Content uploaded by Yuesen Li
Author content
All content in this area was uploaded by Yuesen Li on May 11, 2022
Content may be subject to copyright.
Characterizing player’s playing styles based on Player Vectors for
each playing position in the Chinese Football Super League
Yuesen Li1, Shouxin Zong1, Yanfei Shen1*, Zhiqiang Pu2, Miguel-Ángel Gómez3,
Yixiong Cui1,4*
1. School of Sports Engineering, Beijing Sport University, Beijing, China.
2. Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China
3. Facultad de Actividad Física y del Deporte (INEF), Universidad Politécnica de
Madrid, Madrid, Spain.
4. AI Sports Engineering Lab, School of Sports Engineering, Beijing Sport University,
Beijing, China.
Correspondence: Yixiong Cui, Yanfei Shen; School of Sports Engineering, Beijing
Sport University, Information Road 48, Haidian District, 100084, Beijing, People’s
Republic of China. E-mail: (amtf000cui@gmail.com; syf@bsu.edu.cn).
Abstract
Characterizing playing style is important for football clubs on scouting, monitoring
and match preparation. Previous studies considered a player’s style as a combination of
technical performances, failing to consider the spatial information. Therefore, this study
aimed to characterize the playing styles of each playing position in the Chinese Football
Super League (CSL) matches, integrating a recently adopted Player Vectors framework.
Data of 960 matches from 2016-2019 CSL were used. Match ratings, and ten types of
match events with the corresponding coordinates for all the lineup players whose on-
pitch time exceeded 45 minutes were extracted. Players were first clustered into 8
positions. A player vector was constructed for each player in each match based on the
Player Vectors using Nonnegative Matrix Factorization (NMF). Another NMF process
was run on the player vectors to extract different types of playing styles. The resulting
player vectors discovered 18 different playing styles in the CSL. Six performance
indicators of each style were investigated to observe their contributions. In general, the
playing styles of forwards and midfielders are in line with football performance
evolution trends, while the styles of defenders should be reconsidered. Multifunctional
playing styles were also found in high rated CSL players.
Key words: soccer; match analysis; style of play; performance profile; machine
learning
1. Introduction
The research of performance analysis in association football covers various
aspects such as the identification of key indicators, movement patterns or passing
networks (Lord et al., 2020). Among the wide scope of this research field, the evaluation
of teams’ and players’ match performance has recently become one of the essential
areas that attract scientific attention to improve the body of research (Goes et al., 2021;
Goes et al., 2020; Gudmundsson & Horton, 2017; Pappalardo et al., 2019). Relevant
findings and models could not only assist scouting experts and team stakeholders when
recruiting players, but also help coaches and managers to monitor the evolution of
tactical-technical styles in order to adjust match strategies and control for match
demands (Decroos & Davis, 2020; Zhou et al., 2021).
One of the most commonly used and developed evaluation tool is the rating and
ranking system of teams and players, which could be applied to assess the strength of a
team or a player objectively to a certain degree (Lasek et al., 2013). Technical and
tactical data-driven rating methods such as computing a weighted sum of subindices
(McHale et al., 2012), learning weights out of performance features (Pappalardo et al.,
2019, Li et al., 2020) and assessing behaviors by estimating probabilities (Decroos et
al., 2019) have all been recently proposed and studied. Apart from rating, characterizing
playing styles is also an essential perspective of evaluating, which would provide more
comprehensive insights into teams’ match-play and players’ actual roles, and help teams
to better recruit and avoid mismatches in the transfer market.
From a team level approach, a playing style can be defined as the interactions of
playing patterns (Hewitt et al., 2016) and researchers tends to combine certain attacking
and defending attributes, transitions and set pieces to characterize the style of play for
the purpose of winning (McLean et al., 2017). In particular, Fernandez-Navarro et al.
(2016) used factor analysis via principal components analysis (PCA) to define attacking
and defending styles in elite soccer leagues of Spanish La Liga and English Premier
League. Following the insight and the statistical approach of these studies, Lago-Peñas
et al. (2018) identified five different kinds of playing styles in Chinese Soccer Super
League: possession, set piece attack, counter attacking play and two kinds of
transitional play; while Gómez et al. (2018) compared the playing styles within
different team qualities and match locations based on the data of Greek professional
soccer and claimed that ranking and match location are also important when identifying
teams’ playing styles; and Castellano and Pic (2019) combined playing styles with
match outcomes in La Liga and found that well-designed offensive styles and defensive
minds are keys to win.
However, research on data-driven playing style characteristic at an individual level
is relatively scarce. Some studies have attempted to explore the player similarity in
match-play. For example, Peña and Navarro (2015) stipulated several passing
sequences and tried to find the replacement of the legendary midfielder “Xavi
Hernandez” using dimensionality reduction and unsupervised clustering on different
passing patterns. In addition, Mazurek (2018) searched for the most similar player to
Lionel Messi, using 17 advanced match indicators most relevant to an attacking player.
Moreover, Garcia-Aliaga et al. (2020) attempted to analyse patterns of play and key
performance indicators by position using machine learning methods and verified the
possibility of characterizing players’ positions using only game statistics. Although
these studies provided some preliminary findings, their practical usefulness were
limited due to the fact that a player’s style was only considered as a combination of
technical performances, failing to control for the spatial distribution of these actions.
In light of this phenomena, Decroos and Davis (2020) proposed a Player Vectors
model to express football player’s playing style based on the concept that a player’s
style can be expressed by his/her preferred areas on the field and which action he or she
tends to perform in each of these locations. Using a framework with Non-negative
Matrix Factorization (NMF) as the core, they extracted components from heatmaps,
representing different kinds of playing styles and the similarities of a player’s style with
these components being constructed to the final player vectors, thereby characterizing
playing styles.
Players in different roles have different responsibilities, previous research have
explored and detected player roles based on their active positions (Bialkowski et al.,
2014; Pappalardo et al., 2019). However, it should be noted that there is a wide variety
of playing styles in these positions, and they should play as different roles. Previous
research (Aalbers & Van Haaren, 2019) has classified players European top leagues into
various playing styles using machine learning models, but the players were initially
labeled by experts before the modeling, which might limit the applicability to players
of other tournaments. Therefore, this study aimed to provide a data and domain
knowledge driven solution to characterize players’ playing styles in different positions
in professional football, integrating the state-of-the-art method Player Vectors. Such
approach will allow to analyze the differences among playing styles of players in the
CSL and compare these playing styles with the current tactical requirements for players
in professional football.
2. Methods
2.1 Sample and data source
Data related to match events and player information of all 960 matches from 2016-
2019 CSL where 22 teams and 912 players competed were obtained from a publicly-
assessed football statistics website “whoscored.com”, whose original data are provided
by OptaSports Company. The validity of their data collection has been previously
validated (Liu et al., 2013).
The final match event data-set contains 44 types of actions related to shooting,
organizing, skill, defending and goalkeeping. Each action has its own result and
attribute (See Supplementary Table 1 for detailed definitions).
2.2 Player Vectors
Fixed-size player vectors that characterizes players’ playing style were built
followed by the framework showed in Figure 1.
The framework contains three major parts: Constructing, Compressing and
Assembling. The Constructing part draws normalized smooth heatmaps with the size of
for each player of action . Then, the Compressing part applied a non-negative
matrix factorization (NMF) process to compress the matrix which
columns are the vectorized 1-dimensional heatmap vector with the length of
for each player.
Figure 1. Flowchart of the Player Vectors framework
NMF is a form of PCA where the resulting components contain only positive
values (Lee & Seung, 2000). This results in two matrices and such that:
Where
,
,
. is a user-defined parameter that
refers to the number of principal components for action type . The columns in are
the principal components that represent feature heatmaps. The rows in are
vectors that are the compressed heatmaps in which can be seen as the player playing
style’s similarities to each feature heatmaps. The actions used in this study is showed
in Table 1.
Table 1. Actions used in the Player Vectors framework.
Actions
Definitions
Shot
An attempt to score a goal, made with any (legal) part of the body, either on or off
target
Cross
Any intentional played ball from a wide position intending to reach a teammate in
a specific area in front of the goal.
Dribble
Player dribbles at least 3 meters with the ball.
Pass
Any intentional played ball from one player to another.
Long pass
A lofted ball where there is a clear intended recipient, must be over shoulder height
and using the passes height to avoid opposition players or a long high ball into
space or into an area for players to chase or challenge for the ball
Key pass
The final pass or pass-cum-shot leading to the recipient of the ball having an
attempt at goal.
Interception
Where a player reads an opponent’s pass and intercepts the ball by moving into the
line of the intended pass.
Clearance
A defensive action where a player kicks the ball away from his own goal with no
intended recipient.
Header
Any action using a player’s head, whether it is a pass or a dual in the air.
Recovery
Where a player recovers the ball in a situation where neither team has possession
or where the ball has been played directly to him by an opponent, thus securing
possession for their team.
The assembling part concatenates the for each action type to
and the player vector of a player is the corresponding row in
.
Finally, the similarity of the playing style of each player was calculated as the
Manhattan distance between the player vectors. To illustrate the similarity by
percentage, it was further transformed by:
where is the Manhattan distance between player i and j, is the maximize
distance between each pair of players in data-set.
2.3 Position clustering
A player’s position can be changed in different matches (Pappalardo et al., 2019)
and the actual position is sometimes varied to which given in the starting lineups. To
eliminate the resulting potential impacts as much as possible, it is necessary to detect
the real position of the players.
A total 18,784 of 19,200 average positions of the actions of every starting player
who played more than 45 minutes in each match were extracted from the data-set. A k-
means algorithm was run on those average positions. A silhouette score of each sample
was used to find out the most proper k value among 5 to 10, which is computed as:
where is the average distance between sample i to the other samples in the
cluster and is the distance between i and its nearest cluster (Pedregosa et al., 2011;
Rousseeuw, 1987).
In football matches, side switching is a classical tactic which will lead to a cluster
error for the wide players (being clustered to the center of the field). After finding only
87 times of side switching, these samples were deleted and led the final clustering
samples to 18,699. Further partition of the positions and style clustering were based on
the result of this position clustering process.
2.4 Style clustering
To extract different types of playing styles and the similarity between an outfield
player’s style of play and these representative styles, another n NMF processes were
carried out for the n different positions respectively using different actions. These were
acted on the starting lineups in each single match with goalkeepers excluded. The most
similar style was considered as the playing style of an individual player in a specific
match. Each playing style was defined based on their different preferences observed
from the Player Vectors and referred to a well-known sport video game—Football
Manager. Total appeared numbers (N), Appeared numbers of domestic players (ND)
and foreign players (NF), Win-Loss ratio (W/L), match ratings (R) and 3 performance
indicators (Goals, Shots and Assists) of each style were calculated to compare their
contributions. The numbers of the official starting position which were observed more
than 100 times in each style were also extracted (See Supplementary Table 2). Those
starting positions are Forward (FW), Midfielder (MF), Center Defender (CD),
Left/Right Midfielder (L/RM) and Left/Right Defender (L/RD).
2.5 Player Comparison
Comparing a player’s playing style before and after a specific scenario is valuable
and helpful on evaluating a player’s performance and a club’s strategic decision on
player transfer and recruitment. This study compared three pairs of player’s reports
representing three different scenarios:
(i) The most similar players: Alex Teixeira (Jiangsu Suning) vs. Wei Shihao
(Guangzhou Evergrande). Wei is the most similar player of Teixeira
according to the output of the model;
(ii) Player progressing: Wu lei (Shanghai SIPG) in season 2017 vs. Wu Lei
in season 2018. Wu is considered to be one of the best Chinese players and
won the Player of the season in 2018. Such comparison would help explore
how his match play progressed;
(iii) Midfield partners: Renato Augusto (Beijing Guoan) vs. Jonathan Viera
(Beijing Guoan). Being the two of only four players who performed double
figures for both goals and assists in season 2018 (Whoscored, 2018), these
two midfielders were clearly indispensable for Beijing Guoan. Comparing
their differences in playing style might offer a better understanding of their
tactical differences on the field.
Each player report contains five parts: (i) Player basic information; (ii) the
Similarity between them; (iii) Stats of the player’s styles in each match; (iv) the Figure
of the stats; and (v) the General player vector of the season.
3. Results
3.1 Player Vectors
The NMF process generated a player vector of 44 dimensions for each player, the
component heatmaps are showed in Figure 2.
Figure 2. The heatmaps of each component
3.2 Position clustering
Figure 3 (a) shows the position clusters resulted by the k-means algorithm with a
most proper k=8, which has a stable silhouette score (ss=0.41) among 10 times of
experiments with random sets of centroids that were used to initiate the algorithm. The
clusters located in the two sides were merged into one position manually which led to
the final positions reduced to five: Strikers (ST), Left/Right Wing forwards (L/RW),
Central Midfielders (CM), Left/Right Full Backs (L/RFB), and Central backs (CB). The
actions used in different positions in the following style clustering process were also
showed in Figure 3 (b).
Figure 3. Eight positions clustered by the K-means algorithm and the actions used in different
positions in style clustering
3.3 Style clustering
Table 2 shows the style clustering and descriptive (mean and standard deviation,
SD) results for each style. The definition of each style is illustrated in Supplementary
Table 2.
Table 2. Style clustering and descriptive statistic results of each style.
Position
Style
Appearances
Rating (SD)
Goals (SD)
Shots (SD)
Assists (SD)
Win/Loss
Total
Domestic
Foreign
ST
Second Striker
1394
340
1054
7.22 (.95)
0.36 (.59)
2.86 (2.05)
0.21 (.47)
1.01
Target Man
570
123
447
7.07 (.82)
0.34 (.56)
2.23 (1.49)
0.10 (.31)
0.88
Mobile Striker
646
243
403
7.07 (.88)
0.33 (.59)
2.22 (1.64)
0.15 (.39)
1.12
Poacher
613
154
459
7.38 (1.01)
0.65 (.80)
4.23 (2.01)
0.12 (.34)
1.05
CM
R Defensive Midfielder
760
595
165
6.70 (.59)
0.05 (.22)
0.80 (1.03)
0.05 (.22)
1.04
Playmaker
1549
1062
487
6.83 (.69)
0.09 (.32)
1.14 (1.21)
0.08 (.29)
0.83
L Defensive Midfielder
803
595
208
6.78 (.64)
0.09 (.31)
0.77 (1.01)
0.07 (.26)
0.98
Wide Midfielder
544
373
171
6.85 (.75)
0.14 (.39)
1.17 (1.20)
0.15 (.38)
1.18
L/RW
L Winger
1307
918
389
6.92 (.80)
0.19 (.46)
1.59 (1.48)
0.16 (.42)
1.08
R Winger
1224
928
296
6.85 (.75)
0.18 (.42)
1.46 (1.41)
0.14 (.38)
1.04
Inside Forward
890
450
440
7.01 (.87)
0.24 (.53)
2.03 (1.77)
0.14 (.40)
1.02
L/RFB
R Wing Back
1639
1609
30
6.65 (.59)
0.02 (.13)
0.35 (.63)
0.06 (.25)
1.01
L Wing Back
1613
1603
10
6.70 (.57)
0.01 (.12)
0.40 (.68)
0.07 (.27)
1.03
L Back
347
306
41
6.64 (.54)
0.04 (.21)
0.51 (.82)
0.05 (.21)
0.83
R Back
386
358
28
6.54 (.59)
0.03 (.17)
0.38 (.66)
0.04 (.20)
0.91
CB
R Ball Playing Defender
1750
1410
340
6.60 (.59)
0.03 (.18)
0.31 (.60)
0.01 (.12)
0.98
L Ball Playing Defender
1979
1406
573
6.68 (.61)
0.04 (.21)
0.39 (.68)
0.02 (.13)
0.96
Central Defender
685
510
175
6.80 (.63)
0.03 (.17)
0.30 (.59)
0.02 (.15)
1.66
Notes: ST = Strikers, L/RW = Left/Right Wing forwards, CM = Central Midfielders, L/RFB = Left/Right Full Backs, CB = Central Backs
Strikers
As it was showed in Figure 4 (a), The CSL Strikers were classified into four
playing styles: Poacher (preference for close shot and front header), Second Striker
(preference for Long Shot, Center Dribble and Center Pass), Mobile Striker (high
preference for L/R shot, L/R backline cross, L/R flank dribble and L/R flank pass) and
Target man (high preference for Mid front header). Among a total of 3,223 observations,
the best rated style was Poacher with the highest goals per match, followed by Second
Striker and Mobile Striker. On the other hand, Target Mans had the lowest Win/Loss
ratio among the four styles. A large amount of foreign MFs were found to be Second
Striker and Mobile Striker.
Central Midfielders
The Central Midfielders were clustered into four styles (see Figure 2 (b)):
Playmaker (preference for Dribble, Pass and Long pass in the middle of the field and
no preference for the defensive actions), L and R Defensive midfielders (preference for
Dribble, Pass, Long pass and Ball recovery in the back of the pitch) and Wide midfielder
(preference for Flank dribble, Flank pass, Flank long pass, Ball recovery and Aerial
dual in the mid front area). The best rated style among a total 3,656 observations was
Wide Midfielder who performed the best in goals and assist as well as the Win-Loss
ratio. The second-best rated style was Playmaker while its Win-Loss ratio was the
lowest. R and L defensive Midfielders’ Win-Loss ratio was relatively balanced. The
starting positions with more than 100 observations in all these styles are solely MF.
Left/Right Wing forwards
A total number of 3,421 L and R Wing Forwards were clustered into three styles
which are showed in Figure 2 (c): Inside Forward (preference for Center dribble and
pass, Close shot and Long shot) and L/R Winger (preference for Cross in the flanks and
backline and dribble, pass in the flanks). Among them, Inside Forward got the highest
rating. Most of the L/R Wingers were domestic players, while foreign FWs and
domestic MFs were the majority of Inside Forwards.
Left/Right Full Backs
As it was illustrated in Figure 2 (d), 3985 L and R Full Backs were classified into
four playing styles: L/R Wing Backs (preference for Flank dribbles, passes and Flank,
Backline crosses) and L/R Backs (preference for Back dribbles, passes and recoveries).
L/R Wing Backs performed better both in rating and Win-Loss ratio than L/R Backs.
Almost all Full Backs were domestic and played on their original starting position
(L/RD).
Central backs
Central backs were clustered into three different styles (see Figure 2 (e)): L and R
Ball Playing Defender (preference for Back dribbles, passes and long passes) and
Central Defender (preference for Back header and clearances in all four areas and little
preference for dribbles and passes). Among a total 4,414 observations, Central
Defender had the best rating and Win-Loss ratio. DF was the main part of these three
playing styles where most of the players were domestic.
Figure 4. Playing styles of each position. (Notes: CS = Close Shot, LS = Long Shot, RS = Right
Shot, LS = Left Shot, LBC = L. Backline Cross, RBC = R. Backline Cross, RFC = R. Flank
Cross, LFC = L. Flank Cross, CD = Center Dribble, RFD = R. Flank Dribble, LFD = L. Flank
Dribble, LBD = L. Back Dribble, RBD = R. Back Dribble, LBP = L. Back Pass, RFP = R. Flank
Pass, LFP = L. Flank Pass, CP = Center Pass, RBP = R. Back Pass, LBLP = L. back long pass,
LFLP = L. flank long pass, RFLP = R. flank long pass, CLP = Center long pass, RBLP = R.
back long pass, RFKP = R. far key pass, RKP = R. key pass, LKP = L. key pass, LFKP = L. far
key pass, LBI = L. back interception, RFI = R. flank interception, LFI = L. flank interception,
RBI = R. back interception, MC = Middle clearance, LC = Left clearance, IC = Inner clearance,
RC = Right clearance, BH = Back Header, MFH = Mid Front Header, RBH = R. Back Header,
LBH = L. Back Header, FH = Front Header, LBR = L. back recovery, RFR = R. flank recovery,
LFR = L. flank recovery, RBR = R. back recovery)
3.4 Player comparisons
Alex Teixeira VS. Wei Shihao
Figure 5 (a) shows the comparison report of Teixeira and Wei Shihao, whose
similarity was 93.6%. Clearly the most appeared and rated style of Teixeira was Second
Striker of ST, with an appearance of 13 times and a high average rating of 7.62 points
per game. The second most played style of Teixeira was L Winger, which was the most
appeared style of Wei Shihao, with a rating of 7.18. As for the season general style,
Teixeira had a much higher preference for Long shot, Center dribble, Back dribbles and
Key passes while Wei Shihao preferred more to cross from the right backline.
Wu Lei 2017 VS. Wu Lei 2018
Figure 5 (b) shows the player report of Wu Lei in both 2017 and 2018 which had
a similarity of 92.8%. It is obviously that the most preferred style of Wu Lei transformed
from Winger to Poacher although the average ratings of the flank roles (L Winger, R
Winger and Inside Forward) were higher than that of playing as a Poacher. For the
season general comparison, in season 2018 Wu Lei showed a much higher preference
for Close Shot, Mid front Header and Front header; while the preference for Left shot,
Crosses, Back Passes and Dribbles, all dropped.
Renato Augusto VS. Jonathan Viera
As it is showed in Figure 5 (c), the midfield partner had a similarity of 70.6%.
Among all the 24 matches they started together, Viera played most as a Second Striker
while Augusto preferred more for playing as a Playmaker or playing in the left of the
field as an Inside Forward or a Winger. For the general style of the season, Viera showed
a much higher preference for Close shot, Right shot, Center dribble (even though the
similarity of Augusto was also relatively high) and Key passes, while Augusto did more
on Long shot, Crosses, Flank dribbles, Back passes and Recoveries.
Figure 5. Comparison results of the three pairs of players. (Notes: CS = Close Shot, LS = Long
Shot, RS = Right Shot, LS = Left Shot, LBC = L. Backline Cross, RBC = R. Backline Cross,
RFC = R. Flank Cross, LFC = L. Flank Cross, CD = Center Dribble, RFD = R. Flank Dribble,
LFD = L. Flank Dribble, LBD = L. Back Dribble, RBD = R. Back Dribble, LBP = L. Back Pass,
RFP = R. Flank Pass, LFP = L. Flank Pass, CP = Center Pass, RBP = R. Back Pass, LBLP = L.
back long pass, LFLP = L. flank long pass, RFLP = R. flank long pass, CLP = Center long pass,
RBLP = R. back long pass, RFKP = R. far key pass, RKP = R. key pass, LKP = L. key pass,
LFKP = L. far key pass, BH = Back Header, MFH = Mid Front Header, RBH = R. Back Header,
LBH = L. Back Header, FH = Front Header, LBR = L. back recovery, RFR = R. flank recovery,
LFR = L. flank recovery, RBR = R. back recovery)
4. Discussion
Using match event data, spatial-temporal information of actions and domain
knowledge, the study was to provide a more complex solution to characterize
professional football players’ playing styles of different positions, integrating a recently
adopted Player Vectors framework. Twelve on-the-ball actions were used to generate a
44-dimensional player vector for each player in each match. Players were clustered into
eight different positions, and after performing another NMF process on various
components of these positions, eighteen different kinds of playing styles were finally
discovered in the China Football Super League. The findings provided additional
insights into player evaluation from the perspective of playing style, and could be
applied to player recruitment and match preparation.
4.1 Position Clustering
The Position Clustering process detected eight different roles based on their
average on-the-ball action positions. A silhouette score of 0.41 seemed not so high in
data science resulted by the denseness of the sample, while the algorithm was able to
divide the players into clusters in line with the common knowledge of soccer.
Additionally, the lacking of tracking data also limited the possibility of using more
complex methods which may enhance the clustering effect (Bialkowski et al., 2014).
For these reasons, despite the imperfections of this position clustering method, it can
meet the needs of this study.
4.2 Playing styles of the CSL players
Strikers
This study revealed that almost half of the Center Forwards (Second Striker) in
CSL tended to move deep into the midfield to help teams’ collective organization
(Duarte et al., 2012). However, unlike the Second Strikers, Poachers were required to
focus on the most important task in football: putting the ball in the back of the net
(Guide to FM, 2020), which explained their highest average rating. Moreover, Mobile
Strikers with the highest W/L, who most of them were foreign players started as a striker
or midfielder, were inclined to play on the flanks and to cut into the middle, implying
that assigning best high-level foreign strikers to play on the flank may be a potential
key to goal-scoring and success (Lago-Peñas et al., 2018). Furthermore, Target Mans
had the lowest rating and W/L, which was in line with their duties: being green leaves
to open up spaces for key players by their physicality and aerial presence and join in
the battle in the midfield when the teams were facing unfavorable match situation
(Guide to FM, 2020).
It is interesting to find out that a large number of strikers are Second Strikers. Part
of the reasons may be related to the strategical demands of the match, since nearly two
thirds of CSL strikers are foreign players who make indispensable contribution to teams’
organizing, thus enhancing teams’ success (Gai et al., 2019). By the other hand, nearly
40% of Second Strikers were players starting as a midfielder and were clustered to
Strikers, and 65% of them were also foreign players (e.g., Jonathan Viera). This
indicated that the CSL attacking midfielders not only participated into organizing, but
also moved frequently forward to the attacking third to create threat to the goal.
Left/Right Wing Forwards
L/R Wingers were the team’s sharp knife, they may choose to dribble alongside
the flank line and make backline crosses or attempt to threat the keeper from the wide
side of the box. Inside Forward was a role similar to the Second Striker and the
Playmaker but acting on a wider side of the field, which can explain the high proportion
of players who started as a midfielder in this style. This playing style are space
orchestrators and have become more and more necessary in modern soccer tactical,
especially in an era when super full backs sometimes monopolize the team’s attacking
and defending task on the flank (Crossbarhub, 2020).
It is worthwhile to notice that the appearance number of L/R Wingers were nearly
three times than that of Inside Forwards. The finding echoed with the study by Gai et
al. (2019) that wing players in CSL were prone to play more crosses, while such
behavior was only present in wingers of weak teams in Spanish La Liga (Liu et al.,
2016). This may suggest that most of the CSL Wing Forwards lack the abilities to cut
inside and make direct threats to the opponent goal, which would in turn restrict coaches
from adopting other tactical alternatives. For the Inside Forwards, it is interesting to
find out the majority of them were domestic players who started as a midfielder or
foreign strikers. This might suggest that in the CSL, it was the domestic midfield players
and high-level foreign strikers that performed more dribbling or running inverted to the
threaten area from the flanks, rather than the flank players. Furthermore, one tenth of
the L/R Wingers were players starting as full-backs, which indicated that there were
couples of full backs in the CSL who could make essential contribution to attacking,
but the ratio was relatively small.
Central Midfielders
Midfielders are the primary link of a team when building attacks (Clemente et al.,
2015), Left/Right Defensive Midfielders were mainly focused on regaining possession
and reorganizing from the backfield. In that case, they were reasonable to have a
relatively low rating since most of the rating systems tend to fancy on goal scorers
(Decroos et al., 2019; Pappalardo et al., 2019). Playmakers are the midfielders who are
responsible for providing passes and build up attacking in the center of the pitch.
Although acted as the commander on the field, their W/L was the lowest, despite having
a second-best rating. Moreover, it is worthwhile to notice that Wide Midfielders
performed best among the 4 playing styles which may indicate that owning Midfielders
with almighty ability and flank-preferred acting area were potential keys to win in CSL.
These met the research results of Wu et al. (2020) that instead of sticking in the middle,
winning teams always make a better use of the width of the pitch.
According to studies related to football passing network analysis, midfield players
were always the preeminent players in connecting teammates by passing (Clemente et
al., 2016a; Clemente et al., 2016b, Yu et al., 2020). While this study further notes that
their preferred passing area might be different according to the playing style. Moreover,
most of the Central Midfielders were domestic players who simply focused on
organizing (Playmakers) and midfield defending (Defensive Midfielders), such tactical
role is similar to that of their counterparts Spanish La Liga top teams (Gai et al., 2019;
Liu et al., 2016). A possible reason is that most of the powerful foreign midfielders in
the CSL were assigned into other positions such as Second Strikers and Mobile Strikers,
so that they could offer more direct helps in certain areas, letting their domestic
teammates to locate in the middle of the pitch and to center on organizing and defending.
Such finding was also supported by the evidence that in the CSL, foreign midfielders
outperformed on both goals and assists (Gai et al., 2019). In contrast, strikers in La Liga
might strong enough on goal scoring, thus freeing the midfielders to focus on organizing,
rather than making threats to the goal (Gonçalves et al., 2014).
Left/Right Full Backs
Most of the Full Backs were Wing Backs who started as Left or Right Backs and
intended to fulfill the attacking and defensive duties on the flank side. Another 700
more players played as L/R Backs who focused primarily on their defensive duties
which explained their lower ratings. Particularly, Wing backs’ attacking preferences
(e.g., Crosses) may be suppressed by the pressure from the opposition and had to turn
to L/R Backs, which may be the reason of the lower Win-Loss ratio. What must be
noted is that the preference for crossing is relatively lower than that of dribbling and
passing in the flank sides, which may indicate that most of the CSL Wing backs lacked
the desires on breaking out from the back and crossing the ball into the box.
In contrast, previous research has pointed out that Wing backs were essential for
winning (at both attacking and defending sides) and should contribute more in the final
phase of attacking (Konefał et al., 2015; Wu et al., 2020). Specifically, full backs of
Spanish top teams tended to execute more attacking actions like crossing and passing
than those of the weak teams (Liu et al., 2016). But in the CSL, full backs, almost all
of which were domestic players, were not able to accomplish offensive and defensive
tasks at both ends (Gai et al., 2019), since they might not meet the increasing match
physical demands if they had participated into attacking (Bush et al., 2015; Crossbarhub,
2020; Zhou et al., 2020). Another possible explanation is the lacking of cutting inside
ability in most CSL Wing Forwards, since the former players are supposed to provide
their Full back teammates enough space to break out to the front area and cross the ball
(Crossbarhub, 2020).
Central Backs
Nearly all of the Central backs in the CSL were Ball playing Defenders, and only
683 Central Defender were observed. These were in line with a trend that Central backs
are becoming more and more important on a team’s attacking build up (Clemente et al.,
2015; Korte et al., 2019) and the traditional style one are declining (Ayyagari, 2018).
Clearly, CSL teams were trying to catch up the tactical trend by requesting their Central
backs to convey long passes. However, the minority Central Defenders had a way better
W/L, which may indicate that in soccer leagues with lower technical level like the CSL,
it is better to let the Central back players focusing on defending, rather than keep
thinking about being a long passer (Zhou et al., 2020).
However, Gai et al. (2019) argued that the CSL central defenders preferred more
defensive duties instead of being an additional passing options like the defenders of
English Premier League (Bush et al., 2015). While this study may provide a new insight
from the perspective of playing style that most of the Central backs were trying to
accomplish the passing tasks on the field, but they cannot perfectly fulfill it because of
the lacking of passing abilities.
4.3 Player comparisons
The use of performance profiling in sport allows for a better understanding of
players and teams’ performances. On the one hand, it could visualize the match actions
in different context; on the other hand, the statistical information generated from the
profiling process can be added to justify comparisons and modelling performances
(Butterworth et al., 2013).
Alex Teixeira VS. Wei Shihao
From the high similarity and the general style radar chart, Teixeira and Wei Shihao
did share a lot of common grounds. The differences appeared on the preference for
taking long shots, dribble in the center part and sending key passes where Teixeira’s
weights were much higher than Wei Shihao which can also be proved by the result of
style clustering. These indicated that Teixeira was more involved into the team’s
organization and more like a key player of his team than that of Wei Shihao.
Wu Lei 2017 VS. Wu Lei 2018
Although Wu Lei’s playing style did not change a lot from 2017 to 2018, the
differences in the preferred styles and in some specific components could still show
some clues. He showed a higher preference for playing in the center front as a Poacher
in 2018 than that of 2017. The increased weight on Close shot and Front header
combined with the decreased weight on Left shot and Crosses showed this trend too.
This change means that he was getting closer to the goal line that made a bigger chance
to beat the goalkeeper, which may be the result of his increased match experiences
(Kalén et al., 2019).
Renato Augusto VS. Jonathan Viera
Above all, the most obvious difference was that Viera played as a Second Striker
in almost all his started games in 2018 while Augusto preferred to play in various styles
as a Wing Forward or a Center Midfielder. These indicated that Viera preferred to hide
behind the main striker and not only sending key passes, but also trying to threat the
goal. Augusto was a midfielder who had well-developed physical fitness to enable
himself to cover wide field of the pitch, and such player was exactly that strong teams
in the CSL should recruit, especially in an era of increasing physical demands in top
leagues (Bush et al., 2015).
4.4 Practical application
The current study would provide practical information to club managers and
coaches in the CSL at four different levels. Firstly, at a team tactical level, the
framework can provide references for coaches on tactical arrangement and formation
setting, they may try to enhance the team’s performance and win ratio by combining
certain types of players or changing tactics by sending on substitute players in certain
styles. Secondly, at a club management level, having more knowledge about players’
playing style (e.g., similarities) can be quite helpful in the transfer market in case for
wasting money on wrong players. Thirdly, at an individual level, players will get to
know more about their careers’ development and characteristics about their favorite
positions and styles. Fourthly, it will provide more specific information about a player’s
playing style to broadcast companies, thus enhance the watching experience of fans.
4.5 Limitation and future works
This research only considered on-the-ball actions, other important data
information in soccer analysis like tracking data have not been used. These types of
data can be important on characterizing players’ playing style. On the one hand, Garrido
et al. (2021) have confirmed in a recent study that the heatmap drawn from event data
has a low correlation with the one drawn from tracking data. On the other hand, off-
ball actions like positioning and running are essential for positions like central
defenders who do not have many chances to touch the ball. Therefore, combining
tracking data in the future will be necessary and helpful for making a more precise
characterization of the playing style. Furthermore, this study only considered the
playing style of outfield players but excluded goalkeepers. Since the style of goalkeeper
is more about saving preference and passing only in the back, which the Player Vectors
model could not attach, it is necessary to develop a data-driven framework to
characterize the playing style of goalkeepers in the future.
5. Conclusion
This work characterized and analysed the playing style of CSL players in season
2016-2019 based on the state-of-the-art Player Vectors framework. The playing styles
of the CSL forwards and midfielders were overall match the trend that winning teams
performed better in the flank and players who tend and have the ability to cover as much
of the field area as possible should be attached more importance to the winning of the
match. In contrast, Full backs and Central backs are positions whose playing styles
should be strengthened and reconsidered. Meanwhile, position and style clustering
results proved that in CSL, better players, especially foreign players are becoming more
multifunctional, which was also a widely accepted trend. Furthermore, the player
comparison provided and verified the possibility of further using a data-driven model
like the Player Vectors to analyse and compare the playing styles in certain scenarios.
References
Aalbers, B. & Van Haaren, J. (2019). Distinguishing Between Roles of Football Players
in Play-by-Play Match Event Data. In U. Brefeld, J. Davis, J. V. Haaren & A.
Zimmermann (Eds.), Machine Learning and Data Mining for Sports Analytics
MLSA 2018. Lecture Notes in Computer Science, vol 11330. (pp. 31-41).
Springer, Cham. https://doi.org/10.1007/978-3-030-17274-9_3
Ayyagari, S. (2018, November 21). Why Are Traditional Centre-Backs Diminishing In
World Football. Retrieved December 20, 2021, from https://blog.playo.co/why-
are-traditional-centre-backs-diminishing/
Bialkowski, A., Lucey, P., Carr, P., Yue, Y., Sridharan, S. & Matthews, I. (2014). Large-
scale analysis of soccer matches using spatiotemporal tracking data. Paper
presented at the 2014 IEEE International Conference on Data Mining, Shenzhen,
China.
Bush, M., Barnes, C., Archer, D. T., Hogg, B. & Bradley, P. S. (2015). Evolution of
match performance parameters for various playing positions in the English
Premier League. Human Movement Science, 39, 1-11.
https://doi.org/https://doi.org/10.1016/j.humov.2014.10.003
Butterworth, A., O’Donoghue, P. & Cropley, B. (2013). Performance profiling in sports
coaching: a review. International Journal of Performance Analysis in Sport,
13(3), 572-593. https://doi.org/10.1080/24748668.2013.11868672
Castellano, J. & Pic, M. (2019). Identification and Preference of Game Styles in LaLiga
Associated with Match Outcomes. International Journal of Environmental
Research and Public Health, 16(24), 5090.
https://doi.org/10.3390/ijerph16245090
Clemente, F. M., Martins, F. M. L., Wong, P. D., Kalamaras, D. & Mendes, R. S. (2015).
Midfielder as the prominent participant in the building attack: A network
analysis of national teams in FIFA World Cup 2014. International Journal of
Performance Analysis in Sport, 15(2), 704-722.
https://doi.org/10.1080/24748668.2015.11868825
Clemente, F. M., José, F., Oliveira, N., Martins, F. M. L., Mendes, R. S., Figueiredo, A.
J., . . . Kalamaras, D. (2016a). Network structure and centralization tendencies
in professional football teams from Spanish La Liga and English Premier
Leagues. Journal of Human Sport and Exercise, 11(3), 376-389.
https://doi.org/10.14198/jhse.2016.113.06
Clemente, F. M., Silva, F., Martins, F. M. L., Kalamaras, D. & Mendes, R. S. (2016b).
Performance Analysis Tool for network analysis on team sports: A case study of
FIFA Soccer World Cup 2014. 230(3), 158-170.
https://doi.org/10.1177/1754337115597335
Crossbarhub. (2020, May 23). Midfielder Types in Football. Retrieved December 21,
2021, from https://crossbarhub.com/tactical-analysis/midfielder-types-in-
football/
Crossbarhub. (2020, October 13). Types of Fullbacks. Retrieved December 21, 2021,
from https://crossbarhub.com/tactical-analysis/types-of-fullbacks/
Decroos, T., Bransen, L., Van Haaren, J. & Davis, J. (2019). Actions Speak Louder than
Goals : Valuing Player Actions in Soccer. Paper presented at the Proceedings of
the 25th ACM SIGKDD International Conference on Knowledge Discovery &
Data Mining. Anchorage, AK, USA. https://doi.org/ 10.1145/3292500.3330758
Decroos T., Davis J. (2020) Player Vectors: Characterizing Soccer Players’ Playing
Style from Match Event Streams. In: Brefeld U., Fromont E., Hotho A., Knobbe
A., Maathuis M., Robardet C. (Eds) Machine Learning and Knowledge
Discovery in Databases. ECML PKDD 2019. Lecture Notes in Computer
Science, vol 11908 (pp. 569-584). Springer, Cham. https://doi.org/10.1007/978-
3-030-46133-1_34
Duarte, R., Araújo, D., Correia, V. & Davids, K. (2012). Sports Teams as
Superorganisms. Sports Medicine, 42(8), 633-642.
https://doi.org/10.1007/BF03262285
Fernandez-Navarro, J., Fradua, L., Zubillaga, A., Ford, P. R. & McRobert, A. P. (2016).
Attacking and defensive styles of play in soccer: analysis of Spanish and
English elite teams. Journal of Sports Sciences, 34(24), 2195-2204.
https://doi.org/10.1080/02640414.2016.1169309
Gai, Y., Volossovitch, A., Lago, C. & Gómez, M.-Á. (2019). Technical and tactical
performance differences according to player’s nationality and playing position
in the Chinese Football Super League. International Journal of Performance
Analysis in Sport, 19(4), 632-645.
https://doi.org/10.1080/24748668.2019.1644804
Garcia-Aliaga, A., Marquina, M., Coteron, J., Rodriguez-Gonzalez, A. & Luengo-
Sanchez, S. (2020). In-game behaviour analysis of football players using
machine learning techniques based on player statistics. International Journal of
Sports Science & Coaching, 16(1), 148-157.
https://doi.org/10.1177/1747954120959762
Garrido, D., Burriel, B., Resta, R., del Campo, R. L. & Buldu, J. (2021). Heatmaps in
soccer: event vs tracking datasets. arXiv preprint.
https://arxiv.org/abs/2106.04558
Goes, F., Kempe, M., van Norel, J. & Lemmink, K. A. P. M. (2021). Modelling team
performance in soccer using tactical features derived from position tracking data.
IMA Journal of Management Mathematics, 32(4), 519–533.
https://doi.org/10.1093/imaman/dpab006
Goes, F., Meerhoff, L. A., Bueno, M. J. O., Rodrigues, D. M., Moura, F. A., Brink, M.
S., . . . Lemmink, K. (2020). Unlocking the potential of big data to support
tactical performance analysis in professional soccer: A systematic review.
European Journal of Sport Science, 21(4), 481-496.
https://doi.org/10.1080/17461391.2020.1747552
Gómez, M.-Á., Mitrotasios, M., Armatas, V. & Lago-Penas, C. (2018). Analysis of
playing styles according to team quality and match location in Greek
professional soccer. International Journal of Performance Analysis in Sport,
18(6), 986-997. https://doi.org/10.1080/24748668.2018.1539382
Gonçalves, B. V., Figueira, B. E., Maçãs, V. & Sampaio, J. (2014). Effect of player
position on movement behaviour, physical and physiological performances
during an 11-a-side football game. Journal of Sports Sciences, 32(2), 191-199.
https://doi.org/10.1080/02640414.2013.816761
Gudmundsson, J. & Horton, M. (2017). Spatio-Temporal Analysis of Team Sports.
ACM Computing Surveys, 50(2), 1-34. https://doi.org/10.1145/3054132
Guide to FM. (2020). Central Attack Roles. Retrieved December 26, 2021, from
https://www.guidetofm.com/tactics/central-attack-roles/
Hewitt, A., Greenham, G. & Norton, K. (2016). Game style in soccer: what is it and can
we quantify it? International Journal of Performance Analysis in Sport, 16(1),
355-372. https://doi.org/10.1080/24748668.2016.11868892
Kalén, A., Rey, E., de Rellán-Guerra, A. S. & Lago-Peñas, C. (2019). Are Soccer
Players Older Now Than Before? Aging Trends and Market Value in the Last
Three Decades of the UEFA Champions League. Frontiers in Psychology, 10.
https://doi.org/10.3389/fpsyg.2019.00076
Konefał, M., Chmura, P., Andrzejewski, M., Pukszta, D. & Chmura, J. (2015). Analysis
of match performance of full-backs from selected European soccer leagues.
Central European Journal of Sport Sciences Medicine, 11(3), 45-53.
https://doi.org/10.18276/cej.2015.3-05
Korte, F., Link, D., Groll, J. & Lames, M. (2019). Play-by-Play Network Analysis in
Football. Frontiers in Psychology, 10(1738).
https://doi.org/10.3389/fpsyg.2019.01738
Lago-Peñas, C., Gómez-Ruano, M. & Yang, G. (2018). Styles of play in professional
soccer: an approach of the Chinese Soccer Super League. International Journal
of Performance Analysis in Sport, 17(6), 1073-1084.
https://doi.org/10.1080/24748668.2018.1431857
Lasek, J., Szlávik, Z. & Bhulai, S. (2013). The predictive power of ranking systems in
association football. International Journal of Applied Pattern Recognition, 1(1),
27-46. https://doi.org/10.1504/IJAPR.2013.052339
Lee, D. D. & Seung, H. S. (2000). Algorithms for non-negative matrix factorization.
Paper presented at the Proceedings of the 13th International Conference on
Neural Information Processing Systems, Denver, CO, USA.
Li, Y., Ma, R., Gonçalves, B., Gong, B., Cui, Y. & Shen, Y. (2020). Data-driven team
ranking and match performance analysis in Chinese Football Super League.
Chaos, Solitons & Fractals, 141, 110330.
https://doi.org/10.1016/j.chaos.2020.110330
Liu, H., Gómez, M.-Á., Gonçalves, B. & Sampaio, J. (2016). Technical performance
and match-to-match variation in elite football teams. Journal of Sports Sciences,
34(6), 509-518. https://doi.org/10.1080/02640414.2015.1117121
Liu, H., Hopkins, W., Gómez, M.-Á. & Molinuevo, S. J. (2013). Inter-operator
reliability of live football match statistics from OPTA Sportsdata. International
Journal of Performance Analysis in Sport, 13(3), 803-821.
https://doi.org/10.1080/24748668.2013.11868690
Lord, F., Pyne, D. B., Welvaert, M. & Mara, J. K. (2020). Methods of performance
analysis in team invasion sports: A systematic review. Journal of Sports
Sciences, 38(20), 2338-2349. https://doi.org/10.1080/02640414.2020.1785185
Mazurek, J. (2018). Which football player bears most resemblance to Messi? A
statistical analysis. arXiv preprint. https://arxiv.org/abs/1802.00967
McHale, I. G., Scarf, P. A. & Folker, D. E. (2012). On the Development of a Soccer
Player Performance Rating System for the English Premier League. Interfaces,
42(4), 339-351. https://doi.org/10.1287/inte.1110.0589
McLean, S., Salmon, P. M., Gorman, A. D., Naughton, M. & Solomon, C. (2017). Do
inter-continental playing styles exist? Using social network analysis to compare
goals from the 2016 EURO and COPA football tournaments knock-out stages.
Theoretical Issues in Ergonomics Science, 18(4), 370-383.
https://doi.org/10.1080/1463922x.2017.1290158
Pappalardo, L., Cintia, P., Ferragina, P., Massucco, E., Pedreschi, D. & Giannotti, F.
(2019). PlayeRank: Data-driven Performance Evaluation and Player Ranking in
Soccer via a Machine Learning Approach. ACM Transactions on Intelligent
Systems and Technology, 10(5), 1-27. https://doi.org/10.1145/3343172
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., . . .
Dubourg, V. (2011). Scikit-learn: Machine learning in Python. the Journal of
machine Learning research, 12(85), 2825-2830.
Peña, J. L. & Navarro, R. S. (2015). Who can replace Xavi? A passing motif analysis
of football players. arXiv preprint. https://arxiv.org/abs/1506.07768
Rousseeuw, P. J. (1987). Silhouettes: A graphical aid to the interpretation and validation
of cluster analysis. Journal of Computational and Applied Mathematics, 20, 53-
65. https://doi.org/ 10.1016/0377-0427(87)90125-7
Whoscored. (2018). Super League Player Statistics 2018. Retrieved December 26, 2021,
from
https://www.whoscored.com/Regions/45/Tournaments/162/Seasons/7242/Stag
es/15995/PlayerStatistics/China-Super-league-2018
Wu, Y., Xia, Z., Wu, T., Yi, Q., Yu, R. & Wang, J. (2020). Characteristics and
optimization of core local network: Big data analysis of football matches. Chaos,
Solitons & Fractals, 138, 110136. https://doi.org/ 10.1016/j.chaos.2020.110136
Yu, Q., Gai, Y., Gong, B., Gómez, M.-Á. & Cui, Y. (2020). Using passing network
measures to determine the performance difference between foreign and
domestic outfielder players in Chinese Football Super League. International
Journal of Sports Science & Coaching, 15(3), 398-404.
https://doi.org/10.1177/1747954120905726
Zhou, C., Calvo, A. L., Robertson, S. & Gómez, M.-Á. (2020). Long-term influence of
technical, physical performance indicators and situational variables on match
outcome in male professional Chinese soccer. Journal of Sports Sciences, 39(6),
598-608. https://doi.org/10.1080/02640414.2020.1836793
Zhou, C., Lago-Peñas, C., Lorenzo, A. & Gómez, M.-Á. (2021). Long-Term Trend
Analysis of Playing Styles in the Chinese Soccer Super League. Journal of
Human Kinetics, 79(1), 237-247. https://doi.org/10.2478/hukin-2021-0077
Supplementary materials
Supplementary Table 1. The definitions and available results and attributes of all the events in the data-set
Event name
Definitions
Event results
Event attributes
Simple pass
Any intentional played ball from one player to
another.
Assist，Keypass, Successful, Unsuccessful
Through、simple
Long pass
A lofted ball where there is a clear intended
recipient, must be over shoulder height and
using the passes height to avoid opposition
players or a long high ball into space or into an
area for players to chase or challenge for the
ball
Throw in
Throw in
Goal kick
Goal kick
Freekick pass
A pass made from a freekick
Cross
Any intentional played ball from a wide
position intending to reach a team mate in a
specific area in front of the goal.
Freekick cross
A cross made from a freekick
Corner cross
A cross made from a corner
Shot
An attempt to score a goal, made with any
(legal) part of the body, either on or off target
Goal, On target, Off target, On post, Blocked
Out box, In box
Freekick shot
Any attempts created directly from the free
kick itself (unassisted).
Penalty shot
A shot made from a penalty
Corner shot
A shot made by the corner taker
Chance missed
A big chance opportunity when the player does
not get a shot away, typically given for big
chance attempts where the player shooting
completely misses the ball (air shot) but can
also be given when the player has a big chance
opportunity to shoot and decides not to,
resulting in no attempt occurring in that attack.
Unsuccessful
Simple
Offside
Awarded to the player deemed to be in an
offside position where a free kick is awarded
Unsuccessful
Simple
Own goal
An own goal is usually awarded if the attempt
is off target and deflected into the goal by an
opponent.
Own goal
Simple
Clearance
A defensive action where a player kicks the ball
away from his own goal with no intended
recipient.
Successful, Unsuccessful
Simple
Interception
where a player reads an opponent’s pass and
intercepts the ball by moving into the line of the
intended pass.
Successful
Simple
Tackle
where a player connects with the ball in a
ground challenge where he successfully takes
the ball away from the player in possession.
Successful, Unsuccessful
Simple
Defensive duel
Player has been beaten in one-on-on
Unsuccessful
Simple
Blocked pass
When a player tries to cut out an opposition
pass by any means. Similar to an interception
except there is much less reading of the pass.
Successful
Simple
Blocked shot
A player blocks a shot on target from an
opposing player.
Successful
Simple
Defensive foul
Any infringement that is penalised as foul play
by a referee in defending.
Unsuccessful, Red card, Second yellow card,
Yellow card
Simple
Offensive foul
Any infringement that is penalised as foul play
by a referee in offending.
Simple
Red card
Unsuccessful
Simple
Second yellow
card
Simple
Yellow card
Simple
Aerial duel
This is where two players challenge in the air
against each other.
Successful, Unsuccessful
Simple
Shield ball out
Where a player shields the ball from an
opponent and is successful in letting it run out
of play. Can be offensive (to win a corner or
throw in up field) or defensive (winning a
throw in or goal kick).
Successful
Simple
Ball recovery
where a player recovers the ball in a situation
where neither team has possession or where the
ball has been played directly to him by an
opponent, thus securing possession for their
team.
Successful
Simple
Offside provoked
The deepest player in the defensive line when
an offside has been given.
Unsuccessful
Simple
Foul conceded
any infringement that is penalised as foul play
by a referee.
Successful
Simple
TakeOn
an attempt by a player to beat an opponent
when they have possession of the ball
Successful, Unsuccessful
Simple
Dribble
Player dribbles at least 3 meters with the ball
Successful
Simple
Touch
When the ball bounces off a player and there is
no intentional pass,
Successful, Unsuccessful
Simple
Dispossessed
A player is in possession but not attempting to
“beat” his opponent and get tackled
Unsuccessful
Simple
Error
When a player makes an error, which leads to
a goal or shot conceded. Also used for spills
and attempted claims or saves by a goalkeeper
which directly leads to a second attempt to
score.
Unsuccessful
Simple
Keeper save
A goalkeeper preventing the ball from entering
the goal with any part of his body when facing
an intentional attempt from an opposition
player.
Successful
Simple
Keeper pickup
When the goal keeper picks up the ball and his
side regain possession, similar to recovery,
however, the goal keeper picks the ball up.
Successful
Simple
Penalty faced
A goalkeeper is facing a penalty
Successful, Unsuccessful
Simple
Keeper sweeper
Anytime a goalkeeper anticipates danger and
rushes off their line to try to either cut out an
attacking pass (in a race with the opposition
player) or to close-down an opposition player.
Successful, Unsuccessful
Simple
Keeper smother
A goalkeeper who comes out and claims the
ball at the feet of a forward gets a smother,
similar to a tackle, however, the keeper must
hold onto the ball to award a smother.
Successful
Simple
Keeper punch
A high ball that is punched clear by the
goalkeeper. The keeper must have a clenched
fist and attempting to clear the high ball rather
than claim it.
Successful
Simple
Keeper claim
A high ball played into the penalty area that is
caught by the goalkeeper.
Successful
Simple
CrossNotClaimed
When a goalkeeper comes off his goal line to
claim a high ball (attempting a catch) and
misses the ball.
Unsuccessful
Simple
Supplementary Table 2. The definition and number of appearances of each style in each position
Position
Style
Appearances
Definitions
Total
Start Position
Domestic
Foreign
ST
Second Striker
1394
MF
508
176
332
Strikers or attacking midfielders
who tend to move deep into the
midfield to help team’s organizing
FW
695
101
594
Target Man
570
FW
533
102
431
Strikers who need to open up
spaces for key players by their
physicality and aerial presence
Mobile Striker
646
LM
119
72
47
Strikers who like to play on the
flanks and cutting into the middle
RM
104
59
45
FW
245
59
186
MF
177
52
125
Poacher
613
FW
523
113
410
Strikers focusing on putting the
ball in the back of the net
CM
L/R Defensive Midfielder
1563
MF
1427
1097
330
Midfielders who are good at
tackling, interception and physical
duels
Playmaker
1549
MF
1403
989
414
Midfielders who undertake the
main organizing task of the team
Wide Midfielder
544
MF
317
225
92
Midfielders who have a large
scope of running area and good at
playing on the flanks
L/RW
L/R Winger
2531
L/RM
1523
1126
397
Players who are good at breaking
the defensive line and making
shots or crosses on the flanks
L/RB
230
226
4
MF
481
355
126
FW
115
44
71
Inside Forward
890
LM
126
63
63
Players who tend to build up
attacks on the flanks or carry the
ball to the middle of the field
RM
124
66
58
FW
161
36
125
FB
L/R Wing Back
3252
L/RM
311
297
14
Full backs who tend to break out
from the back and making crosses
or passes
L/RB
2599
2578
21
MF
252
249
3
L/R Back
733
L/RB
283
283
0
Full backs who tend to focus on
defending
DF
239
201
38
CB
L/R Ball Playing Defender
3729
DF
3279
2411
868
Central backs who undertake the
backfield carrying and passing of
the team
MF
278
241
37
Central Defender
685
DF
611
449
162
Central backs who tend to focus
on defending
Notes: ST = Strikers, L/RW = Left/Right Wing forwards, CM = Central Midfielders, L/RFB = Left/Right Full Backs, CB = Central Backs,
FW = Forward, MF = Midfielder, CD=Center Defender, L/RM = Left/Right Midfielder, L/RD = Left/Right Defender
