Introduction to football analytics 2021 - Complex Systems in ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
12 Introduction
to football
analytics
2021
INTRODUCTION
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
1: Complex Systems
— Angel Ric1,2, Sam Robertson3 & David in Sport Research
Sumpter4,5 Group, Institut Nacional
d’Educació Física de
INTRODUCTION
Catalunya (INEFC),
Lleida, Spain
2: FC Barcelona
3: Victoria University,
The incorporation of new technology into sport has resulted in an increase in Australia
both the volume of data and the variety of its types. This has led to a need for 4: Uppsala University
completely new areas of expertise (such as that provided by data analysts, 5: Hammarby IF
computer scientists and mathematicians) in organisational departments, as
well as adaptation by those already holding established roles, such as sport
scientists, strength and conditioning coaches and performance analysts. The
link between sports and computer sciences is now well established, and few
top-level professional structures resist its integration. However, despite the
interdisciplinarity of many staff teams, the knowledge and linguistic barriers
between different scientific fields mean that interaction between professionals
can often be difficult. One way to help address this problem is for experiential
and scientific knowledge, historically based on reductionism, to move towards
an understanding of sport as a complex phenomenon.
13
The complexity of football lies mainly
in the number of interactions between
teammates and their environment, in
particular their opponents, but is also
found in the context of the game as a
whole. Traditionally, performance has
been evaluated by isolating system
components including the physical and
physiological; the technical (derived
from the players’ on-ball events); or
tactical aspects (the relationship between
teammates and opponents). These
“performance components” cannot
be seen at the same level but instead
in a nested way, because the inherent
processes within them emerge on
different timescales (figure 1). It is in
a competitive environment, evolving
over several years, where cooperative
processes are situated. The team
cooperates at the timescale of weeks
and months as well as more immediately
at the timescale of the match’s 90-plus
minutes to achieve the main objective
of winning. Interpersonal synergies
between players are then formed and
dissolved on a scale of seconds and tens
of seconds on the pitch, allowing them
to score a goal or avoid conceding (Ric et
al. 2016). These team patterns then make
the context for the individual actions that
emerge in a smaller timescale (seconds
or milliseconds). Therefore, context must
be understood as the environment for an
organism-environment relationship at all
scales and in performance analysis, from
seeing the team as an organism (Duarte
et al. 2012), through to the player himself
within the team, and even down to the
organs or systems within the athlete (Pol
et al. 2020).
INTRODUCTION
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
14
Figure 1. Nested levels and scales of performance analysis.
INTRODUCTION
FOOTBALL ANALYTICS 2021 THE ROLE OF CONTEXT IN TRANSFERRING ANALYTICS TO THE PITCH For these levels of performance et al. 2019), as well as the behaviour have on mental or emotional pressure 15 analysis, tracking data (team), event of the team’s performance in relation (Bransen et al. 2019). data (player) and biometric data to the opponent’s disposition and the (organ) is collected systematically. ball (Fernández & Bornn, 2020). The It is common for media and fans to The integration of many of these value of these behaviours (individual consider players in a position within data sets is essential and has or collective) derives from the context a system of play or team formation. been recommended within clubs in which they take place. For example, However, within the same position(s), or institutions for proper data the value of actions-decisions with different roles exist. Because of management (Linke et al. 2018). FIFA, or without a ball has been quantified the dynamism of the game, these in its aim to assist associate members according to variables such as positions are not static and depend or other stakeholders, is developing location, distance to the goal or the on specific contexts that can be standards for the unification of angle, that vary in tens of second. On identified within the game. Among these data sources. Related to this, the other hand, the effect of the style them, moments in which a team has the assignment of unique player and system of play, or the number the ball and others in which it tries IDs aims to avoid duplication and of players on each line of pressure to recover it can be differentiated. incompatibility between different and defensive organisation, implies Similarly, playing strategies such as data sources and competitions, thus contextual changes on a larger pressure or withdrawal to recover providing simplicity for joint data scale of seconds or tens of seconds. the ball or more elaborate direct play analysis at all levels of performance. Furthermore, changes in situational when possessing the ball can also be At a higher level, where performance variables, already noted above, play an identified (Castellano & Pic, 2019). is affected by competition (e.g. the important role in the effects they can league), the effects of situational or contextual variables on the behaviour of players and teams should be considered (Gómez et al. 2013). In fact, during FIFA competitions, performance can be affected by macro-considerations such as the location and climate of where the competition takes place. However, variables that influence at shorter time scales (e.g. match location, half, score, team formation, etc.) have also been extensively studied. Traditionally, these variables have been represented as frequencies of individual actions or aggregations of collective behaviour derived from event and tracking data respectively. However, this type of analysis makes it difficult to consider the context in the description of the results. Consequently, recent statistical models have emphasised estimating the value of those actions (Decroos INTRODUCTION
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
16 Furthermore, the ball location of the team that is in possession of challenge for both data and sports
between the different confrontation the ball. So much so that FC Barcelona scientists. This will be best achieved
lines (Castellano et al. 2007) and has been clearly differentiated from through collaborative efforts.
the depth of these lines will also other teams (Gyarmati et al. 2014)
determine more specific contexts of because of how the team makes use In summary, one of the challenges
interaction. These structures can be of it through passing, as an action by for researchers and practitioners is
practised and learnt over the time which the players relate and interact. to break down the linguistic barriers
scale of months, but can change in the Although the Barça style of play has between the areas of knowledge or
blink of an eye. been characterised by high passing fields of science in order to allow
percentages, this is not an end in better and more efficient information
Identifying the most common itself, but a means through which the transmission. The adoption of a
formations during these game players organise themselves around transdisciplinary theoretical approach
contests, and the changes derived the ball. will allow a better understanding of
from the transitions between them, all scales of the emergent processes
are some of the challenges that some It is precisely this organisation that arising from football environments
researchers are already beginning defines one of the most common (e.g. talent identification and
to address. The stability of these drills during the club’s team training development, performance analysis,
formations, and also derived from sessions, the “rondo”. Despite the injury risk assessment, etc.). The
the values of width and depth of extensive research on the use of identification of essential variables
the team, dispersion of the players small-sided games (Clemente et al. through current analytical techniques
respect to their team centroid and the 2020), the effects from rondos or (e.g. machine learning) from the
depth of the rear line, characterise its other common training drills such inductive approach would allow the
organisation. In turn, this definition as positional games (Casamichana modelling of simulations derived
makes possible to characterise Gómez et al. 2018) on player behaviour from those variables (deductive
defensive disruption as a function or the dynamics of the coordination approach) to capture the system’s
of the variation from the previous between them are still unknown. performance/intelligence (Hristovski &
set of variables, allowing players’ Little research has addressed the Balagué, 2020). The conceptualisation,
performance to be classified according need to identify values, parameters quantification and modelling, through
to the impact that their passes have or patterns of play that allow the entropy, could one day elucidate a
on the opponent’s defensive (dis) design of training scenarios that are paradigm shift in sports performance
organisation (Goes et al. 2019). representative of the game. There are that might allow its unification at
Most football teams are organised already some publications which, all levels of analysis. Under the
according to the location of the goals, through the identification of physical nested relationship between levels,
because of two main objectives: to demands (e.g. distance covered) the context-dependent analysis of
score a goal and protect their own according to the game context, performance will be presented in the
goal. However, the location of the help to design drills that fit with the following chapters by offering to the
ball is essential to define the context identified demands (Bradley & Ade, reader a wide view of the research
of the game and, in some cases, it 2018). However, making training carried out by some of the most
is the element that determines the drills that meet not only physical but renowned experts from that topic.
organisation of the players, regardless also behavioural demands is still a
INTRODUCTION
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
REFERENCES
Bradley, P. S., & Ade, Probabilistic Model of nisms: Implications org/10.4324/978020 Psychology, 7. https://
J. D. (2018). Are the Development of of Sociobiological 3806913 doi.org/10.3389/ 17
Current Physical Play in Soccer. Quality Models of Behaviour fpsyg.2016.01513
Gyarmati, L., Kwak, H.,
Match Performance and Quantity, 41(1), for Research and
& Rodriguez, P. (2014).
Metrics in Elite Soccer 93–104. https://doi. Practice in Team
Searching for a Uni-
Fit for Purpose or is org/10.1007/s11135- Sports Performance
que Style in Soccer.
the Adoption of an 005-3148-0 Analysis. Sports
2014 KDD Workshop
Integrated Approach Medicine, 42(8),
Castellano, J., & Pic, on Large-Scale
Needed? International 633–642. https://doi.
M. (2019). Identifica- Sports Analytics.
Journal of Sports org/10.2165/
tion and Preference arxiv:1409.0308
Physiology and 11632450-
Performance, 13(5), of Game Styles in 000000000-00000 Hristovski, R., &
656–664. https:// LaLiga Associated Balagué, N. (2020).
with Match Outco- Fernández, J., &
doi.org/10.1123/ Theory of Cooperati-
mes. International Bornn, L. (2020).
ijspp.2017-0433 ve-Competitive Inte-
Journal of Environ- SoccerMap: A Deep
lligence: Principles,
Bransen, L., Robbere- mental Research and Learning Architecture
Research Directions,
chts, Pi., Van Haaren, Public Health, 16(24), for Visually-Inter-
and Applications.
J., & Davis, J. (2019). 5090. https://doi. pretable Analysis in
Frontiers in Psycholo-
Choke or Shine? org/10.3390/ijer- Soccer. Proceedings
gy, 11, 2220. https://
Quantifying Soccer ph16245090 of the 2020 European
doi.org/10.3389/
Players’ Abilities Conference on Machi-
Clemente, F. M., fpsyg.2020.02220
to Perform Under ne Learning and Prin-
Mental Pressure. Afonso, J., Castillo, ciples and Practice of Linke, D., Link, D.,
Proceedings of the D., Arcos, A. L., Silva, Knowledge Discovery & Lames, M. (2018).
13th MIT Sloan Sports A. F., & Sarmento, H. in Databases, 1–16. Validation of Elec-
Analytics Conference, (2020). The effects https://bitbucket.org/ tronic Performance
1–25. http://www. of Small-Sided ghentdatascience/ and Tracking Systems
sloansportsconferen- Soccer Games on ecmlpkdd20-papers/ (EPTS) Under Field
ce.com/wp-content/ Tactical Behavior and raw/master/ADS/ Conditions. PLoS ONE,
uploads/2019/02/ Collective Dynamics: sub_1006.pdf 13(7). https://doi.
Choke-or-Shi- A Systematic Review. org/10.1371/journal.
Chaos, Solitons and Goes, F. R., Kempe,
ne-Quantifying-Soc- pone.0199519
Fractals. https:// M., Meerhoff, L. A., &
cer-Players-Abili-
doi.org/10.1016/j. Lemmink, K. A. P. M. Pol, R., Balagué, N.,
ties-to-Perform-Un-
chaos.2020.109710 (2019). Not Every Pass Ric, A., Torrents, C.,
der-Mental-Pressure.
Can Be an Assist: A Kiely, J., & Hristovski,
pdf Decroos, T., Van Data-Driven Model to R. (2020). Training or
Casamichana Gómez, Haaren, J., Bransen, Measure Pass Effecti- Synergizing? Complex
D., Gómez Diaz, A., L., & Davis, J. (2019). veness in Professional Systems Principles
Cos Morera, F., & Mar- Actions Speak Louder Soccer Matches. Big Change the Unders-
tín García, A. (2018). Than Goals: Valuing Data, 7(1), 57–70. ht- tanding of Sport
Wildcard Players Player Actions in tps://doi.org/10.1089/ Processes. Sports
During Positional Ga- Soccer. Proceedings big.2018.0067 Medicine - Open,
mes. Apunts. Educa- of the ACM SIGKDD 6(1), 1–13. https://doi.
International Confe- Gómez, M. Á.,
cion Fisica y Deportes, org/10.1186/s40798-
rence on Knowledge Lago-Peñas, C., &
133, 85–97. https:// 020-00256-9
Discovery and Data Pollard, R. (2013).
doi.org/10.5672/
Mining. https://doi. Situational Variables. Ric, A., Torrents,
apunts.2014-0983.
org/10.1145/329250 In T. McGarry, P. C., Gonçalves, B.,
es.(2018/3).133.06
0.3330758 O’Donoghue, & J. Sampaio, J., &
Castellano, J., Sampaio (Eds.), Hristovski, R. (2016).
Hernández-Mendo, Duarte, R., Arajo, D., Routledge Handbook Soft-Assembled
A., Morales-Sánchez, Correia, V., & Davids, of Sports Perfor- Multilevel Dynamics
V., & Anguera, M. T. K. (2012). Sports mance Analysis (pp. of Tactical Behaviors
(2007). Optimising a Teams as Superorga- 277–287). https://doi. in Soccer. Frontiers in
INTRODUCTION
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
18 Always
think before
computing!
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
— Rafel Pol1,2 & Natàlia Balagué3 1: Real Federación
Española de Fútbol
(RFEF)
INTRODUCTION
2: Complex Systems
in Sport Research
Group, Institut Nacional
d’Educació Física de
Catalunya (INEFC),
What could we tell analysts to help them better connect theory and practice? University of Lleida,
Let’s start from the beginning: why do we analyse the game? Spain
3: Complex Systems
in Sport Research
Group, Institut Nacional
d’Educació Física de
Catalunya (INEFC),
University of Barcelona,
Spain
19
R: There are three types of analysis that
practically all professional clubs apply
today. The first is oriented toward the
scouting of talented players that the
club may want to sign. This entails
analysing the market to gain insight and
decide whether or not players might be
an attractive choice for certain positions.
The second is analysis of opponents
(teams or players), which is aimed at
understanding the characteristics and
weaknesses of rivals to prepare for
competition against them. The third is
player performance analysis, which is
currently focused on kinematic variables
(e.g. total distance covered, velocity,
etc.). I am not aware of any elite club in
the world that does not use this level of
analysis.
N: This means that there are three
organisational levels (club, team
and player) that are involved in such
performance analysis. Any decision
made at a higher level (club), such as
signing new players, affect those made
at lower levels (e.g. team strategies,
players actions) and vice versa. That
is, the composition of the roster will
impact playing systems and competitive
strategies, which, in turn, bind the
performance and actions of individual
players (Balagué et al. 2019); Pol et al.
2020). Likewise, lower-level events
(e.g. an injured or underperforming
player) shape higher-level ones (change
of playing strategies, signing of new
players). It is important to take into
account that these three levels cannot
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
20
be understood in isolation when skilled dribbler may be invaluable in she won’t just lose her/his reputation
thinking about performance analysis. a team that covers a lot of distance as the best centre-back but will come
due to its playing style. However, if we off as a shoddy defender. In contrast,
R: In fact, these three levels share a put that same player in a team that consider a fast and small centre-back,
common limitation, inherited from spends a lot of time in possession of such as Javier Mascherano, who
an overly reductionist brand of sports the ball and that generally forces the played with Barça. In the context of
science: they underestimate context. opponent to retreat, s/he will lose that team, he was undoubtedly one
that running space and, suddenly, of the best centre-backs in the world.
a. Club level: When signing players, no longer seem like such a skilled However, he probably would have
we often forget that some of them dribbler. Likewise, we might say that a had more difficulty performing at
perform better in certain clubs player is the best centre-back without that level had he been on a team that
and worse in others. Big data and specifying that his or her team plays played defence closer to its own goal.
statistics are used to confirm that a close to its own goal, leaving little In fact, when playing with Argentina,
player is good at performing certain space behind them. This player might which does indeed play defence
actions; it is then inferred that s/he be very good at confrontation but closer to their own goal, he was more
will perform other actions equally might not be a very fast runner. If s/ of a midfielder than a centre-back.
well. But we must also examine the he is signed by a team that plays Players who are deemed the best in
context in which s/he is performing in possession of the ball in the their position may end up seeming
these actions exceptionally well. opponent’s half, s/he will be forced second-rate when the context is
There are also other areas outside to run long distances, face multiple changed.
the context of the match that allow players at once in a larger space and
b. Team: When analysing an opponent,
us to predict whether or not a player enjoy fewer one-on-one challenges
there is a general tendency to make
will adapt well to the club. A fast and in her/his territory. As a result, he or
statements such as “this team usually
CHAPTER 1FOOTBALL ANALYTICS 2021 THE ROLE OF CONTEXT IN TRANSFERRING ANALYTICS TO THE PITCH presses in this way” or “they play quantification is very uninformative. For We can gather a great deal of 21 with the ball like this”. The idea is example, the simplest way to objectify information through conversations to offer an average, or a general perceived exertion (PE) is with the Borg with athletes about how they feel in rule, of how the team plays. But scale (RPE 6-20 or CR-10) (Borg, 1998). similar situations. This also may help “this is how they press” can only be Based on my experience in different the player become more aware of her/ confirmed when the opponent does situations and with different teams, the his condition and learn to express it as not force them to play differently. information obtained from PE ratings a comparison to her/himself and not For example, this statement will means very little if the players do not to an external scale. But I think that if I no longer hold true when playing value this type of monitoring and if it is wanted to publish these data I would against a team with the resources not adequately contextualised. We must be required to quantify them based to resist such pressure. Identifying encourage and harness athletes’ ability on an external universal reference an “average” or a “summary” in this to perceive and integrate information to objectify the differences between way does not provide information and put it at the service of monitoring. performance on different days and about what a team will do in a These are exceptional means of injury across different weeks. That would context in which their standard prevention and promote autonomy and be difficult to publish, don’t you think, playing style is no longer effective. self-management of workloads (Pol et Natalia? This analysis, therefore, becomes al. 2018). useless if, as a team, we manage to overcome our rival. And since that is the general objective, it is useless all around. Nevertheless, averages are compiled for teams to identify their characteristics and label them, despite the fact that these characteristics will disappear as soon as the context of the game changes. c. Player: When analysing the performance of individual players, we once again seek out these averages and compare them with the team average, sometimes with the player averages and sometimes with what is typical in that level of competition. These averages offer little information about whether or not a player is capable of adapting to the demands arising from the context of the match. At all three levels of analysis there is an attempt to extract quantitative data and little emphasis on the context. Absolute, decontextualised CHAPTER 1
FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
22
N: Right, sports sciences place a depending on the context. Since Changing Perspectives in Sport
greater emphasis on quantitative every competition and every rival Performance,” (Balagué y Torrents,
research, easier to objectify, and relies is different, it would be a mistake 2005). It might be useful to look
too heavily on group data averages. to compare kinematic information, back on its message to clarify what
obtained either at individual or it means performance analysis from
However, it is possible to collective level, without an adequate a complex perspective. Certain study
operationalise psychological or contextualisation and interpretation. variables such as entropy, big data or
sociological information extracted In this sense, examining the dynamics more advanced processing techniques
from interviews and focus on of change of the variables under (e.g. deep learning) are generally
individual data obtained from time study is more informative to recognise associated with the analysis of the
series (Molenaar, 2004; Nesselroade whether performance stagnation or game as a complex phenomenon.
and Molenaar, 2010: Rose et al. 2013). deterioration is taking place. This is However, processing large volumes
The main experimental model used applicable to all the aforementioned of data and variables compiled from
to test research hypotheses in sports levels and requires updated these techniques is of little use if
sciences is the comparison of group assumptions and deep theoretical there is not accompanied by a deep
data means through inferential understanding of the examined understanding of the process under
statistics. The fact that the averages processes. study. It must also be noted that
obtained do not correspond to any merely processing common variables
specific individual is generally not R: Most clubs have entire departments in the study of complex systems does
taken into account when interpreting devoted to generating analysis tools not necessarily mean that a complex
and applying research results. In fact, that, in practice, add no value to the analysis of the game is performed.
neglecting the idiosyncrasy of the work of the coaches. The problem For example, the entropy measures,
structure of time-dependent variations lies in developing applications or usually associated with the study
within a single individual and reports without first thinking about the of complex systems, can be used
assuming the equivalence between underlying theoretical assumptions. to answer a very simple question
inter- and intra-individual variability If the phenomena to be analysed are about cause and effect relationships
is a common mistake of this type not properly understood, it is difficult that has nothing to do with a
of research that leads to erroneous to create effective applications. The complex perspective. Essentially,
interpretations and inadequate first thing to do is to improve our in this perspective, we are not just
practical applications (Rose, 2016, understanding about the principles changing the type of data analysis
Balagué et al. in press). that lead the behaviour of the systems or study variables but the type of
we deal with: the game, the team, questions, assumptions and research
Ignoring the context in which complex the players. Only then a right answer interpretations. This is something we
adaptive systems interact can be very to the question “what do we need to need to emphasise so that resources
misleading. In particular, when such analyse?” will emerge. With the current are not wasted. Otherwise, the same
context is non-stationary and fast approach, countless resources are mistake will be made over and
changing, as during sports games. For being wasted. over, just with more sophisticated
example, the total distance covered technology and more data or new
by a player or her/his running velocity N: Years ago, we published an article variables, but still without improving
may be functional or un-functional titled “Thinking Before Computing: the understanding of the phenomena
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
Statistics itself, which
does have clear
limitations, is not the
main problem
23
under study. effects (Amrhein et al. 2019). Even main problem. It is the misuse and
the American Statistical Association misinterpretation of statistics in an
R: In fact, this lack of reasoning is has spoken out against the misuse oversimplified research context. There
more common in science than you of inferential statistics and its role is a general belief that exceeding a
might think. The journal Nature in promoting erroneous and biased threshold for statistical significance
recently published an article findings (Wasserstein et al. 2016; means a result is real (Wasserstein
signed by 850 scholars of different McShane et al. 2019). This tendency is et al. 2016; Amrhein et al. 2019,
academic ranks that denounces not exclusive to performance analysis, McShane et al. 2019). Without a deep
the misuse of statistics in research. but pervasive throughout sports understanding of the assumptions
Specifically, it calls for the retirement science. behind the research hypotheses,
of statistical significance to put methodological decisions, data
an end to publications that report N: Statistics itself, which does analysis and interpretation, the
“hyped” results and dismiss crucial have clear limitations, is not the research is no longer practical but
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
rather may add to the confusion and that merely creating multidisciplinary to the concept of constraints but to all
24
harm the advance of knowledge. If teams will not overcome the obstacles the DST concepts; valid for all levels of
we do not understand the theory, of studying a phenomenon that analysis (see Fig. 1).
we cannot propose useful practical belongs to no specific discipline:
applications. We need to be aware neither the game nor the player fit However, despite having powerful
of the norm in sports science: within these disciplines. Therefore, tools such as Dynamic Systems Theory,
isolating variables, comparing it is not just about adding new it is quite challenging to evaluate the
group averages, obtaining results professionals and perspectives from dynamics of change in collaborative
based on statistical significance different branches, but about sharing and competitive processes in which
and putting them into practice as a common approach to how the rivals never behave in exactly the
general truths. All this represents a systems that we work with function same way and there are no static
significant bias for the advancement – whether players or teams – and situations, as in football.
of knowledge. The reductionist making contributions from that
tendency, inherited from medicine common vision. R: Perhaps we should change the
and psychology, and over-simplified type of question or the objective of
theoretical assumptions limit the N: Right, there are transdisciplinary performance analysis. If we have a
evolution of our understanding of initiatives that pursue real knowledge deep understanding of the dynamic
complex phenomena linked to sports integration, but multidisciplinarity mechanisms of players/teams to
performance. is in fact non integrative (Balagué adapt to the context of a match or
et al. 2017). It is necessary to clarify training session, and develop them
It is possible, however, to turn what a real integration means, and during training sessions, we could
away from tradition and opt for which type of approaches may even forget about analysing the
other models, such as those of satisfy it. Some authors suggest that opponent. The issue is that training
complex systems, widely applied training-methodology departments sessions are often focused on the rival,
in several emerging branches of should integrate knowledge on the on their weaknesses and strengths as
biology. Of course, this option does basis of an ecological dynamics detected from analysing their game.
mean overcoming the resistance approach (Rothwell et al. 2020). This In response, optimal, predetermined
and difficulties that come with any perspective, however, focuses on solutions are prepared. But if we
paradigm shift. We now find ourselves the level of perception and action, changed the objective from having
in the midst of that transition, forcing while neglecting other relevant the team or player learn specific
models that are to some extent levels related to strength and solutions to improving their potential
incompatible to coexist. However, conditioning (e.g., biochemical and for adaptation and diversity (Pol et
the difficulties associated with such physiological). In line with trans- or al. 2020) and their ability to resolve
paradigm shifts should not discourage cross-disciplinarity, some authors unpredictable situations – such as
the pioneers because they are on the sustain that integration requires a those that happen regardless of how
right track. Interdisciplinary work is common scientific vocabulary shared we train – we would not even need to
being encouraged in many scientific across disciplines (Balagué et al. 2017); analyse the opponent or, at least, not
branches and the contributions of Hristovski et al. 2017). In this direction, in the way we currently do. We could
physicists, mathematicians and Glazier (2017) proposes the unification focus instead on analysing the state
computer scientists, collaborating of knowledge based on the concept of of our team and deciding whether
with sports scientists, are key in constraints and Balagué et al. (2017) we should generate new training
performance analysis. suggest the Dynamic Systems Theory challenges or a sense of confidence by
as an ideal integrative framework. DST strengthening scenarios already under
R: However, it is important to highlight does not limit the common vocabulary control. But, of course, for that we
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
would need the deep understanding that is not accustomed to competing
25
of the theoretical foundations of at a high frequency qualifies for a
performance that we mentioned European competition, is that the
earlier. coaches underestimate that they
can no longer train as much as they
N: Even separating training and used to. If they do not address the
competition would no longer make process as a whole, understand that
sense. There is training in competition competition is part of the process,
and competition in training, so and eliminate a significant part of the
why shouldn’t we view them as training, they end up overwhelming
continuous, adaptive processes with the team. Even if the coach is
disturbances and constraints that are changed, it becomes very difficult to
more or less intense and/or global? revive the team. This situation repeats
itself every year. Figure 1.
R: In fact, training while competing is Structure and organisation of
the best type of training; this becomes N: There is a fundamental problem scientific fields sharing the same
concepts and principles from Dyna-
very clear in teams that play every in our understanding about the mic Systems Theory.
three days. One of the main problems systems we deal with: it is generally v
that usually crop up after a team assumed that they are dominated by
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
More information
does not mean more
knowledge
26 the dynamics of their components, put into question. These common For example, specialisation does not
instead of being dominated by the and artificially built barriers can be encompass the fact that the levels
dynamics of the interactions among reduced under the proposed paradigm of analysis previously mentioned
such components, as occurs in shift, and specifically, under the – club, team and player – do not
complex systems (Delignières and umbrella offered by the Dynamic exist independently of each other
Marmelat, 2012). This explains why Systems Theory. but are related by circular causality
training focuses on components and (Balagué et al. 2019). In cooperative
underestimates the counterproductive R: The key to knowledge integration and competitive environments,
effects this may have on competition. is a shared understanding of the maintaining and developing the
For example, fragmenting and phenomenon to be studied. You can potential for functional diversity/
stabilising excessively positional play have a team comprising professionals unpredictability – also defined as
in attack and defence (separately) from different disciplines working intelligent cooperative-competitive
during trainings may harm the on an integrated job, but if the behaviour – at different levels is
transitions during competitions. biomechanist tells you that a player crucial (Hristovski and Balagué,
Moreover, there is a crucial missed a pass because s/he shot from 2020). In other words, having a
training goal in team sports that the wrong angle, while the physical smart club, a smart team, and smart
is often underestimated by this trainer says that s/he was worn out players means evading and escaping
component-dominant approach: by the end of the game because s/he situations of reduced possibilities of
developing interpersonal synergies played for too many minutes and did goal achievement in different contexts.
among players. These synergies, not rest enough, and the physiologist Opponents try to reduce that diversity
which are the basis of the team’s agrees with the physical trainer and potential intervening at all levels, and
functional diversity potential and adds that the results of the last blood the goal is to maintain and develop it
competitiveness (Pol et al. 2020), analysis were not good enough, what despite these disruptions. Functional
require collective challenges and do you do? You say you are doing players are those that contribute to
constant interaction between multidisciplinary and integrated work the collective goal, and functional
players to overcome them. If training because every specialist provides teams are those that escape and
objectives change their focus, the line their view on the phenomenon, but quickly regain the diversity potential
between training and competition will if all these specialists understand it despite their opponents’ actions. One
be blurred. Technology must serve differently, and look at it from their fundamental advantage is that this
that understanding, not the other way own perspective, it becomes very potential can be measured using a
around. Different departments of the difficult to generate new knowledge. common variable: information entropy.
club, even members of the coaching In such cases, a multidisciplinary team
staff, often analyse competition is of little or no use and may even R: In conclusion, to improve the
and training separately. That is, the detract from the work. If we do not methodologies of game analysis
performance department and the change our understanding, any other we must start by improving our
physical trainers focus on analysing efforts we make are useless. understanding of the phenomena
players and their physical and under study. When our understanding
physiological response during training N: More information does not changes, the questions we ask
through GPS data, and analysis mean more knowledge. Scientific ourselves change; and when the
departments focus on the competition specialisation has provided a lot questions change, the way we analyse
(event data and movement tracking) of detailed information, but this phenomena necessarily changes as
through video analysis. Perhaps does not necessarily imply greater well.
this fragmentation should also be comprehension and knowledge.
CHAPTER 1FOOTBALL ANALYTICS 2021
THE ROLE OF CONTEXT IN TRANSFERRING
ANALYTICS TO THE PITCH
REFERENCES
Amrhein, V., Green- CritRevBiome- with P-Technique
land, S., & McShane, B. dEng.2013006727 Factor Analysis and 27
(2019). Scientists Rise the Idiographic Filter.
Glazier P. (2017).
Up Against Statistical Behavior Genetics,
Towards a Grand
Significance. Nature, 40(6), 776-783.
Unifying Theory of
567, 305-307. https://
Sports Performance. Rose, T. (2016). The
doi.org/10.1038/
Hum. Mov. Sci., End of Average: How
d41586-019-00857-9
56:139–56. https:// to Succeed In a World
Balagué, N. y Torrents, doi.org/10.1016/j. That Values Sameness.
C. (2005). Thinking humov.2015.08.001 USA: HarperCollins
Before Computing:
Hristovski, R., Aceski, Rothwell, M., Davids,
Changing Perspectives
A., Balagué, N., Seifert, K., Stone, J. A., O’Sulli-
in Sport Performance.
L., Tufekcievski, A., van, M., Vaughan, J.,
International Journal
Aguirre, C. (2017). Newcombe, D.J., Shu-
of Computer Science in
Structure and Dy- ttleworth, R. (2020). A
Sport, 4, 2, 5-13.
namics of European Department of Metho-
Balagué, N., Torrents, Sports Science Textual dology Can Coordinate
C., Hristovski, R., Kelso, Contents: Analysis Transdisciplinary
JAS. (2017). Sport Of ECSS Abstracts Sport Science Support.
Science Integration: (1996–2014). European Journal of Expertise, 3,
An Evolutionary Journal of Sport Scien- 1, 55-65. http://shura.
Synthesis. European ce, 17, 1, 19-29. http:// shu.ac.uk/25980/
Journal of Sport Scien- dx.doi.org/10.1080/174
Wasserstein, R. L., &
ce, 17, 1, 51-62. https:// 61391.2016.1207709
Lazar, N. A. (2016).
doi.org/10.1080/17461
Hristovski, R., Balagué, The ASA’s Statement
391.2016.1198422
N. (2020) Theory of on p-Values: Context,
Balagué, N., Pol, R., Cooperative-Com- Process, and Purpose.
Hristovski, R., Torrents, petitive Intelligence: American Statistician,
C., Ric, A. (2019). On Principles, Research 70(2), 129-133. https://
The Relatedness and Directions and Appli- doi.org/10.1080/00031
Nestedness of Cons- cations. Frontiers in 305.2016.1154108
traints. Sports Medici- Psychology, 11. https://
ne Open, 5, 6. https:// doi.org/10.3389/
doi.org/10.1186/ fpsyg.2020.02220
s40798-019-0178-z
McShane, B. B., Gal, D.,
Balagué, N., Hristovski, Gelman, A., Robert, C.,
R., Almarcha, M. C., & Tackett, J. L. (2019).
García-Retortillo, S., Abandon Statistical
Ivanov, P. (in press) Significance. American
Network Physiology Statistician, 73(sup1),
of Exercise: Vision and 235–245. https://doi.or
Perspectives. Frontiers g/10.1080/00031305.2
in Physiology. 018.1527253).
Borg, G. A. (1998). Molenaar, P. C. (2004).
Borg’s Perceived A Manifesto on Psy-
Exertion and Pain Sca- chology as Idiographic
les. Champaign (IL): Science: Bringing
Human Kinetics. the Person Back Into
Scientific Psychology,
Delignieres, D., and
This Time Forever.
Marmelat, V. (2012).
Measurement, 2(4):
Fractal Fluctuations
201-218.
and Complexity:
Current Debates and Nesselroade, J. R.,
Future Challenges. Molenaar, P. C. (2010).
Crit. Rev. Biomed. Eng. Analyzing Intra-Per-
40, 485–500. https:// son Variation: Hybri-
doi.org/10.1615/ dizing the ACE Model
CHAPTER 1You can also read
