Open 2007
Figure 4.30 Age-dependent course of critical goal situations (CGS) and perturba-
tions per match and CGS per goal and perturbations per CGS
Figure 4.31 Perturbation profiles of football teams of different age groups
Figure 4.32 Colour-coded recurrence plot for a football match
Figure 4.33 Recurrence plots of nine randomly selected football matches
Figure 5.1 The informational coupling between competition and training
Figure 5.2 Illustration of the concept of comprehensive performance analysis
Figure 5.3 Illustration of the concept of considering interactions between match
analyst and coach, staff, and athletes as being embedded in a
social context
List of Figures xvii
Figure 5.4 Illustration of the analogy between Qualitative Content Analysis and
Qualitative Game Analysis
Figure 5.5 Steps of QGA
Figure 5.6 Conceptual model of strategy development including feedback through
strategy check
Figure 5.7 The trimodal communication model of Merten
Figure 5.8 Organizational sequence and results of video-based tactic test and
match behaviour of service
Figure 5.9 Different roles of match analysts in a sports club
Figure 5.10 Design and user interfaces of a table tennis match analysis software
Figure 5.11 Different roles in the staff of a professional football club
Figure 5.12 Architecture of a club information system
List of Tables
xix
List of Boxes
xxi
xxii List of Boxes
This chapter deals with basics of performance analysis embedded in a broader per-
spective of training and exercise science starting with central definitions and con-
cepts. Although this is an arduous and tiresome, not at all easy-going start, these are
necessary prerequisites not only for the course of the book but also to show the
contrast to different approaches of performance analysis. After this, a paragraph
outlines the particularities of game sports compared to other types of sports. It will
be proven that in performance analysis literature, these differences are not suffi-
ciently acknowledged, and sometimes even a tacit identification of sports and game
sports is found. The chapter closes with a comparison of different approaches
including UK’s notational analysis and US sports analytics.
In a classical academic tradition, this textbook starts with definitions necessary for
creating a common background of understanding. Sports science is not famous for
being a very analytical science, but the roots of this approach can be seen in the
conceptual systems going back to Soviet Union and GDR (German Democratic
Republic) with their acknowledged strong definitions and systematics. In the field
of performance analysis, these traditions meet an application field with completely
different traditions, frequently associated with professional (US- and West
European) team sports, specific head-coach personalities and scouts and analysts
relying on gut feelings. As things have changed dramatically in modern professional
team sports, nowadays being characterized by large, scientifically educated staffs, it
is about time to raise the conceptual framework of performance analysis on a cor-
responding level, starting with a precise understanding of the terms used.
A first comment must state that this suggested definition represents an “open”
concept of performance analysis, open to different methods including modern
machine learning approaches, but the grass-rooted coach’s out-dated gut feeling is
still included (he for sure assesses the competition!). Purposes of PA range from
establishing statistical laws on the structure of performances to supporting the foot-
ball coach in deriving recommendations for tactical changes in the last 10 min with
one goal behind.
A second comment refers to the notion that PA is not limited to scientific activi-
ties. Assessing performances is something that happens in sports practice as well,
and thus, a much broader framework is drawn. Differences and commonalities of
PA conducted in academic and practical settings are issues that will turn up through
the whole of this book.
As mentioned in the introduction, the scientific home of book and author is training
and exercise science in the sense of German “Trainingswissenschaft”. This disci-
pline of sports science aims at providing scientific support for practical action in
sports (Hohmann et al. 2020). The scientific subject of training and exercise science
consists of three areas: competition, training, and athletes’ capabilities including the
interrelationships between them (see Fig. 1.1).
The terms “training” and “competition” need no further explanation. With “capa-
bilities”, personal properties of athletes are denoted that are required for acting
1.1 Definitions and Concepts 3
A first conclusion on the nature of training may be derived from this scheme:
The basic function of training, preparing athletes for competition, may be
achieved only indirectly via impacting on the required levels of athletes’ capa-
bilities. Only these capabilities have a direct impact on behaviour in competi-
tion, for example, my jumping capability allows me to accomplish a header
after a corner. Using a belligerent metaphor, one could say: In training the
weapons for the battle are sharpened, but the battle is decided on the battle-
ground. This holds true to an even larger extent for game sports compared to
other groups of sports as will be explained in the section on the nature of game
sports below.
Analysing the relations between the three big topics allows for identifying two
main tasks of performance analysis (PA): First, PA has to identify determinants of
success in competition and to support conclusions on required levels of athletes’
capabilities. Second, PA has to analyse behaviour in competition to identify
strengths and weaknesses of the athletes. The latter is conducted with regard to
deriving practical recommendations for training. These two different tasks give rise
to the important distinction between practical and theoretical PA being introduced
in the next paragraph.
from statistical laws but from sound reconstruction of behaviour and interpretations
relying on a broad information based on the whole relevant context, for example,
game situation, opponent, one’s own tactics, preparation, motivation, and injuries/
rehabilitation. In Chap. 5, dedicated to details and methods of PPA, it will be
explained why qualitative methods play an important role in this process.
In the aspects discussed above and others (see Table 1.1), there are differences in
many respects between TPA and PPA. These differences in aim, method, sample,
setting, and design make a clear distinction more than appropriate. Especially what
scientific methodology is concerned, differences cannot be underestimated. The dis-
tinction between TPA and PPA as two different research areas promises also to
clarify many problems discussed in the so-called practical impact debate.
Many of the issues brought up in the practical impact debate could be solved
partially by omitting misunderstandings of the nature of scientific findings but most
important by acknowledging the differences between the two sub-disciplines of PA,
TPA and PPA!
Nevertheless, it would create a wrong impression if these two areas were per-
ceived as being totally independent. First, people engaged in PPA should have and
usually have an education in TPA. Ideally, during their education, they are made
aware of the differences between practical and scientific work concerning both con-
ceptual and more hands-on aspects of each area. Second, there are specific relation-
ships: Results of TPA provide a valuable background for doing even more informed
PPA by giving, for example, a normative framework for analyses of one’s own team.
On the other hand, because of its close connection to the “real world”, PPA should
be the place where new hypotheses, new explanations, and new determining vari-
ables appear that are candidates for an investigation within the framework of TPA.
Scientifically founded PA should start with a solid idea, concept, or model of its
subject, that is, performances in sports. In Fig. 1.2, it is illustrated that athletes’
performance in competition is of top interest. Compared to environmental factors,
the capabilities of the athletes are the primary influencing factor from a practical
point of view, because they may become targets of training as shown in Fig. 1.1. In
the following section, the model will be introduced more in detail before it will be
tailored to game sports.
Competition comprises the overall result as well as parts of performance. Parts of
performance are obtained by the following:
What is a game sport? The term is not very common in English, where, for example,
expressions like “team sports” or “individual sports” are more in use with team
sports sometimes implicitly taken for game sports. But obviously, there are sports
run by teams (rowing, relays) not being games and most net games are individual
sports. The term “formal games” (consisting of net games, invasion games, and
striking/fielding games) created by Read and Edwards (1992) and adopted by
Hughes and Bartlett (2002) and Hughes and Franks (2004) was never popular and
is nowadays reserved for serious games and eSports. Moreover, since PA deals also
10 1 Basics
with other sports than game sports, it is a good start to look for a definition of these
sports when aiming to point out the peculiarities of PA in game sports.
Searching for a more apt definition of game sports, it is helpful to consider them
as a group of sports (Aristotle’s genus proximum) with special characteristics (dif-
ferentia specifica). From a PA perspective, it is convenient to distinguish groups of
sports according to their leading determinant of performance. There is a suggestion
by Thiess and Schnabel (1976), two leading sports taxonomists in GDR, comprising
five groups of sports, that is, endurance and strength sports, game and combat
sports, and technical sports, the latter with the sub-groups of artistic and driving-
flying-shooting sports.
What is the typical property of game sports as a group of sports? There is an early
answer in the author’s doctoral thesis:
Explanations:
• The definition follows the classical structure: Game sports form a group of sports
with a distinguishing property from other groups of sports.
• The distinguishing property of game sports is given by the specific way of inter-
action between the (two) parties.
• The specific way of interaction is given by the simultaneous striving of the par-
ties for their aim (offence) and by preventing the opponent from reaching his one
(defence).
• The only way to distinguish game sports from combat sports is connected with
the different nature of the aims in these two groups of sports. While in combat
sports the aim is a manipulation of the opponent’s body (hitting, kicking, stab-
bing, throwing on the mat, etc.), aims in game sports are symbolic acts per-
formed with the game object (e.g. to throw the ball in a basket or to hit a winner).
According to Read and Edwards (1992), three sub-groups of game sports with
variants of the typical interaction process are distinguished:
1. Invasion games such as football codes (soccer, rugby, American, Australian, and
Gaelic football), handball, basketball, and several others: The two teams are on the
pitch at the same time (different: American football with only the special teams
facing each other on the pitch!). The aim is to bring the game object (the ball) into
a certain space, typically a goal or a basket, but also an “end-zone” marked by a
line or through a “goal” built up by two posts and a bar. There are phases of ball
possessions either finishing with rule-based events (goal, shot at goal, offensive
1.2 Performance Analysis in Game Sports 11
foul, out of bounds, and others) or turnovers. There is physical contact between the
two parties, and we typically have a back and forth on the pitch, because the play-
ers group themselves more or less around the playing object.
2. Net games such as volleyball, tennis, table tennis, and squash (the net is a wall
here!): The game is made up of rallies ending with a point for one party. Rallies
consist of alternating strokes (different: volleyball, a team net game!) by the two
parties. The aim of the rally is to hit a winner or to force the opponent to hit an
error. The parties being separated by a net (different: squash!) are not in physical
contact. The sequence of alternating actions is the typical interaction in this
group of game sports, making something such as “alternating strokes games” a
better (there is a net in football as well!), but inconvenient, denomination.
3. Pitcher/batter games such as baseball, cricket, and softball: In these game sports,
we find a special interaction structure given by the fact that the aims of the teams
on the pitch are not the same. For example, in baseball, the batter’s team tries to
score runs and the pitcher’s team tries to avoid this and getting players out. These
roles are changed in the second part of an inning. In the light of our definition,
the teams are simultaneously striving for their respective aims, but these aims are
different ones over certain periods of the match.
Summing up these considerations on the nature of game sports, one may state
that their constitutive property is interaction. From this basic statement, some fur-
ther conclusions on their nature with important consequences for PA may be drawn:
Achieving a symbolic aim against the resistance of the opponent requires a plan,
a strategy how to do this. Such a plan has to take one’s own and the opponent’s
capabilities into account. These plans may also be called the tactics of a player or a
team, showing that in game sports, tactics is of outstanding importance compared to
other groups of sports. In the case of team games, tactics is even more dominant
since within-team cooperation demands for an additional dimension of planning or
tactics.
With respect to success, one may state that if a team makes an unsuccessful
move, it will very likely try a different move to avoid failure thereafter. On the other
hand, if there is a successful move or action, the opponent has every reason to
change his behaviour. In sum, it is very unlikely to find a constant playing pattern
all over a match; behaviour changes over time, and it is dynamic!
Taken together with the constitutive aspect of interaction, the nature of game
sports is best captured by treating them as dynamic interaction processes with
emerging behaviour. This concept was suggested by several researchers in the area,
already (e.g. Lames (1991), Passos et al. (2011)). The consequences for PA may
hardly be underestimated. For example, the frequent practice in PA to describe a
performance with summative statistics on the frequency of a certain behaviour dur-
ing a match (“stats”) and taking it as a measure of the capability of one player or one
team is a questionable method in the light of this concept, as the first practice does
not reflect dynamics and the second does not reflect interaction!
The resulting behaviour in a game perceived as a dynamic interaction process is
only weakly connected to the levels of abilities of the players. It spontaneously
12 1 Basics
emerges from the interaction and is not repeatable and hard to predict (see Box
“Where is the ball 10 s after a corner?” in Chap. 4). Emergence, the development of
new structures from complex interactions between sub-systems, is a crucial concept
for understanding game behaviour. This concept has already been introduced earlier
(Kelso 1995; Glazier 2010; Duarte et al. 2013), and PA is still looking for appropri-
ate methods to deal with it (see Sect. 4.3.1 in Chap. 4 on TPA).
Reflections on the nature of game sports like in the last section have far-reaching
consequences for PA as already denoted. The general model for a performance in
sports (Fig. 1.2) has to be adapted to reflect the conditions in game sports. In
Fig. 1.3, the observable behaviour in competition is depicted as the result of the
dynamic interaction of the two sets of athletes’/players’ capabilities.
A problem caused by the nature of game sports is the fact that we only can “see”
the result of the interaction process between the parties on the pitch and we do not
directly perceive the capabilities of the parties. On the other hand, drawing infer-
ences on players’ capabilities is a most important aim of PPA. In the light of the
previous considerations, it is no surprise that it requires specially designed methods
that allow looking behind the curtain of the interaction process (see Chap. 5 on PPA).
Lames and McGarry (2007) compared the structure of performance in a game
sport with the one of a 100 m sprint. The time scored in a 100 m sprint may be inter-
preted as an expression of the sprinter’s potential. It may even be conceived as a
realization of his capabilities at the occasion of the performance under scrutiny. It is
obvious that there is a very close relationship between his capabilities and his per-
formance. It is a comparatively easy task for PA to identify strengths and
1.2 Performance Analysis in Game Sports 13
weaknesses of a sprinter given the kinematics of his 100 m performance, for exam-
ple, with reaction time, peak acceleration, peak velocity, and sprint endurance index.
This does not hold true at all for performances in game sports. Since observable
behaviour emerges from the dynamic interaction with the opponent, PA in game
sports has to solve two basic problems: taking into account the changes in perfor-
mance over time of a match and estimating the performance taking into consider-
ation the behaviour of the opponent.
Performance Indicators
As pioneers of performance analysis already mentioned (in Germany:
Hagedorn 1972), performance in game sports is expressed by the frequency of
actions either directly associated with reaching the aim of the game, for exam-
ple, scoring or preventing a goal, or being supportive to this. Therefore, per-
formance indicators, being single or combined action variables that describe
some or all aspects of performance (Hughes and Bartlett 2002), are used to
assess performance in game sports. O’Donoghue (2010) added necessary
conditions for a variable becoming a performance indicator borrowed from
business and engineering: it must have been proven to be a valid indicator for
an important aspect of performance and possess three metric properties,
namely, an objective measurement procedure, a known scale of measurement,
and a valid means of interpretation.
Confronting this notion of performance indicators with the nature of game sports,
both constitutive properties, dynamics and interaction, create problems. A typical
performance indicator consisting of (normalized) frequencies of actions in a match
is first, a summative statistic and as such static and not dynamic, and is second,
attributed to the performance of one player or one team and thus does not reflect the
interaction.
Sampaio and Leite (2013) mention this problem and optimistically state that
continuing development is required in the forthcoming years, whereas the author
sees it as a basic “dilemma of game sport research” (Lames 1991) and consequences
are to be drawn concerning basic concepts of research strategies and methodologies
to be explained in the remainder of this book.
Of course, classical approaches in PA have perceived these problems as well.
One way of dealing with the problem of dynamics is to refer to the “law of large
numbers” that will cure dynamic changes in behaviour based on a large number of
observations. This might seem to be acceptable in TPA, but in PPA, we are inter-
ested in the “true value” of a performance of one player in one game. This is one
reason why in PPA qualitative research methodology is applied.
What the influence of opposition is concerned, a suggestion in TPA is to intro-
duce opponent’s strength as a confounding variable (O’Donoghue 2009; Lago
2009). One must state though, that this procedure is far away from analysing behav-
iour emerging from the interactions between the opponents. Instead, emergence
14 1 Basics
means the situational mutual shaping of behaviour, which is hardly accounted for by
introducing another summative, static variable “level of opponent”. For the purpose
of PPA, a much deeper analysis of how the qualities of the opponent impact the
performance of one’s own players is required bringing qualitative methods with in-
depth reconstruction and interpretation into play (see Chap. 5 on PPA).
Finally, it is worth noting that the impact of interaction and dynamics on perfor-
mance in game sports varies between the different game sports.
In net games, there is a tight interaction induced by the alternating sequence of
strokes. Each stroke, except service, can be seen as highly influenced by or as an
answer to the prior stroke. For a service, we must acknowledge that its impact
strongly depends on the qualities of the return player as well.
In invasion games, trying to invade a certain space against a defence implies
strong interactions, too. In addition, it is maybe a good idea to distinguish between
high-scoring games (basketball, handball) and low-scoring ones (soccer, field and
ice hockey). In high-scoring games (basketball and handball roughly show around
40% successful ball possessions), attacks are more successful, and points are scored
on a regular base with fewer opportunities for the defence to interfere to a well-
planned and executed attack. In low-scoring games though, we have additional
components that make the success of attacks unlikely, such as chance and instability
or chaotic phenomena (see Chap. 4).
In pitcher/batter games, actions are strictly sequential and frequently analysed in
isolation (pitching game, batting game). Apart from the (crucial) pitcher-batter
interaction, success depends to a great extent on individual skill. This means that in
this group of sports, where performance is made up of sequential, rather indepen-
dent actions (pitching, batting, catching, running), the connectedness of game
behaviour and individual skill is comparatively close, and interaction is lower than
in the other two families of game sports. In this light, it is not a surprise that the
longest tradition in recording summative static PIs is found in baseball.
Performance analysis has a rather long history with respect to disciplines of sports
science. Nevertheless, only in recent years technological developments have
increased the potential of PA dramatically. This may be documented with the num-
ber of publications found in Google Scholar with the keyword “performance analy-
sis in football” given in Fig. 1.4. There is an exponential growth until 2013 and a
levelling off at a high level since then. The modelling method of using regression
functions is discussed below in Sect 4.2 in Chap. 4.
In the remainder of this paragraph, historically and actually relevant approaches
of PA are described.
1.3 Approaches in Performance Analysis 15
Fig. 1.4 Hits in Google Scholar for “Performance Analysis Football”; from 1990 to 2013, an
exponential trend fits data with R2 = 0.998
The term “classical performance analysis” is chosen here, because this is a rela-
tively old approach of performance analysis but a basic one for more recent devel-
opments. It was established in the late 70ies/early 80ies of last century in the form
presented here by Manfred Letzelter from Johannes Gutenberg University at Mainz,
Germany. He merged analytical models on the structure of performances in sports
with empirical research methodology, which was a big achievement for applied
sports science in these times. There are three main characteristics of this approach:
In his seminal paper, the method of classical performance analysis was described
in three steps (Letzelter and Letzelter 1982):
16 1 Basics
Fig. 1.5 The general structural model for sports performances. (With permission of Limpert-
Verlag from Hohmann et al. (2020))
Fig. 1.6 Empirical structure of performance in cross-country skiing. (Reproduced with permis-
sion of Philippka Verlag from Ostrowski and Pfeiffer (2007))
An example for step 2 of the above agenda of classical performance analysis, the
examination of the empirical structure, is given in Fig. 1.6. This study on Nordic or
cross-country skiing (Ostrowski and Pfeiffer 2007) shows the empirical relations
between four levels of the general model. The complex performance is given by the
running time on a cross-country skiing trail. Parts of performance are given by the
time used for parts of the trail with different slopes (uphill, flat, downhill with and
without use of arms). The level of specific performance prerequisites is represented
by three capabilities, running endurance on a skiing treadmill and strength endur-
ance on an arm-pull ergometer for diagonal and parallel pulls. Finally, physiological
parameters such as heart rates, watts per kilogram bodyweight, and VO2 at certain
intensities specified by lactate levels in the blood are given representing the level of
general performance prerequisites.
The small figures depicted in Fig. 1.6 show the respective R2-values of a regres-
sion with all variables from the lower level as independent variable and the con-
struct on the higher level as dependent variable. One may see, for example, that
skiing treadmill endurance performance is more determined by physiological vari-
ables than arm pull strength endurance obviously requiring specific skills. On the
next level, it is interesting to note that only uphill and flat parts are substantially
influenced by the capabilities under examination and that we have a substantial
amount of unexplained variance. On the other hand, the time needed for these parts
18 1 Basics
of the trail determines to a great extent the overall running time, with the exception
of passive downhill that shows almost no impact. It becomes obvious from this
example that results of a study performed under the paradigm of classical perfor-
mance analysis create valuable background information for training, especially how
to distribute efforts and resources to the many parts of the system.
Despite its conceptional merits, there has been some criticism on classical per-
formance analysis. First, there are some general methodological issues. Empirical
models are based on empirical samples, that is, the athletes that happen to be in a
sample “create” in a way the resulting model of the structure of performance. At this
point, we run into the general problem of samples in top-level sports: at the very top,
the samples are necessarily too small to allow for decent statistical evaluation.
Especially when a differentiated performance model with many variables is to be
filled with empirical content, this becomes an almost insurmountable problem.
Ostrowski and Pfeiffer (2007) built their model on n = 31 athletes and several (nine)
single regressions are run, because a more appropriate structural equation model
does not work at this ratio of sample size vs. number of variables.
Other critical aspects are the variables included in the studies. The claim of clas-
sical performance analysis is that all hypothetically relevant variables are included,
which is very demanding. In practice, the selection of variables is less determined
in a deductive manner, that is, all variables that are hypothetically relevant are
included, but rather in an inductive way, that is, the variables one has at hand are
included in the study. Ostrowski and Pfeiffer (2007) used the results of three tests
applied in routine testing in Nordic skiing as representation of the performance
prerequisites. Maybe the perceived gap with respect to explained variance between
performance prerequisites and performances in different parts of the trail is due to
missing skill or explosive power tests at the level of performance prerequisites.
Another methodological concern is associated with the way statistical methods
treat independent and dependent variables. The linear relationship typically
assumed and assessed by traditional statistical methods may not be adequate for
modelling some well-known relationships between performance prerequisites and
performance, for example:
• A covariation between variables does not reveal the basic mechanisms that are
responsible for the statistical covariation.
• Cases of biased correlations are discussed in research methodology, for example,
correlations induced by moderator or mediator variables.
1.3 Approaches in Performance Analysis 19
Knowing the underlying mechanisms would be very helpful for practice for
designing training interventions. In the same vein, one must admit that empirical
results frequently show ambiguities, for example, is it really justified to devote more
resources to a performance prerequisite that has a common variance of R2 = 0.55
with a complex performance than to one that correlates with R2 = 0.53, especially
given the methodological problems mentioned above?
The most severe objection against classical performance analysis comes from its
basic assumption that performance is determined by the performance prerequisites
of the athletes or teams. As explained in the paragraph on the nature of game sports
above (see Fig. 1.3), this is not the case for game sports and combat sports, where
performance must be considered as the result of the interaction process between the
two parties. Moreover, aspects such as dynamics and emergence, both being consti-
tutive of game and combat sports, are not an issue in classical performance analysis.
Taken together, classical performance analysis was a great innovation for perfor-
mance analysis at the time of its foundation. It was a big support in striving for
academic dignity of applied sports science that was at stake in these times. For game
and combat sports though, this is a conceptually inapplicable approach. Its merits
are to be seen in individual sports, where overall performance is to a great extent
determined by performance prerequisites, for example, track and field or swimming
disciplines. Methodological improvements, for example, capturing nonlinear rela-
tionships with neural network methods or finding design solutions for the sample
problem of top-level sports, will result in further valuable contributions of classical
performance analysis to applied sports science.
The title of two textbooks “Notational Analysis in Sports” (Hughes and Franks
1997, 2004) gave the name to a very early and influential school of performance
analysis predominantly located in the UK. Hughes and Bartlett give a definition:
“Notational analysis is an objective way of recording performance, so that critical
events … can be quantified in a consistent and reliable manner. This enables quan-
titative and qualitative feedback that is accurate and objective” (2008, p. 9).
From this definition, two aspects can be derived being characteristic for nota-
tional analysis. First, it is about event detection which is used to be the only relevant
method of performance analysis for a long time. Only later, technological progress
made position detection available for performance analysis, which since then is the
second standard method in PA (see Chaps. 2 and 3 addressing these standard
methods).
The second aspect of interest is the purpose of notational analysis, which is seen
in providing feedback about the events that took place in competition. Emphasizing
the feedback function of notational analysis may be due to the impact of two
research traditions, first, studies on coaches’ ability to correctly reproduce a match,
and second, the investigation of motor learning processes in movement science.
20 1 Basics
In very early studies, Franks and his co-workers were able to demonstrate that
coaches’ memory of what had happened on the pitch is far away from being flawless
(Franks and Miller 1986, 1991) and, in addition, coaches tend to be very confident
in their judgements (Franks 1993). These proven perceptual weaknesses of coaches
watching a match of course gave rise to a demand for accurate and objective feed-
back which was met by notational analysis. The second source of demand for feed-
back is to be found in movement science, where it is unanimously acknowledged
that feedback, for example, intrinsic or extrinsic feedback, is essential for any motor
learning process. Taking training in game sports as such a learning process, the
importance of feedback becomes obvious.
The position taken in this book is that feedback is of course important, but
in Chap. 5 on PPA, the restriction on just giving feedback will be given up. In
addition, the differences between a motor learning process and cognitive learn-
ing processes will be mentioned there with respect to consequences for the
need and the structure of feedback in PPA. Instead of being restricted to merely
giving feedback, the position taken in this book opens performance analysis to
the derivation of practical measures as well. In notational analysis, deriving
these measures is an exclusive task of the coach as it becomes clear in their
early concepts of the coupling of athletes’ performance and practice (Maslovat
and Franks 2008).
The restriction on giving feedback made it hard for notational analysis to deal
with some specific conceptual issues of game sports. The distinction between
behaviour in competition and players’ capabilities, which is essential for practical
performance analysis, is not important when one just aims at describing matches.
Also, dynamics and interaction in game sports were initially not of interest to
notational analysis and the search for summative PIs for the whole match pre-
vailed (Hughes and Bartlett 2002; Nevill et al. 2008). In contrast to within-match
dynamics, which was treated more or less only via modified score lines (Hughes
2004; Hughes et al. 2013), match-to-match variability became more an issue
(Hughes et al. 2001b). The desire for PIs that had “stabilized,” i.e., that match-to-
match fluctuation of PIs is below an acceptable level allowing to speak about a
“normative profile” was expressed. A normative profile is required when infer-
ences about strengths and weaknesses are to be made. A considerable number of
studies were conducted to determine empirically the number of matches required
to obtain a reasonably stable estimate for the mean values of PIs. O’Donoghue
and Ponting (2005) made clear that this question could be answered also analyti-
cally by a probability estimation assuming the law of large numbers in statistics.
It is interesting to see that researchers who have great experiences as players,
coaches, or performance analysts and for sure have an excellent understanding
of the structure of performances in game sports fail to adopt the corresponding
concepts in their scientific work. Nevertheless, notational analysis is to be
acknowledged as a pioneering approach in PA that has remained very influential
ever since.
1.3 Approaches in Performance Analysis 21
In US sports, especially in the professional clubs of the big five leagues (NBA,
MLB, NFL, MLS, NHL), there is a long tradition of analysing sports data. The so-
called sabermetrics (SABR, Society for American Baseball Research) in baseball,
for example, go back until the early 1980s. They replaced the mere accumulation of
baseball statistics, which may be traced back for baseball until 1845, when the box
score was introduced (Lewis 2003). Interestingly, more advanced PIs from saber-
metrics were first appreciated by fans and media and became a successful business
model, for example, for STATS (Sports Team Analysis and Tracking Systems) Inc.
and ESPN (Entertainment and Sports Programming Network) before the clubs took
notice of it and applied it in their daily work (Lewis 2003). A breakthrough in this
direction was the book Moneyball by Michael Lewis describing the successful use
of sports analytics by manager Billy Beane from the Oakland A’s. Beane managed
to arrive at great decisions, for example, in drafting new players, giving his team a
position in competition that was much better than to be expected from the financial
resources available.
The standard book on sports analytics is Alamar (2013). He describes sports
analytics as the search for patterns in sports data to optimize decisions in sports. A
distinguishing feature of sports analytics in comparison to the two other approaches
mentioned above is that it is not dedicated or restricted for use in sports practice
(Miller 2016) but maybe even more relevant to media and club management (Fried
and Mumcu 2017). In addition, there was hardly any formative involvement of
sports science in the development of sports analytics, instead highly qualified
experts from other sciences, especially informatics (typically being nerds in their
respective sports), applied their knowledge and methods to sports data. Taken
together, sports analytics has become a giant field of business (Harrison and
Bukstein 2017; Link 2018).
Link (2018) gives an impressively long list of stakeholders in sports analytics
with their specific interests in it. This list contains large businesses that pursue spe-
cial interests in sports analytics:
• Betting industry: Optimizing betting odds relies very much on sports analytics,
but also betting frauds may be detected with it.
• E-games: The e-game industry uses sports analytics to make tactical behaviour
in their games as realistic as possible and/or to map individual or team character-
istics onto their avatars.
Historically, recording events in sports competitions is most likely the earliest activ-
ity in PA. In this chapter, action detection in PA is put in the tradition of behavioural
research as a relatively recent development. As notational analysis may be seen as a
special case of behavioural observation, this link opens a broader perspective on the
issue. All observational methods share two important properties: they are limited to
the perceivable surface of behaviour and the measurement instrument is the human
observer. Based on these commonalities, a paragraph is dedicated to the design of
observational systems in sports. Dealing with the central methodological issue of
agreement between the human observers requires introducing some special tech-
niques that maybe are not common in a typical methodological education in sports
science. Finally, examples of observational systems for action detection in sports
are given.
One might ask why to treat observational methods in a special chapter in a text-
book on performance analysis in the age of data providers, action feeds, and artifi-
cial intelligence. The simple answer is that one is relying on competency in these
methods in case there is a deeper interest in details than provided by action feeds or
when there is no coverage by data providers either because of a too low performance
level or in game sports that are not commercially attractive for data providers.
Moreover, pursuing an AI approach in sports does not absolve from testing data
quality, in the case of event data typically applying observational validation methods.
Textbooks of notational analysis often do not mention that methods and designs of
notation systems for sports analyses must be seen as a special case of observational
methods. This perspective, though, allows making use of decades of methodologi-
cal developments and considerations that have led to a complex body of knowledge.
The last blow was given to the Golden Horde by Girei, Khan of the
Crimea, Ivan’s faithful ally, against whom a mortal hatred was
cherished by Ahmed’s descendants. Girei attacked the Golden
Horde at Sarai, its capital, and destroyed it completely. Ahmed’s son,
then Khan of the Horde, sought refuge among the Nogais. Later on
he went to the Sultan at Tsargrad, and at last to his famous ally, the
King of Poland. There he was put in prison, however, and the king
sent word to Mengli Girei that as long as he remained in peace his
erstwhile disorderly neighbor would be retained in durance.
Thus in 1505 ended the Golden Horde, or the Horde of Sarai, which
had so bitterly oppressed Russia for more than two hundred and
forty years. The continuation of the Horde was the small Astrakhan
Kingdom, once a vassal state in Batu’s mighty empire.
THE END.
[483]
[Contents]
By JEREMIAH CURTIN.
OPINIONS
[Contents]
With Etched Frontispiece. Crown 8vo. Cloth, gilt top, $2.00 net.
CONTENTS.
Introduction. Shaking-Head.
The Son of the King of Erin Birth of Fin MacCumhail.
and the Giant of Loch Léin. Fin MacCumhail and the
The Three Daughters of Fenians of Erin in the
King O’Hara. Castle of Fear Dubh.
The Weaver’s Son and the Fin MacCumhail and the
Giant of the White Hill. Knight of the Full Axe.
Fair, Brown, and Trembling. Gilla na Grakin and Fin
The King of Erin and the MacCumhail.
Queen of the Lonesome Fin MacCumhail. the Seven
Island. Brothers, and the King of
The Shee an Gannon and France.
the Gruagach Gaire. Black, Brown, and Gray.
The Three Daughters of the Fin MacCumhail.
King of the East and the Cucúlin.
Son of a King in Erin. Oisin in Tir na n-og.
The Fisherman’s Son and
the Gruagach.
The Thirteenth Son of the
King of Erin.
Kil Arthur.
NOTICES.
[Contents]
HERO-TALES OF IRELAND 📘.
BY JEREMIAH CURTIN.
CONTENTS.
Elin Gow, the Swordsmith Balor on Tory Island.
from Erin, and the Cow Balor of the Evil Eye.
Glas Gainach. Art, the King’s Son, and
Mor’s Sons and the Herder Balor Beimenach, Two
from Under the Sea. Sons-in-law of King Under
Saudan Og and the the Wave.
Daughter of the King of Shawn MacBreogan and
Spain; Young Conal and the the King of the White
Yellow King’s Daughter. Nation.
The Black Thief and King The Cotter’s Son and the
Conal’s Three Horses. Half Slim Champion.
The King’s Son from Erin, Blaiman, Son of Apple, in
the Sprisawn, and the Dark the Kingdom of the White
King. Strand.
The Amadan Mor and the Fin MacCool and the
Gruagach of the Castle of Daughter of the King of the
Gold. White Nation.
The King’s Son and the Fin MacCool, the Three
White-Bearded Scolog. Giants, and the Small Men.
Dyeermud Ulta and the Fin MacCool, Ceadach Og,
King in South Erin. and the Fish-Hag.
Cud, Cad, and Micad, Fin MacCool, Faolan, and
Three Sons of the King of the Mountain of Happiness.
Urhu. Fin MacCool, the Hard
Cahal, Son of King Conor, Gilla, and the High King.
in Erin, and Bloom of Youth, The Battle of Ventry.
Daughter of the King of
Hathony.
Coldfeet and the Queen of
Lonesome Island.
Lawn Dyarrig, Son of the
King of Erin and the Knight
of Terrible Valley.
OPINIONS.
These are thrilling hero-tales. No extract can do the
stories justice. Any one taking up the volume will not be
likely to lay it down without reading it.—The Cincinnati
Commercial Gazette.
[Contents]
BY JEREMIAH CURTIN.
CONTENTS.
OPINIONS.
[Contents]
By JEREMIAH CURTIN
CONTENTS
Introduction. Ilhataina.
Olelbis. Hitchinna.
Olelbis and Mem Loimis. Tirukala.
Norwan. Sukonia’s Wives and the
Tulchuherris. Ichpul Sisters.
Sedit and the Two Brothers The Finding of Fire.
Hus. Haka Kaina.
Hawr. Titindi Maupa and Paiowa,
Norwanchakus and Keriha. the Youngest Daughter of
Kele and Sedit. Wakara.
Kol Tibichi. The Two Sisters, Haka Lasi
The Winning of Halai Auna and Tsore Jowa.
at the House of Tuina. The Dream of Juiwaiyu and
The Hakas and the Tennas. his Journey to Damhauja’s
Country.
The Flight of Tsanunewa
and Defeat of Hehku.
The First Battle in the World
and the Making of the Yana.
OPINIONS
This eBook is for the use of anyone anywhere at no cost and with
almost no restrictions whatsoever. You may copy it, give it away or re-
use it under the terms of the Project Gutenberg License included with
this eBook or online at www.gutenberg.org ↗️.
Metadata
The Mongols in
Title:
Russia
Jeremiah Curtin Info
Editor:
(1835–1906) https://viaf.org/viaf/37277243/
File generation 2024-01-20
date: 17:49:53 UTC
Language: English
Original
publication 1908
date:
Revision History
2024-01-01 Started.
Corrections
1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside
the United States, check the laws of your country in addition to
the terms of this agreement before downloading, copying,
displaying, performing, distributing or creating derivative works
based on this work or any other Project Gutenberg™ work. The
Foundation makes no representations concerning the copyright
status of any work in any country other than the United States.