Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Sports Sciences

ISSN: 0264-0414 (Print) 1466-447X (Online) Journal homepage: http://www.tandfonline.com/loi/rjsp20

Dribbling speed along curved paths predicts


attacking performance in match-realistic one vs.
one soccer games

Robbie S Wilson, Nicholas M A Smith, Solange de Paula Ramos, Fabio


Giuliano Caetano, Marcio Aparecido Rinaldo, Paulo Roberto Pereira
Santiago, Sergio Augusto Cunha & Felipe Arruda Moura

To cite this article: Robbie S Wilson, Nicholas M A Smith, Solange de Paula Ramos, Fabio
Giuliano Caetano, Marcio Aparecido Rinaldo, Paulo Roberto Pereira Santiago, Sergio Augusto
Cunha & Felipe Arruda Moura (2018): Dribbling speed along curved paths predicts attacking
performance in match-realistic one vs. one soccer games, Journal of Sports Sciences

To link to this article: https://doi.org/10.1080/02640414.2018.1544110

Published online: 23 Nov 2018.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=rjsp20
JOURNAL OF SPORTS SCIENCES
https://doi.org/10.1080/02640414.2018.1544110

SPORTS PERFORMANCE

Dribbling speed along curved paths predicts attacking performance in


match-realistic one vs. one soccer games
Robbie S Wilsona, Nicholas M A Smithb, Solange de Paula Ramosc, Fabio Giuliano Caetano c
,
Marcio Aparecido Rinaldoc, Paulo Roberto Pereira Santiago d, Sergio Augusto Cunha e
and Felipe Arruda Moura c
a
School of Biological Sciences, The University of Queensland, St Lucia, Australia; bSchool of Life and Environemntal Sciences, University of Sydney,
Brisbane, Australia; cSport Sciences Department, State University of Londrina, Londrina, Brazil; dSchool of Physical Education and Sport of Ribeirão
Preto, University of Sao Paulo, Ribeirão Preto, Brazil; eCollege of Physical Education, University of Campinas, Campinas, Brazil

ABSTRACT ARTICLE HISTORY


This study assessed whether a new, closed-skill dribbling or sprinting task could predict attacking Accepted 19 September 2018
performance in soccer. Twenty-five male players were recruited from the Londrina Junior Team Football
KEYWORDS
Academy in Brazil and asked to either dribble the ball or sprint through five custom circuits that varied Soccer; football; dribbling;
in average curvature (0–1.37 radians.m−1). These measures were then validated using a realistic one vs. one vs. one; skill
one competition in which each player acted as attacker or defender in turn (N = 1330 bouts). Sprinting
(ICC = 0.96) and dribbling (ICC = 0.97) performances were highly repeatable for individual players.
Average dribbling speed decreased non-linearly with increasing circuit curvature (F = 239.5; P < 0.001)
from 5.19 ± 0.11 ms−1 on the straightest path to 2.13 ± 0.03 ms−1 on the curviest. Overall, dribbling but
not sprinting performance predicted attacking success in the one vs. one competition, explaining more
than 50% of the variation in attacking success alone (rp = 0.70; P < 0.001). In conclusion, our new closed-
skill dribbling assessment is a valid and reliable protocol to predict a soccer player’s success in attacking
performance in one vs. one situation, and can be used to identify talented players.

Introduction top-level players were significantly faster than less experi-


enced players in a dribbling test around different obstacles.
One of the most important skills in soccer is the ability to take on
Yet not every study finds differences in dribbling among elite
opposition players and dribble the ball past them (Zago et al.,
and amateur levels. Coelho e Silva et al. (2010) found no
2016). By beating defenders, players can create space for team-
significant differences across a range of soccer skills (including
mates, open up direct scoring opportunities, and pull defenders
dribbling) when comparing different levels of under-14 soccer
out of their preferred positions. As a result, dribbling is one of the
players.
most effective attacking skills in soccer, and top professional
One limitation of many dribbling tests is that they do not
clubs seek to obtain and develop players who excel at it.
test the variety of dribbling techniques used in matches. In
Improvement of dribbling is considered a pivotal attribute in
tight positions, players can approach and beat opponents with
the development of young players (Huijgen, Elferink-Gemser,
close control of the ball and rapid changes in direction, while
Post, & Visscher, 2010), and though current methods for testing
in open or wide positions players may use both control and
dribbling skill can be used to identify talent (Waldron & Worsfold,
running speed to beat opponents. Another criticism of cone-
2010), they may have little relevance to performance in compe-
based dribbling tests is they do not incorporate the number of
titive matches (Ali, 2011; O’Reilly & Wong, 2012; Vilar, Araújo,
mistakes into analyses and only require close ball control
Davids, & Travassos, 2012).
when near cones (Ali, 2011). Finally, dribbling is a complex,
Dribbling performance is usually assessed by quantifying
multi-player action that relies on the perceptual, cognitive,
the time an individual player takes to dribble around a set of
and motor skills of both the attacker and the defender (Ali,
cones (Haaland & Hoff, 2003; Huijgen et al., 2010; Reilly &
2011; Unnithan, White, Georgiou, Iga, & Drust, 2012; Vilar et al.,
Holmes, 1983; Russell, Benton, & Kingsley, 2010). These types
2012), so closed-skill tests like dribbling speed must also be
of tests, known as closed-skill, can provide repeatable mea-
able to predict individual performance in match-realistic situa-
sures of individual dribbling performance (Ali, 2011) and can
tions that involve an attacker and defender. If closed-skill tests
even distinguish among players of different representative
of dribbling performance are validated under match-realistic
levels (Reilly, Williams, Nevill, & Frank, 2000). For example,
situations that involve both attackers and defenders then such
a stepwise-discriminant analysis showed that closed-skill
tests can be meaningfully compared among individuals of
tests, such as slalom dribbles among cones, could distinguish
different ages, populations, and nations for more efficient
among elite and sub-elite groups of players (Reilly et al., 2000;
talent identification protocols.
Vaeyens et al., 2006). In addition, Rösch et al. (2000) found that

CONTACT Robbie S Wilson r.wilson@uq.edu.au School of Biological Sciences, The University of Queensland, St Lucia 4072, Australia
© 2018 Informa UK Limited, trading as Taylor & Francis Group
2 R. S. WILSON ET AL.

Most assessments of soccer-specific skills, or even broader involved in the study and was in accordance with ethical
talent identification protocols, focus more on comparisons protocols for the University of Queensland and University of
among groups (elite versus non-elite) rather than among indivi- Londrina.
duals (Ali, Foskett, & Gant, 2008; Ali et al., 2007; Dardouri et al.,
2014; Meylan, Cronin, Oliver, & Hughes, 2010; Reilly et al., 2000;
Study design
Rostgaard, Iaia, Simonsen, & Bangsbo, 2008; Unnithan et al.,
2012; Vaeyens et al., 2006; Waldron & Worsfold, 2010). We designed our experiment to test the dribbling and sprinting
Comparing a group of elite athletes with a group of non-elite performance of all players and quantified performance in a one
athletes provides only a coarse understanding of the attributes vs. one attacking competition across three two-hour sessions as
required to reach elite status. Furthermore, in team sports, pre- a part of their regular training schedule. The one vs. one com-
dicting whether or not an individual will become “elite” is less petition was conducted during sessions 1 and 3 of the experi-
important than predicting how they will perform within elite ment, while the dribbling and sprinting performances were
competitions. The success of an individual depends on that measured during session 2 of the experiment. The dribbling
individual’s ability relative to others within its team or league; and sprinting performance of only 22 individuals was assessed,
in other words, identifying which players could become profes- thus our analyses that explored predictors of one vs. one attack-
sional is not the same as identifying which individuals could ing performance were performed on these 22 individuals only.
become the best professional players. To identify the best players Before testing during any session, players proceeded through
requires that player performance is analysed on an individual their normal 15 min warm-up routine with their coach.
level – a process that is under-utilised in sports science (Wilson
et al., 2016) yet routine in evolutionary biology (Arnold, 1983;
Dribbling performance
Endler, 1986; Kingsolver et al., 2001). Wilson et al. (2017) recently
quantified variation among individual soccer players in skill, We designed our dribbling test to be low-cost, simple to set-
balance, and athleticism, and found they could predict players’ up and requiring limited equipment, so that it could be repli-
success in 11-a-side competitive soccer games. They also found cated by any football player or coach across the world. Our
that an individual’s skill was the best predictor of performance in new assessment protocol measured dribbling speed of indivi-
competitive matches and contribution to the success of the dual players through circuits that require close ball control
whole team, based on a social network analysis of ball move- and penalise players for mistakes. The dribbling performance
ment (Wilson et al., 2017). of each player was tested along five different 30 m long paths
Here, we developed a new assessment protocol of drib- that varied in curvature from 0 radians.m−1 (straight path) to
bling performance that extends previous methods and then 1.37 radians.m−1 (highly curved path). In addition, Each path
tested its association with individual performance in match- consisted of an inner 1 m-wide channel, with the outer bound-
realistic one vs. one games. First, dribbling speed was assessed aries marked with 6 mm black and yellow plastic chain (Kateli,
through circuits that varied in curvature, thus better represent- Brazil). Paths consisted of a series of straight sections that
ing the diversity of dribbling techniques used in match were interspersed with turns that were always 1 m in diameter
situations. Second, we quantified the ability of individual and either 45° (1/8 of a circle), 90° (1/4 of a circle), 135° (3/8 of
players to beat opponents with the ball by staging one vs. a circle), or 180° (1/2 a circle) (Figure 1). Path 1 had no turns
one competitions where players in turn acted as attackers and and consisted of a straight 30 m long path. Path 2 had 6 turns,
defenders. The competition replicates situations where an with five 90° turns and one 135° turn, with a total curvature of
attacker would take on a single defender in order to get 0.37 radians.m−1. Path 3 had 10 turns, with three 45° turns,
around them into a more dangerous attacking position. This two 90° turns and two 135° turns and three 180° turns, with
situation occurs frequently in matches, and players who excel a total curvature of 0.67 radians.m−1. Path 4 had 15 turns, with
in this ability may be better attackers in open-play football. We four 45° turns, four 90° turns and two 135° turns, and five 180°
then tested whether our dribbling assessment protocol could turns, with a total curvature of 1.03 radians.m−1. Path 5 had 20
significantly predict one vs. one attacking performance; and if turns, with five 45° turns, five 90° turns and four 135° turns,
so, whether dribbling was a better predictor of attacking and six 180° turns, with a total curvature of 1.37 radians.m−1.
performance than sprinting performance. Dribbling speed was quantified by recording the total time
taken for an individual to dribble the ball through each 30 m
path. Each player started with the ball behind the starting
Materials and methods position and proceeded through the circuit as fast as possible.
Time was stopped when both the player and ball crossed the
Participants
finish line. Players were asked to complete the path as quickly
We recorded the performance of 25 players from the Junior as possible without going out of the 1 m wide pathway. If the
Team Football Academy of Londrina State University, in ball went outside the path then the test was stopped and the
Londrina, Brazil. Since 2001, Junior Team Football Academy individual repeated the trial after a minimum of 2 minutes
has been used to identify young players for Brazilian profes- rest. If the ball touched the side of the path (the plastic chain)
sional teams. The average age of participants in this study was this was recorded as an error (including faults requiring retest-
16.6 years (SD = 1.1; range 15.4–18.2 years old), and players ing), and the total number of errors was recorded for each
were all high standard and competed in their state’s junior individual for each trial. Although there was no maximum
competition. All players and guardians gave consent to be number of faults allowed per individual, few faults were
JOURNAL OF SPORTS SCIENCES 3

Players moved through each test station in groups of four


or five, and each group visited all five dribbling performance
stations on three separate occasions. The order of the circuit
the players progressed through was Path 1, Path 4, Path 2,
Path 3 and then Path 5. However, each group of players was
randomly assigned a starting station, so that the order of
testing was different for each group. All individuals within
a group completed the path at a given station once, before
the entire group moved on to the next.
The first component of a principal component analysis (PCA)
based on measures of dribbling speed along each of the five
paths (PCD1) explained 80.9% of the variation observed in the
data (Table 1). All vectors of PCD1 loaded in the same direction,
and because larger positive values were indicative of higher
dribbling speed, PCD1 represented a measure of overall drib-
bling performance. The second component of the PCA (PCD2)
explained 8.1% of overall variation, with positive values asso-
ciated with high dribbling speed along the straightest path
(Table 1). We also separately calculated a PCA based on each
individual’s dribbling speeds on all five paths for trials 1, 2 and 3.
Repeatability of these overall measures of dribbling perfor-
mance, based on the first two dimensions of the PCAs, were
then calculated across each three trials.

Sprinting performance
The sprinting performance of each player was tested along the
same five 30 m paths used for the dribbling tests (curvature
range: 0 to 1.37 radians.m−1). On day 2, after all dribbling tests
were completed, players were given a 20 min rest and again
visited the same paths in groups of four or five. The testing
protocol was identical to that described for dribbling above.
The first component of a PCA based on measures of sprinting
speed along each of the five paths (PCS1) explained 74.2% of the
variation observed in the data (Table 1). All vectors of PCS1
loaded in the same direction, and because larger positive values
were indicative of higher sprinting speed, PCS1 represented
Figure 1. The five 30 m-long paths used to assess dribbling and sprinting
a measure of overall sprinting performance. The second compo-
performance for the individual players in our study. The curvature of each
path was (a) 0 radians.m−1, (b) 0.37 radians.m−1, (c) 0.67 radians.m−1, (d) 1.03 nent of the PCA explained 13.7% of the variation and positive
radians.m−1, and (e) 1.37 radians.m−1. The outline of each path was marked with values of PCS2 were associated with high sprinting speed along
6 mm plastic chain and the width of each path was 1 m, and any 45° (1/8 of
the straightest path (Table 1). We also separately calculated
a circle), 90° (1/4 of a circle), 135° (3/8 of a circle) or 180° (1/2 a circle) turns had
a radius of 1 m. The number of 0.5 m long sections (grey) along each path a PCA based on each individual’s sprint speeds on all five paths
shows the lengths between each turn, with a 90° turn represented in green and for trials 1 and 2. Repeatability of these overall measures of sprint
45° in yellow.
performance, based on the first two dimensions of the PCAs,
were then calculated across the two trials.

observed and not likely to affect interpretation of the results.


Repeatability was calculated by comparing a player’s dribbling Table 1. Principal components analysis matrix of dribbling and sprinting per-
speed of each path on their first, second and third attempts. formance (N = 22) along the five paths that differ in curvature from 0 to 1.37
radians.m−1. Factor loadings of each measured variable and the direction in
Two researchers used stopwatches to record the time taken by which they contribute towards the components are shown. See the text for
a player to complete the circuit and these two times were a description of each trait.
averaged and taken as a player’s average time over the path. Dribbling Performance Sprinting Performance
Dribbling by the players and time on the stopwatches were Path curvature (radians.m−1) PCD1 PCD2 PCS1 PCS2
started by counting down from 3 to 1, ensuring the timing of 0 0.75 0.65 0.77 0.56
a player’s movement and beginning of the stopwatches was 0.37 0.96 −0.10 0.92 −0.17
0.67 0.91 −0.22 0.95 0.00
well sequenced. Times were converted to average speed 1.03 0.92 −0.25 0.77 −0.55
through the 30 m paths. 1.37 0.95 0.05 0.87 0.19
Total variance 80.9 8.1 74.2 13.7
4 R. S. WILSON ET AL.

One vs. one attacking competition zone, using any legal tactic, and to touch the ball with any
part of their foot inside the scoring zone. Each bout ended
One vs. one attacking games were conducted in a tournament
when either: (i) the ball was touched by the attacker in the
design (Stuart-Fox, 2006; Wilson, Angilletta, James, Navas, &
scoring zone (successful attack), (ii) the ball was kicked out of
Seebacher, 2007; Wilson et al., 2017), thus allowing a detailed
the field or scoring zone (unsuccessful attack), or (iii) a total
hierarchy of success to be determined for each player. Each
time of 30 s had elapsed without a result (unsuccessful attack).
individual player competed against 8 to 14 opponents in the
If the defender illegally fouled the attacker, then the attack
one vs. one matches. Each match between pairs of players
was deemed successful, but because this occurred in only 15
consisted of 10 individual bouts, with each individual being
out of 1330 total bouts the exclusion or inclusion of these data
the attacker on five occasions and the defender on five occa-
would not have affected the results.
sions, with a total of 1330 individual one vs. one bouts across
the 25 individuals. The first player to attack in a pair was
chosen at random, and after this, players alternated position- Statistical analyses
ing as attacker and defender. Thus, each player in a pair had
the opportunity to attack five times and score up to five Estimates of repeatability for dribbling and sprinting perfor-
points. Individuals were randomly placed into five groups of mance across trials, and one vs. one attacking performances
five individuals, with all individuals in a group playing each across days, were calculated using intra-class correlation coeffi-
other once. cients (Quinn & Keough, 2002). Correlations among perfor-
After this first round, all players were then randomly dis- mances were conducted using Pearson’s product moment
tributed to another five groups of five individuals, with all correlations. We used a linear mixed effects model to determine
individuals playing each other once. These first two rounds how the curvature of each path affected the dribbling and
were conducted in session 1 of the experiment so that all sprinting performance score. Random intercepts were fit for
players competed against 8 unique opponents, providing each player’s ID. To assess the differences between each path
them with 40 attacking and 40 defending opportunities. we used Tukey’s contrasts and a Holm correction to adjust
Players were allowed to rest for 5–10 minutes when switching p-values. To assess the effects of dribbling ability and sprinting
between competitors. In session 3, using a smaller number of speed on the probability that a player was successful in a one-on-
players (N = 17) we conducted an additional number of paired one contest, we fitted a binomial, non-linear mixed effects model
competitions (N = 4–6) that extended our sample size and with a logistic link function. We modelled PCD1, PCD2, total
allowed us to test the repeatability of attacking performance number of errors when dribbling, PCAS1 and PCS2 as fixed effects.
across days. Data from sessions 1 and 3 were used in all Random intercepts were fit for each player’s ID. We used
subsequent analyses. a stepwise approach based on the Akaike’s information criterion
Matches were held on a field 15 m wide and 25 m long, (AIC) to select the most likely model describing the data. All
with a 4 m × 15 m area at one end of the field defined as the analyses were carried out in R 3.4.3 (R Core Team, 2017). Data
scoring zone (Figure 2). Players were allowed to use whatever are reported as means ± standard deviations.
dribbling strategy they preferred, so that attackers with good
close control could lure defenders into close situations near
Results
the scoring zone while those with higher running speeds and
poorer control could attempt to kick the ball past the defen- Repeatability of dribbling and sprinting performance
ders and re-gather the ball in the scoring zone. At the begin-
The dribbling performance of individual players was highly
ning of each bout, the attacker was positioned behind the line
repeatable among trials as measured by intra-class correlation
at the opposite end of the field to the scoring zone, with the
coefficients (ICC), which ranged from an ICC of 0.79 for drib-
defender at the centre of the field. The attacker’s objective
bling along the curviest path to an ICC of 0.90 for the straight-
was to dribble the ball past the defender into the scoring
est path (Table 2). Performance was also very highly

Figure 2. The dimensions of the one vs. one competition field, with the starting position of the attacker and defender shown.
JOURNAL OF SPORTS SCIENCES 5

Table 2. Intra-class correlation coefficients (ICC) of dribbling and sprinting Table 4. Correlation matrix between measures of dribbling performance for
performance along the five different paths that varied in curvature from 0 to each of the five different paths that vary in curvature from 0 to 1.37 radians.
1.37 radians.m−1. The ICCs for the composite measures of dribbling and sprint- m−1 and the sprinting performance and number of errors. In addition, the first
ing performance from the first and second components of principal components components of a principal component analysis (PCA) on dribbling was corre-
analyses (PCA) are provided. lated with the first component of a PCA conducted on sprinting performances
Path curvature and also total number of errors across all paths. All correlations are Pearson’s
(radians.m−1) Dribbling (ICC) CI (± 95%) Sprinting (ICC) CI (± 95%) product moment correlations.
0 0.88 0.75–0.94 0.73 0.37–0.89 Sprinting performance Dribbling errors
0.37 0.86 0.72–0.94 0.83 0.59–0.93 Path curvature Correlation P Correlation P
0.67 0.87 0.75–0.94 0.89 0.73–0.95 (rads.m−1) coefficient (r) value coefficient (r) value
1.03 0.90 0.80–0.96 0.88 0.72–0.95 0 0.69 0.0003 −0.39 0.07
1.37 0.79 0.59–0.91 0.77 0.46–0.90 0.37 0.63 0.002 −0.45 0.03
PC1 0.97 0.94–0.99 0.96 0.92–0.98 0.67 0.56 0.001 −0.37 0.08
PC2 0.72 0.43–0.87 0.69 0.27–0.87 1.03 0.67 0.0007 −0.56 0.007
1.37 0.45 0.04 −0.24 0.28
PC1 0.71 0.002 −0.56 0.007
repeatable when all dribbling measures were combined as
PCD1 (ICC = 0.97) (Table 2).
The sprinting performance of individual players was also
P < 0.001)(Figure 3(a)) from 5.19 ± 0.11 m.s−1 on the straight-
highly repeatable among trials and tasks, and ranged from an
est path to 2.13 ± 0.03 m.s−1 on the curviest one. Based on
ICC of 0.73 for running along the straightest path to an ICC of
post-hoc tests, the average dribbling speed along the straight-
0.89 for the path with a curvature of 0.67 radians.m−1 (Table 2).
est path was significantly faster than all other paths (Appendix
Performances were highly repeatable when all sprinting mea-
Table 1). Players that exhibited greater numbers of errors
sures were combined as PCS1 (ICC = 0.96) (Table 2).
when dribbling the ball along each path had a tendency to
have slower average dribbling speeds (Table 4). However, the
Correlations between performances across different path only significant negative correlations between average drib-
curvatures bling speed and the number of errors was along the second
curviest path (N = 22; r = −0.55; P < 0.01) (Table 4) and
An individual’s dribbling speed along any given path was the second straightest path (N = 22; r = −0.46; P < 0.05). An
significantly positively correlated with their dribbling speed individual’s composite dribbling speed across all paths (PCD1)
along any other path, with correlations ranging from 0.53 to was significantly negatively correlated with the total number
0.95 (Table 3). Sprinting speed along any curved path was also of errors exhibited across all paths combined (N = 22;
positively associated with speed along any other curved path r = −0.58; P < 0.01) (Table 4).
(correlations range from 0.55 to 0.89) (Table 3). The only non- The average sprinting speed of individuals also significantly
significant association was between the straight path and decreased non-linearly with increasing curvature of the 30 m
the second curviest path (N = 22; r = 0.36; P > 0.05). paths (df = 194; F = 1737.8; P < 0.001)(Figure 3(b)). Based on
Dribbling speed was positively correlated with sprinting post-hoc tests, the average sprinting speed along a given path
along each path (Table 4), and was highest along the straight- significantly differed from speeds along all other paths, except
est path (N = 22; r = 0.69; P < 0.001), and weakest along the between the two curviest paths (Appendix Table 2). Average
curviest path (N = 22; r = 0.45; P < 0.05). speeds decreased from 6.93 ± 0.11 m.s−1 on the straight path
to 3.33 ± 0.05 m.s−1 on the curviest one (Figure 3(b)).

Effect of path curvature on dribbling and sprinting


performance One vs. one dribbling competitions and attacking
Average dribbling speed along the 30 m paths decreased non- performance
linearly with increasing curvature (df = 304; F = 239.5; In the one vs. one dribbling competition, attackers successfully
beat defenders and reached the scoring zone 41 ± 17% of the
Table 3. Correlation matrix of measures of dribbling and sprinting performances time (mean ± standard deviation). The best attacker suc-
across the five different paths that vary in curvature from 0 to 1.37 radians.m−1. ceeded on 71% of attempts (N = 60), while the poorest was
Significant correlations between pairs of performance traits are denoted by * for successful on only 9% of attempts (N = 65). For those indivi-
P < 0.01, ** for P < 0.001, and *** for P < 0.0001. All correlations are Pearson’s
product moment correlations. duals that competed in the one vs. one competition on days 1
Path curvature and 3, the repeatability of proportional success was an ICC of
(rads.m−1) 0 0.37 0.67 1.0 0.73. The proportion of attempts that were successful for each
Dribbling Performance individual on day 1 was also significantly positively correlated
0.37 0.66*** with their proportion of attempts that were successful on day
0.67 0.56** 0.92***
1.0 0.53* 0.88*** 0.84*** 3 (N = 17; r = 0.57; P < 0.05).
1.37 0.72*** 0.87*** 0.79*** 0.88*** Dribbling performance (PCD1) and total number of errors
Sprinting Performance when dribbling were the only significant predictors of success
0.37 0.59*
0.67 0.71** 0.89*** in the one vs. one competition in the multivariate model
1.0 0.36 0.74** 0.69** (Appendix Tables 2 and 3; Table 5). Specifically, players with
1.37 0.69** 0.73** 0.79*** 0.55* higher PCD1 values and lower number of errors were more likely
6 R. S. WILSON ET AL.

a b

Figure 3. (a) Dribbling speed, and (b) sprinting speed, along five 30 m paths that varied in curvature from 0 radians.m−1 (straight path) to 1.37 radians.m−1.
Curvature significantly affected both dribbling and sprinting speed (see results in text). Significant differences between paths based on post-hoc tests are denoted
by different letters in each panel, such that dribbling along the straight path is significantly different to dribbling speed on all other paths. Dots represent individual
datum points and horizontal lines are the means.

Table 5. Summary results from the binomial logistic regression, testing the attacking matches was significantly positively correlated with
effects of dribbling ability and sprinting speed on the probability that a player
their overall dribbling performance (PCD1) (N = 22; r = 0.70;
was successful in a one vs. one contest. The factors included in the final model
are the first two components of a principal components analysis conducted on P < 0.001) (Figure 4(a)) and overall sprinting performance (PCS1)
all five measures of dribbling speed (PCD1 and PCD2), the first component of the (N = 22; r = 0.49; P < 0.05)(Figure 4(b)). Neither PCD2 (N = 22;
principal components analysis conducted on all five measures of dribbling
r = 0.16; P > 0.05) nor PCS2 were significantly associated with
speed (PCS1), and total number of errors that occurred across all five dribbling
paths. The significant factors in this model are the metric of overall dribbling attacking success (N = 22; r = −0.01; P > 0.05).
performance and number of errors when dribbling. As PCD1 increases and
number of errors decreases then so does the probability of success for an
individual player.
Path (radians.m−1) Estimate Std. Error z value Pr(>|z|) Discussion
Intercept −0.007 0.169 −0.043 0.965 In this study, we developed a novel and low-cost series of
PCD1 0.247 0.065 3.770 0.0001
PCD2 0.087 0.127 0.685 0.493 dribbling skill tasks and then explored the relationship
PCS2 0.218 0.143 1.524 0.127 between dribbling speed with performance in a match-
errors −0.036 0.013 −2.769 0.005 realistic one vs. one competition between an attacker and
a defender. We were able to rank players based on measures
of one vs. one attacking ability that was repeatable across
to be successful in the one vs. one competition. However, in the
days. Players used a variety of tactics to beat defenders,
univariate correlations, a player’s success in one vs. one
including the use of close dribbling to navigate around

a b

Figure 4. Relationship between the proportion of successful attempts in the one vs. one competition for each individual with their overall metric of dribbling speed
based on the first component of a PCA conducted on individual dribbling speeds across all five paths (PCD1) (N = 22; rp = 0.70; P = 0.0002)(a) and overall measure of
sprinting speed based on the first component of a PCA conducted on individual sprinting speeds across all five paths (PCS1) (N = 22; rp = 0.49; P = 0.02)(b).
JOURNAL OF SPORTS SCIENCES 7

defenders or longer kicks combined with fast sprint speeds to for dribbling performance and previously for multivariate
run past them. We found that attacking success in the one vs. assessments (Wilson et al., 2016), could markedly improve
one competition was best predicted by overall dribbling per- talent identification and development protocols (Höner et al.,
formance across drills, and more than 50% of the variation in 2017, 2015). Recent work (Wilson et al., 2017) shows that the
attacking success was explained by our composite measure of best players in matches are those with higher levels of skill,
dribbling performance alone. Although sprint speed was posi- rather than athletic traits like speed, agility, strength, or sta-
tively correlated with dribbling speed, we found that dribbling mina. A bias toward athletic traits is now holding back the
speed and total number of errors were the driving factors in field of soccer science, and a forward focus on skill assessment
attacking performance based on our multivariate statistical could revolutionise the structure of junior development pro-
model. Sprint capabilities may provide an upper limit to drib- grams, just as developments in exercise physiology have trans-
bling speed, but a player’s ability to manipulate and control formed the development of power and endurance of Olympic
the ball when moving at speed remains the best predictor of athletes and the fitness training of professional footballers.
attacking ability in one vs. one matches. Taken together, our In our study, we designed and tested the reliability of
closed-skill dribbling tasks can be used to identify which a multi-component, complex dribbling test that is more
players will excel at taking on and beating opponents in detailed than any other dribbling test previously used. In
match-realistic situations. Because our metric of dribbling addition, we validated our dribbling tests by showing they
speed is independent of other player’s performances (unlike can predict players performances in realistic 1 vs 1 attacking
one vs. one performance), it provides an efficient method for situations, allowing one to predict the performances of indivi-
predicting an individual’s relative attacking performance. Thus, dual players from different countries, ages and experiences
our dribbling test is low-cost, simple to set-up and requires that are independent of the defenders they pay against. We
limited equipment, and can be used to compare individuals analysed an elite Brazilian junior football team because of the
from across developmental stages, ages, playing levels, and cultural importance they place upon dribbling and attacking
nations. skills. Our study participants were all highly proficient at drib-
Closed-skill tests must be able to reliably measure an indi- bling but we still observed substantial variance among indivi-
vidual’s performances so that any variation detected among duals. The next logical step for our research would be to
individuals is meaningful. Performances in our five dribbling compete various coaching methodologies and to then quan-
tasks were highly repeatable within individuals, with intra- tify any improvement in dribbling speed along our multiple
class correlation coefficients (ICC) ranging between 0.79 to curved paths. Then, further assessments may highlight
0.90. These ICCs are similar to those from previous studies whether any coaching-mediated improvements in dribbling
that measured dribbling speed through sets of cones (Ali, speed also translate into subsequent improvements in attack-
2011). However, our combined metric of dribbling perfor- ing performance in one vs. one competitions. The use of
mance, based upon all paths (PCD1), had the very high ICC of experiments to assess the relative benefits of alternate coach-
0.97, indicating that our method is an extremely reliable tool ing methodologies for match-realistic performance could sub-
for ranking players’ dribbling abilities – a skill pre-requisite to stantially improve coaching programs and the development of
attacking success in matches. junior technical development. We believe this underlies many
The ability to take on defenders and beat them using close of the advantages for research programs that quantify skill
ball control and dribbling speed is such an important offensive and its relevance to match-realistic performance.
strategy in soccer, that it is surprising that comparatively little
research has been devoted to assessing and improving this
skill or, in fact, many of the other fundamental skills of soccer Disclosure statement
(Höner, Leyhr, & Kelava, 2017; le Gall, Carling, Williams, & Reilly,
No potential conflict of interest was reported by the authors.
2010; Reilly et al., 2000; Vaeyens, Lenoir, Williams, &
Philippaerts, 2008). Unfortunately, most quantitative protocols
for talent identification in soccer focus more on generic ath- ORCID
letic traits such as speed, strength, agility and endurance, Fabio Giuliano Caetano http://orcid.org/0000-0003-3602-9684
rather than skill (le Gall et al., 2010; Reilly et al., 2000; Paulo Roberto Pereira Santiago http://orcid.org/0000-0002-9460-8847
Vaeyens et al., 2008). In fact, the vast majority of scientific Sergio Augusto Cunha http://orcid.org/0000-0003-1927-0142
research occurs on soccer players still focuses on how to Felipe Arruda Moura http://orcid.org/0000-0002-0108-7246
improve speed, strength and fitness, and how to facilitate
rapid recovery after matches or following injury (Ali, 2011;
Höner et al., 2017; Höner, Votteler, Schmid, Schultz, & Roth, References
2015; Wilson et al., 2016; Wilson et al., 2017). This is probably Ali, A. (2011). Measuring soccer skill performance: A review. Scandinavian
because athletic traits are faster, easier, and more reliably Journal of Medicine & Science in Sports, 2, 170–183.
measured (Wilson et al., 2016), but also because of disagree- Ali, A., Foskett, A., & Gant, N. (2008). Validation of a soccer skill test for use
ment in the perceived relevance of certain skills (Ali, 2011) and with females. International Journal of Sports Medicine, 29, 917–921.
a long-standing resistance of coaches to manipulating coach- Ali, A., Williams, C., Hulse, M., Strudwick, A., Reddin, J., Howarth, L., …
Hirst, M. (2007). Reliability and validity of two tests of soccer skill.
ing regimes and measure their impact on skill uptake. Despite Journal of Sports Sciences, 25, 1461–1470.
these potential barriers, the development of a comprehensive Arnold, S. J. (1983). Morphology, performance and fitness. American
protocol for soccer-specific skill, including that provided here Zoologist, 23, 347–361.
8 R. S. WILSON ET AL.

Coelho e Silva, M. J., Silva, M. J., Figueiredo, A. J., Simoes, F., Seabra, A., Rösch, D., Hodgson, R., Peterson, T. L., Graf-Baumann, T., Junge, A.,
Natal, A., & Malina, R. M. (2010). Discrimination of U-14 soccer players Chomiak, J., & Dvorak, J. (2000). Assessment and evaluation of football
by level and position. International Journal of Sports Medicine, 31, performance. American Journal of Sports Medicine, 28, S29–S39.
790–796. Rostgaard, T., Iaia, F. M., Simonsen, D. S., & Bangsbo, J. (2008). A test to
Dardouri, W., Selmi, M. A., Sassi, R. H., Gharbi, Z., Rebhi, A., & Moalla, W. evaluate the physical impact on technical performance in soccer.
(2014). Reliability and discriminative power of soccer-specific field tests Journal of Strength Cond Res, 22, 283–292.
and skill index in young soccer players. Science & Sports, 29, 88–94. Russell, M., Benton, D., & Kingsley, M. (2010). Reliability and construct
Endler, J. A. (1986). Natural selection in the wild. Princeton: Princeton validity of soccer skills tests that measure passing, shooting, and
University Press. dribbling. Journal of Sports Sciences, 28, 1399–1408.
Haaland, E., & Hoff, J. (2003). Non-dominant leg training improves the bilateral Stuart-Fox, D. (2006). Testing game theory models: Fighting ability and
motor performance of soccer players. Scand J Med Sci Sports, 13, 179–184. decision rules in chameleon contests. Proceedings of the Royal Society of
Höner, O., Leyhr, D., & Kelava, A. (2017). The influence of speed abilities London Series B: Biological Sciences, 273, 1555–1561.
and technical skills in early adolescence on adult success in soccer: A Unnithan, V., White, J., Georgiou, A., Iga, J., & Drust, B. (2012). Talent
long-term prospective analysis using ANOVA and SEM approaches. identification in youth soccer. Journal of Sports Sciences, 30, 1719–1726.
PLoS ONE, 12, e0182211. Vaeyens, R., Lenoir, M., Williams, A. M., & Philippaerts, R. M. (2008). Talent
Höner, O., Votteler, A., Schmid, M., Schultz, F., & Roth, K. (2015). identification and development programmes in sport - Current models
Psychometric properties of the motor diagnostics in the German foot- and future directions. Sports Medicine, 38, 703–714.
ball talent identification and development programme. Journal of Vaeyens, R., Malina, R. M., Janssens, M., Van Renterghem, B., Bourgois, J.,
Sports Sciences, 33, 145–159. Vrijens, J., & Philippaerts, R. M. (2006). A multidisciplinary selection
Huijgen, B. C., Elferink-Gemser, M. T., Post, W., & Visscher, C. (2010). model for youth soccer: The ghent youth soccer project. British
Development of dribbling in talented youth soccer players aged 12–- Journal of Sports Medicine, 40, 928–934.
19 years: A longitudinal study. Journal of Sports Sciences, 28, 689–698. Vilar, L., Araú jo, D., Davids, K., & Travassos, B. (2012). Constraints on
Kingsolver, J. G., Hoekstra, H. E., Hoekstra, J. M., Berrigan, D., Vignieri, S. N., competitive performance of attacker-defender dyads in team sports.
Hill, C. E., … Beerli, P. (2001). The strength of phenotypic selection in Journal of Sports Sciences, 30, 459–469.
natural populations. American Naturalist, 157, 245–261. Waldron, M., & Worsfold, P. (2010). Differences in the game specific skills of elite
le Gall, F., Carling, C., Williams, M., & Reilly, T. (2010). Anthropometric and and sub-elite youth football players: Implications for talent identification.
fitness characteristics of international, professional and amateur male International Journal of Performance Analysis in Sport, 10, 9–24.
graduate soccer players from an elite youth academy. Journal of Science Wilson, R. S., Angilletta, M. J., James, R. S., Navas, C., & Seebacher, F.
and Medicine in Sport, 13, 90–95. (2007). Dishonest signals of strength in male slender crayfish (Cherax
Meylan, C., Cronin, J., Oliver, J., & Hughes, M. (2010). Talent identification in dispar) during agonistic interactions. The American Naturalist, 170,
soccer: The role of maturity status on physical, physiological and technical 284–291.
characteristics. International Journal of Sports Science & Coaching, 5, 571–592. Wilson, R. S., David, G. K., Murphy, S. C., Angilletta, M. J., Jr, Niehaus, A. C.,
O’Reilly, J., & Wong, S. H. S. (2012). The development of aerobic and skill Hunter, A. H., & Smith, M. D. (2017) Skill not athleticism predicts
assessment in soccer. Sports Medicine, 42, 1029–1040. individual variation in match performance of soccer players 284.
Proceedings of the Royal Society of London Series B: Biological Sciences,
Quinn, G. P., & Keough, M. J. (2002). Experimental design and data analysis for
284, 20170953
biologists. Cambridge, UK: Cambridge University Press.
Wilson, R. S., James, R. S., David, G., Hermann, E., Morgan, O. J., Niehaus, A.
R Core Team. (2017). R: A language and environment for statistical comput- C., Hunter, A. H., and Smith, M. (2016). Multivariate analyses of indivi-
ing. Vienna, Austria: R Foundation for Statistical Computing. https:// dual variation in soccer skill as a tool for talent identification and
www.R-project.org/ development: Utilizing evolutionary theory in sports science. Journal
Reilly, T., & Holmes, M. (1983). A preliminary analysis of selected soccer of Sports Sciences. doi:10.1080/02640414.2016.1151544.
skills. Physical Ed Rev., 6, 64–71. Zago, M., Piovan, A. G., Annoni, I., Ciprandi, D., Iaia, F. M., & Sforza, C.
Reilly, T., Williams, A. M., Nevill, A., & Frank, A. (2000). A multidisciplinary (2016). Dribbling determinants in sub-elite youth soccer players.
approach to talent identification in soccer. Journal of Sports Sciences, 18, Journal of Sports Sciences, 34, 411–419.
695–702.

You might also like