Content uploaded by Oliver R Runswick
Author content
All content in this area was uploaded by Oliver R Runswick on Jan 25, 2024
Content may be subject to copyright.
1
Visual exploratory activity in elite women’s soccer: An analysis of the
1
UEFA Women’s European Championship 2022
2
3
James Feista*, Naomi Datsond, Oliver R. Runswickb, Alice Harkness-Armstrongc
4
and Chris Pococka
5
6
aInstitute of Applied Sciences, University of Chichester, U.K
7
bDepartment of Psychology, Institute of Psychiatry Psychology & Neuroscience, King’s
8
College London, U.K.
9
cSchool of Sport, Rehabilitation and Exercise Sciences, University of Essex, U.K.
10
dDepartment of Sport and Exercise Sciences, Manchester Metropolitan University, U.K.
11
12
13
Correspondence concerning this article should be addressed to James Feist, Institute of
14
Applied Sciences, University of Chichester, Chichester, PO19 6PE. Email: J.Feist@chi.ac.uk
15
16
Open Science Framework Project Link:
17
https://osf.io/2cnh7/?view_only=613db7aaada144e1abfefb648a0b272f
18
19
ORCID
20
James Feist https://orcid.org/0009-0007-7708-925X
21
Naomi Datson https://orcid.org/0000-0002-5507-9540
22
Oliver Runswick https://orcid.org/0000-0002-0291-9059
23
Alice Harkness-Armstrong https://orcid.org/0000-0002-7258-4469
24
Chris Pocock https://orcid.org/0000-0001-5929-7273
25
26
27
28
29
2
Visual exploratory activity in elite women’s soccer: An analysis of the
30
UEFA Women’s European Championship 2022
31
Recent research has developed understanding of the technical and tactical
32
determinants of success in elite women’s soccer, however a lack of research
33
exists on analysing how elite female players visually explore their environment to
34
support skilled soccer performance. This study aimed to describe the visual
35
exploratory activity (VEA) of elite female central midfield players and
36
understand the relationships between VEA, performance with the ball and
37
specific contextual and situational factors. Thirty female central midfield players
38
(M age = 26.7 years, SD = 3.8) from the eight teams who competed in the knock-
39
out stages of UEFA Women’s European Championship 2022 were analysed.
40
Television broadcast and UEFA tactical footage were combined to analyse
41
players across the seven knock-out stage matches, totalling 1,038 individual ball
42
possessions. The mean scan frequency before receiving the ball was 0.35
43
scans/second. Results showed pitch location when receiving the ball to be the
44
main predictor of scan frequency, which in turn predicted action result (p =
45
0.003) and turn with the ball (p = 0.003). Scan frequencies were lower compared
46
to men’s elite and academy players. This study sets a platform for experimental
47
research to further our understanding of VEA and performance with the ball in
48
women’s soccer.
49
Key Words: central midfielders, scan frequency, women’s soccer (football),
50
visual perception, exploratory activity.
51
52
53
54
55
56
57
58
59
60
3
Introduction
61
Soccer is a dynamic invasion game that requires players to have awareness of the movements
62
of the ball, teammates, and opposition players (Pokolm, 2021). Skilled performance requires
63
players to visually explore their surroundings to identify opportunities for action (McGuckian
64
et al., 2018). Visual exploratory activity (VEA), or ‘scanning’ as typically referred to by
65
coaches (Eldridge et al., 2023), when one’s team is in possession of the ball can be defined as
66
a head or body movement where a player’s face is temporarily directed away from the ball to
67
locate teammates, opposition players or empty space, before engaging with the ball (Jordet et
68
al., 2020). In men’s soccer, players who engage in more frequent VEA typically perform
69
more successful actions with the ball (e.g., higher pass success rates; Aksum et al., 2021;
70
Jordet et al., 2020). However, there is a lack of understanding of how elite female players
71
engage in VEA to support skilled performance. Recent literature has identified differences
72
between male and female soccer in both tactical elements such as pass accuracy, ball
73
recovery time (Pappalardo et al., 2021) and the start and development of possession
74
(Mitrotasios et al., 2022), as well as specific physiological characteristics (de Araújo et al.,
75
2020). With research identifying tactical differences between men’s and women’s soccer, and
76
VEA having previously been shown to be important for skilled performance in men’s soccer,
77
there is a need to analyse VEA and performance with the ball in elite women’s soccer.
78
Women’s soccer is currently experiencing a dramatic increase in popularity and
79
professionalism (Griffin et al., 2020; Okholm Kryger et al., 2021), with the UEFA Women’s
80
European Championship 2022 (UEFA Women’s Euro 2022) setting a record aggregate
81
attendance of 574,875 and the Championship having been watched by over 365 million
82
people globally (UEFA, 2022). To date, research into women’s soccer has largely focused on
83
the physical (Vescovi et al., 2021) and physiological demands of the game (Datson et al.,
84
2014; Martínez-Lagunas et al., 2014). Current literature has also emphasised the importance
85
4
of understanding technical and tactical characteristics of the game (de Jong et al., 2020; de
86
Jong et al., 2022). When compared to elite men’s soccer, elite women’s soccer appears to
87
adopt a more attacking style of play, possession is lost more frequently, and passes are
88
performed with less accuracy (Bradley et al., 2014; Garnica-Caparros & Memmert, 2021).
89
These differences could have potential links to how a player visually explores their
90
environment to inform their subsequent action with the ball. Previous literature has also
91
found the more successful teams in elite women’s soccer are those who maintain longer
92
spells of possession (Ivn-Baragao et al., 2022; Maneiro et al., 2020; Soroka & Bergier,
93
2010), make more passes resulting in goal scoring opportunities (Kubayi & Larkin, 2020),
94
and have high interconnectivity with more successful ball transfers and effective ball
95
movements (de Jong et al., 2022). It appears that successful women’s teams are those which
96
are highly interconnected and able to effectively transfer and move the ball quickly to create
97
goal scoring opportunities, factors which are related to the ability to effectively pick up
98
information from the environment. Also, with central midfield players being mainly located
99
in central areas of the pitch, research has suggested central midfielders may play a crucial
100
role in successful ball transfers in teams that achieve international or domestic success (de
101
Jong et al., 2022). These findings demonstrate the importance of maintaining possession and
102
highlight the potential significance of VEA in elite women’s soccer.
103
Previous studies investigating VEA in elite men’s soccer have focused on
104
understanding the influence of contextual factors on performance with the ball (Aksum et al.,
105
2021; Jordet et al., 2020; Pokolm et al., 2022). Notable findings have identified positive
106
relationships between VEA and pass success (Aksum et al., 2021; Jordet et al., 2013), as well
107
as VEA being constrained by pitch location, playing position, phase of play (McGuckian et
108
al., 2020) and opponent pressure (Jordet et al., 2020). More specifically, research has found
109
central midfield players have higher scan frequencies compared to wide players (Pocock et
110
5
al., 2019), central defenders and strikers (Aksum et al., 2020; Jordet et al., 2020). Central
111
midfielders are frequently required to pass the ball forwards and turn with the ball,
112
highlighting the need for players to have a 360-degree visual input to pick up information
113
(Phatak & Gruber, 2019). Despite comprehensive observational studies into VEA in soccer
114
(see Jordet et al., 2020; Pokolm et al., 2022), there is limited evidence on the influence of
115
game-related contextual and situational factors on a female soccer player’s VEA and
116
subsequent performance with the ball.
117
The importance of VEA in soccer and the influence of contextual factors can be
118
conceptually explained through the cyclical relationship between perception and action
119
(Gibson, 1979). Soccer players engage in eye, head, and body movements to pick-up the most
120
relevant environmental information, therefore recognising affordances (Pokolm et al., 2022;
121
McGuckian et al., 2018). Affordances can be defined as opportunities for action that the
122
environment provides an individual in relation to an individual’s action capabilities (Gibson,
123
1979; Fajen et al., 2009). For example, a central midfielder may receive a pass from a team-
124
mate whilst facing the goal their team is defending. Prior to the ball arriving, the player may
125
search their environment for affordances, dependent on the location of teammates, opposition
126
players and empty spaces located behind them. Through exploring the environment and
127
depending on their own action capabilities, the player can recognise relevant affordances (e.g.
128
turn; pass; dribble) and prospectively control actions with the ball (Fajen et al., 2009). A
129
higher frequency of VEA could therefore underpin the search for more information to act
130
upon, which may be linked to more effective performance with the ball.
131
Despite increasing interest in elite women’s soccer, a lack of empirical evidence
132
exists to investigate VEA and its contribution to successful performance. Previous literature
133
has highlighted the need further investigate the technical and tactical match-play
134
characteristics in women’s soccer to gain a more holistic insight into match performance
135
6
(Harkness-Armstrong et al., 2022). Taking this into account alongside the growing body of
136
work into VEA in elite men’s soccer (see Eldridge et al., 2023; Jordet et al., 2020), there is a
137
need to understand the role VEA plays in elite women’s soccer, particularly in central
138
midfield players. Therefore, the aim of this study is to describe the visual exploratory
139
activity of elite female central midfield players and understand the relationships between
140
VEA, performance with the ball and specific contextual and situational factors in elite
141
women’s soccer.
142
Figure 1 presents the hypothesised relationships based on current research evidence.
143
Each arrow represents a potential relationship between VEA (scan frequency), contextual
144
(“on or around the ball”), situational (“off the ball”) and performance with the ball factors. It
145
is hypothesised that central pitch locations will elicit higher scan frequencies compared to
146
wide locations on the pitch and players experiencing higher amounts of opponent pressure
147
will result in lower scan frequencies compared to players experiencing lower amounts of
148
opponent pressure. Higher scan frequencies will be observed when the score line is a draw
149
compared to winning and losing and when the score line is losing compared to winning.
150
These hypotheses are informed by findings in men’s soccer from Jordet et al. (2020) who
151
found game standing to be significantly related to scan frequency. Therefore, it is predicted in
152
the current study that higher scan frequencies will be observed when the score line is a draw
153
compared to winning and losing and when the score line is losing compared to winning.
154
Higher scan frequencies will be observed in earlier stages of a game (between 0-15 minutes)
155
compared to later stages of a game (>75 minutes) as well as in the final compared to the
156
semi-finals and quarter final matches. Due to limited research that currently exists
157
investigating the influence of contextual factors on scan frequency, these hypotheses are
158
presented following tendencies and trends found in previous literature in elite men’s soccer
159
(see Fernandes et al., 2020; Jordet et al., 2020).For the factors that may be influenced by scan
160
7
frequency, it is hypothesised that higher scan frequencies will result in more successful
161
actions with the ball, more forward passes compared to sideways and backwards passes, more
162
passes completed over greater distances (e.g., 15-34 m and 35 m +) compared to shorter
163
passes (e.g., 0-14 m), more forwards and sideways orientated body positions compared to a
164
backwards orientated body position and more turns with the ball.
165
Method
166
167
**Insert Figure 1 near here**
168
169
170
Participants
171
Participants were thirty female central midfield players (M age = 26.7 years, SD = 3.81) from
172
the eight teams who competed in the knock-out stages of UEFA Women’s Euro 2022.
173
Knock-out stage matches were selected to include the top eight European teams in the
174
championship, similar to the approach of Aksum et al. (2021). All players satisfied the
175
following inclusion criteria: i) to have competed in a minimum of two out of three group-
176
stage matches, ii) to have accumulated a minimum of 150 minutes playing time over the
177
course of the UEFA Women’s Euro 2022 and iii) to be classified as a central midfield player
178
based upon UEFA tactical line-ups for each knock-out stage match. The current
179
investigations inclusion criteria is similar to that reported in previous literature into VEA in
180
men’s soccer (Phatak & Gruber, 2019). Due to the observational nature of the study involving
181
elite female soccer players in their natural sport setting (matches of an international
182
tournament), no informed consent was gained due to data being analysed using broadcast
183
8
footage which was publicly available. Ethical approval was granted from the lead author’s
184
institution.
185
Footage
186
Two types of match footage were obtained for the current study. Firstly, broadcast
187
angle footage was obtained through screen-recording publicly available televised footage, as
188
well as UEFA tactical camera (wide angle) footage. The video quality of the UEFA tactical
189
camera footage was 1920x1080 (‘Full HD’). All footage was then imported into Hudl
190
Sportscode (Hudl, Nebraska, USA) with the broadcast footage synced and aligned with
191
UEFA wide angle footage which resulted in split screen footage being generated to enable all
192
players to be on screen at all times. All matches were analysed on a Dell Computer
193
(Windows 10) at a resolution of 1920 x 1080 connected to an Apple MacBook Pro (Version
194
12.6.3). In instances where a player was visible on the broadcast footage and then left the
195
televised picture, the remainder of the instance was analysed using the UEFA tactical camera
196
footage. A total of 402 instances were analysed solely using broadcast footage and 636
197
instances analysed using the broadcast footage and tactical camera footage specifically
198
provided by UEFA.
199
Procedures
200
Prior to data collection, pilot testing was conducted by analysing the Women’s FA
201
Cup Final 2022 to allow the researcher to identify any issues with the operational definitions
202
and code window. As a result of the pilot test, minor changes were made to the operational
203
definitions of action type, turn with the ball and line break. A ball possession in the current
204
study was defined as a player receiving the ball from a teammate and performing an action
205
with the ball (e.g., a pass). For an instance (individual ball possession) to be included in the
206
final analysis all instances were required to meet specific inclusion criteria (see Table 1).
207
9
208
**Insert Table 1 near here**
209
210
Data Collection
211
Phase 1
212
All knockout matches were coded using a bespoke code window, which included the
213
creation of a ‘contextual’ (including performance with the ball factors) and ‘situational’
214
window. All instances were labelled with situational factors (“off the ball”) which were then
215
edited to capture the final 10 seconds prior to the analysed player receiving the ball from a
216
teammate, or from the point when the analysed player’s team won possession of the ball
217
during the 10 s interval. The 10 second cut-off point was chosen to allow comparison to be
218
made with previous studies into VEA in elite male soccer (see Aksum et al., 2021; Jordet et
219
al., 2020).
220
Phase 2
221
Contextual factors (“on or around the ball”) were labelled with the exception of pass
222
distance and opponent pressure. To adequately analyse the players’ scans, the magnifying
223
trackable zoom feature tool was utilised to track the analysed players’ scans throughout the
224
10 second interval. This magnifying zoom feature was placed over the individual player
225
being analysed and then tracked the movements of the player through the 10 second interval.
226
This feature was used in combination with reducing the speed of each instance by 50% in
227
order to accurately capture all head and or body movements of the analysed players between
228
the 5-10 second interval. The ruler feature was utilised to measure pass distance and
229
opponent pressure which uses 3-D calibration technology where X and Y coordinates track
230
10
the movements of the players and the ball. Similar methods have been previously used by
231
OPTA statistics which has demonstrated high levels of reliability (Bradley et al., 2007; Liu et
232
al., 2013). Known distances of goal-line to six yard box, goal-line to 18-yard box and goal-
233
line to penalty spot were measured and checked for accuracy of the ruler. For all passes
234
analysed, the ruler measuring feature in Hudl Studio which uses 3-D tracking and calibration
235
technology measured the exact point from which the ball left the analysed player’s foot and
236
was then either received by a teammate, intercepted by an opposition player or the exact point
237
in which the ball left the pitch. Instances were then cut to the exact point at which the
238
analysed player received the ball from a teammate. Finally, instances from Sportscode
239
timelines were exported as CSV files and transferred to Microsoft Excel (Microsoft
240
Corporation, Washington, USA, Version 16.67) for data analysis purposes.
241
Measures and Variables
242
The following variables were analysed: scan frequency, body orientation and
243
performance with the ball. Performance with the ball was split up into action type (with the
244
final action of a pass also including pass distance, pass direction and lines broken), action
245
result and turn with the ball. Pass distance categories were classified following previous
246
research in elite women’s soccer (Mara et al., 2012). We present scan timing across the final
247
five seconds prior to ball contact as a result of all 1,038 instances analysed capturing the
248
analysed players VEA across the final five seconds prior to ball contact because a number of
249
instances were six, seven, eight and nine seconds in duration. To gain a more comprehensive
250
insight into how elite female central midfield players visually explore their environment, the
251
contextual variables of opponent pressure and pitch location were also investigated. All
252
operational definitions were informed by previous research and validated by a UEFA A
253
License Football Coach with 22 years soccer coaching experience (see supplementary
254
material for list of operational definitions).
255
11
Data Analysis
256
Reliability
257
An independent observer with three years’ experience as a soccer analyst performed
258
additional coding on all variables to assess inter-rater reliability. A total of 156 individual ball
259
possessions were re-analysed across two matches for both inter and intra-rater reliability
260
totalling 15% of the entire sample, similar to samples presented in previous research in elite
261
men’s soccer of 10% and 8.2%, respectively (Aksum et al., 2021; Jordet et al., 2020). Intra-
262
rater reliability was completed following a six-week gap to minimise the chances of any
263
potential learning effects. Intra-class correlations (ICC) were utilised for the continuous
264
variable of number of scans which formed the basis variable for scan frequency. ICC were
265
assessed following Cicchetti (1994) criteria to understand the strength of agreement between
266
two separate coders and repeated observations of the same coder (see Table 2). For all
267
remaining categorical variables, Cohen’s Kappa values (Cohen, 1960) were produced for
268
both inter and intra-rater reliability with the strength of agreements classified following
269
Landis and Koch (1977) criteria (see Table 3).
270
**Insert Table 2 near here**
271
272
**Insert Table 3 near here**
273
274
Statistical Analysis
275
All statistical analysis was conducted in RStudio (R Core Team. R, 2022). As a result
276
of only achieving a moderate agreement for the variable of ‘lines broken’ for inter-rater
277
reliability, the variable was not included in any statistical modelling. To achieve the first aim
278
12
of the study, descriptive statistics were presented for VEA (scan frequency) against each
279
variable of interest. To achieve the second aim, a linear mixed model (LMM) was built to
280
understand the relationships between scan frequency and contextual and situational variables,
281
with mixed effects logistic regression models developed to understand relationships between
282
scan frequency and performance with the ball variables.
283
A LMM was developed with the dependent variable of scan frequency and fixed
284
effects of pitch location, opponent pressure, game state, stage of competition and time in the
285
game. Repeated measurements in the data were accounted for within the random effects
286
structure of subject (player) nested within fixture. The lme4 package (see Bates et al., 2014)
287
in RStudio was used to fit the LMM. The emmeans package was used provide estimate
288
means for each variable, and the results were reported as mean ± SE. Tukey's pairwise
289
comparisons were conducted to identify differences between individual fixed effects, with
290
statistical significance set at p < 0.05. The effsize package was used to calculate effect size
291
(ES), which was classified as trivial (<0.2), small (0.2-0.59), moderate (0.6-1.19), large (1.2-
292
1.99), or very large (>2.0) (Batterham & Hopkins, 2006). Effects were considered unclear if
293
the 90% confidence intervals included both positive and negative values below 0.2 (Hopkins
294
et al., 2009). The assumptions of linearity, normality of the distribution of the model and
295
homogeneity of variance were verified visually.
296
Mixed effects logistic regression models with separately considered dependent
297
variables (performance with the ball variables, see Figure 1) were developed with scan
298
frequency as a fixed effect. Mixed effects ordinal logistic regression was performed for the
299
variable pass distance using the Ordinal package (Christensen, 2018). Mixed effects binomial
300
logistic regression was performed for the variables of action result and turn with the ball
301
using the lme4 package (Bates et al., 2014). Mixed effects multinomial logistic regression
302
was performed for the variables of pass direction, action type and body orientation using the
303
13
Mclogit package (Elff, 2021). The random effects structure of subject (player) nested within
304
fixture was maintained. The summary and anova functions in RStudio were used to produce
305
estimates, standard errors, z-values, and p-values for the separate models built. Through
306
utilising mixed effect models, this enabled us to examine the condition/factor of interest
307
while accounting for variability within and across participants and items (Brown, 2021). For
308
all models developed the assumptions of the distribution of the model, linearity and
309
homogeneity of variance were verified visually, with the assumption of proportional odds
310
satisfied visually for the mixed effects ordinal logistic regression. Statistical significance was
311
set at p < 0.05.
312
Results
313
Description of VEA behaviours
314
Central midfielders (n = 30) recorded a mean scan frequency of 0.35 0.17
315
scans/second (scans/s) prior to receiving the ball in the knock-out stages of the UEFA
316
Women’s Euro 2022 (n instances = 1,038). An average of 34 instances per player were
317
analysed across the knockout stages (SD = 21.75, min = 7, max = 93). Across the final five
318
seconds prior to receiving the ball, the highest mean scan frequency was observed in the final
319
1-2 seconds before ball contact (see Figure 2). Table 4 presents mean scan frequencies and
320
the number of instances per variable for all analysed variables except pitch location. Figure 3
321
presents mean scan frequencies, standard deviations and the number of instances analysed
322
across twelve pitch locations.
323
324
**Insert Figure 2 near here**
325
326
327
14
**Insert Table 4 near here**
328
329
330
331
332
333
**Insert Figure 3 near here**
334
335
Pitch Location and Scan Frequency
336
Table 5 shows the estimated means ± SE for all individual pitch locations. Pairwise
337
comparisons revealed players performed significantly more scans in DMCL compared to
338
ACL pitch locations (0.39 ± 0.02 vs 0.26 ± 0.04 scans/s; p = 0.019; moderate ES: 0.83 ±
339
0.22) and in DMCR compared to ACL pitch locations (0.39 ± 0.02 vs 0.26 ± 0.04 scans/s; p =
340
0.013; moderate ES: 0.84 ± 0.22). Significant differences were also identified in DMR
341
compared to ACL pitch locations (0.39 ± 0.02 vs 0.26 ± 0.04 scans/s; p = 0.044; moderate
342
ES: 0.83 ± 0.24) and in DMCL compared to AMCL (0.39 ± 0.02 vs 0.39 ± 0.02 scans/s; p =
343
0.007; small ES: 0.48 ± 0.12) and in DMCR compared to AMCL (0.39 ± 0.02 vs 0.31 ± 0.02
344
scans/s; p = 0.004; small ES: 0.49 ± 0.12). Figure 4 presents all small and moderate effect
345
sizes (90% Confidence intervals) within pitch location. All other pairwise comparisons
346
revealed trivial or unclear effect sizes.
347
Opponent Pressure and Scan Frequency
348
There were no significant differences between different amounts of opponent
349
pressure. Pairwise comparisons revealed players performed more scans when opponent
350
15
pressure was at 7-9 m compared to 0-3 m (0.35 ± 0.03 vs 0.31 ± 0.02 scans/s; small ES: 0.22
351
± 0.12), however the difference was not statistically significant (p > 0.05). Table 5 presents
352
estimated means ± SE for all opponent pressure categories.
353
Game State and Scan Frequency
354
No statistical differences, nor substantial effect sizes were identified between
355
winning, drawing and losing. Table 5 displays the estimated means ± SE for all game state
356
categories.
357
Time in the Game and Scan Frequency
358
Table 5 presents the estimated means ± SE for all different time in the game
359
categories. Pairwise comparisons revealed players performed more scans between 0-15
360
minutes compared to 90-105 minutes (0.34 ± 0.02 vs 0.30 ± 0.03 scans/s; small ES: 0.26 ±
361
0.15) and between 16-30 minutes compared to 90-105 minutes (0.35 ± 0.03 vs 0.30 ± 0.03
362
scans/s; small ES: 0.31 ± 0.16), however the differences was not statistically significant (p >
363
0.05). Players also performed more scans between 45-60 minutes compared to 90-105
364
minutes (0.35 ± 0.02 vs 0.30 ± 0.03 scans/s; small ES: 0.31 ± 0.16) and between 61-75
365
minutes compared to 90-105 minutes (0.34 ± 0.02 vs 0.30 ± 0.03 scans/s; small ES: 0.28 ±
366
0.16), however the differences were not statistically significant (p > 0.05). All other pairwise
367
comparisons revealed trivial or unclear effect sizes.
368
369
**Insert Figure 4 near here**
370
371
16
Stage of competition and Scan Frequency
372
No statistical differences, nor substantial effect sizes were identified between the quarter-
373
final, semi-final and final. Table 5 displays the estimated means ± SE for all the three stages
374
of competition.
375
376
377
**Insert Table 5 near here**
378
379
Scan Frequency and Action Type
380
Results show scan frequency significantly predicted action type for dribble vs pass (β
381
= -3.68, z = -2.55, p = 0.011), with a higher scan frequency associated with a decrease in the
382
odds of choosing to dribble over pass. Higher scan frequencies were also associated with a
383
significant decrease in the odds of choosing to shoot over pass, (β = -2.36, z = -2.48, p =
384
0.013). No relationship was observed for receiving vs pass (β = -1.36, z = -1.81, p = 0.070).
385
The action type ‘pass’ was labelled as the reference category.
386
Scan Frequency and Action Result
387
A positive relationship was identified between scan frequency and action result
388
indicating for every one-unit increase in scan frequency, the log odds of the result of the
389
action with the ball being successful increase by 1.33, while keeping all other factors
390
constant. Players demonstrated a higher scan frequency when possession was maintained (M
391
= 0.36 ± 0.17 scans/s) compared to when possession was lost after their action with the ball
392
(M = 0.32 ± 0.18 scans/s). The mixed effects binomial logistic regression model was
393
17
statistically significant χ² (1) = 8.81, p = 0.003. An unsuccessful action (‘0’) was labelled as
394
the reference category.
395
Scan Frequency and Pass Direction
396
Significant relationships were found between scan frequency and the direction of a
397
pass. Results show scan frequency significantly predicted pass direction for forwards pass vs
398
backwards pass, (β = -0.85, z = -0.43, p = 0.049), with a higher scan frequency associated
399
with a decrease in the odds of performing a forward pass. No relationship was observed when
400
comparing sideways vs backwards passes (β = -1.07, z = -1.77, p = 0.077). Higher scan
401
frequencies were associated with a significant decrease in the likelihood of not performing a
402
pass compared to performing a backwards pass, (β = -2.59, z = -4.14, p < 0.001). The pass
403
direction ‘backwards’ was labelled as the reference category.
404
Scan Frequency and Pass Distance
405
Scan frequency was not found to have a significant effect on pass distance (β = -0.30,
406
SE = 0.36, p = 0.403).
407
Scan Frequency and Body Orientation
408
A relationship was identified between scan frequency and body orientation. Higher
409
scan frequencies were associated with a significant decrease in the likelihood of having a
410
forward body orientation compared to a backwards body orientation, (β = -1.63, z = -3.16, p =
411
0.002). No relationship was observed when comparing sideways vs backwards body
412
orientation (β = -0.84, z = -1.79, p = 0.073). A ‘backwards’ body orientation labelled as the
413
reference category.
414
Scan Frequency and Turn with the ball
415
18
A positive relationship was found between scan frequency and the probability of
416
turning with the ball indicating for every one-unit increase in scan frequency, the log odds of
417
performing a turn with the ball increase by 1.58, while keeping all other factors constant.
418
Players exhibited a higher scan frequency when performing a turn with the ball (M = 0.39 ±
419
0.18 scans/s) compared to when no turn with the ball was performed (M = 0.35 ± 0.17
420
scans/s). The mixed effects binomial logistic regression model was statistically significant χ²
421
(1) = 9.02, p = 0.003. No turn with the ball (‘0’) was labelled as the reference category.
422
Discussion
423
The purpose of this study was to describe the visual exploratory activity (VEA) of
424
elite female central midfield players and understand the relationships between VEA,
425
performance with the ball and specific contextual (“on or around the ball”) and situational
426
(“away from the ball”) factors. Results showed that pitch location was a significant predictor
427
of scan frequency where players performed a higher number of scans in central defensive
428
midfield pitch locations compared to defensive midfield wide and attacking central or wide
429
pitch locations. Additionally, scan frequency was found to be a significant predictor for a
430
number of performance with the ball variables. Higher scan frequencies resulted in an
431
increased likelihood of performing a successful action with the ball and performing a turn
432
with the ball. VEA appears linked to a player’s performance with the ball and seems to vary
433
depending on contextual demands (i.e. pitch location).
434
The first aim of this study was to describe VEA in elite central midfield players
435
across the knock-out stages of the UEFA Women’s EURO 2022. Players performed on
436
average 3-4 scans in the final 10 seconds before receiving the ball (scan frequency = 0.35
437
0.17 scans/s). This average is lower than that of similar studies investigating VEA in elite
438
male youth (0.42 scans/s in U17 and U19’s; Aksum et al., 2021) and professional male soccer
439
19
(0.44 scans/s; Jordet et al., 2020) players. These differences in findings could be linked to the
440
higher passing tempo in elite men’s soccer compared to elite women’s soccer (Mitrotasios et
441
al., 2022), which could in turn influence the frequency and timing of how a player needs to
442
scan their environment. For example, we might expect a relationship between a higher
443
passing tempo and an increase in scan frequency, as a result of players being required to have
444
a greater understanding of their environment due to the ball arriving at their feet quicker.
445
Similarly, with a slower passing tempo, we may expect lower scan frequencies due to players
446
potentially having more time and space to scan their environment and so may perform scans
447
of a longer duration that are less frequent. Aksum et al. (2021) found U19 male soccer
448
players conducted their final scans significantly closer to ball contact compared to U17
449
players. It was suggested that the increase in tempo demands of the U19 game may provide
450
an explanation for this finding (Aksum et al., 2021). Therefore, it could be suggested that a
451
slower passing tempo may lead players to scan their environment less frequently.
452
In line with previous work (e.g., Aksum et al., 2021), we measured scan frequency up
453
to ten seconds prior to receiving the ball. To understand the timing of a ‘final scan’ before
454
completing an action, we also measured the timing of scans in the final five seconds relative
455
to a player receiving the ball. Data showed that scan frequency was highest in the final 1-2
456
seconds prior to receiving the ball. A potential explanation for this finding is that players may
457
direct their attention away from the ball in the final two seconds before ball contact to receive
458
the most up to date information from the environment to subsequently inform their actions
459
with the ball (Aksum et al., 2021; McGuckian et al., 2018). Previous research in men’s soccer
460
has found players that perform more scans in the one and two seconds prior to receiving the
461
ball were more likely to turn with the ball (McGuckian et al., 2018). Therefore, in the context
462
of the current study, by performing scans closer to receiving the ball, central midfield players
463
may become more attuned to dynamically evolving game situations to enable them to make
464
20
the most appropriate action with the ball using the most relevant information (Aksum et al.,
465
2021). Rather than players simply increasing scan frequency, there appears a need to
466
understand where and when players should scan to inform coaching interventions.
467
In line with our hypotheses, results showed pitch location when receiving the ball to
468
be a significant predictor of scan frequency. The highest scan frequencies were observed in
469
defensive midfield central left and right locations, with findings also showing higher scan
470
frequencies observed in defensive midfield wide locations, compared to attacking wide
471
locations. These findings align with current literature on VEA in elite men’s youth (Aksum et
472
al., 2021) and adult male soccer (Jordet et al., 2020) which has found players scan more
473
frequently in central pitch locations compared to peripheral pitch locations. Central midfield
474
players are often required to drop deeper to collect the ball and so are required to have a
475
greater awareness of their surroundings due to also being surrounded by teammates (Aksum
476
et al., 2021; Jordet et al., 2020). More specifically, central midfield players may also be more
477
inclined to perform a greater number of scans in defensive midfield central pitch locations
478
due to the potentially detrimental consequences of losing possession (Jordet et al., 2020).
479
This current finding can be further explained by research that has found the more successful
480
women’s teams appear to be highly centralised and interconnected, with suggestions that
481
midfielders play a crucial role in performing a high volume of passes through central areas of
482
the pitch (de Jong et al., 2022). Taken together, it seems players when receiving the ball in
483
defensive central midfield positions scan their environment more frequently to identify
484
multiple passing options, with this pitch location being particularly important for progressing
485
play and starting attacks. Therefore, pitch location appears an important variable when
486
understanding VEA in elite women’s soccer.
487
Contrary to our hypotheses, no relationship was observed between opponent pressure
488
and scan frequency. Findings revealed players appeared to perform more scans when
489
21
experiencing less defensive pressure (i.e. when the distance to the nearest opponent was 7-9
490
metres away compared to 0-3 metres away), however only a small and non-significant effect
491
was identified (0.04; small ES: 0.22 ± 0.12). Research has found central locations of the pitch
492
tend to be highly congested, with playing spaces in the centre of the pitch observed to be
493
wider than they are deeper, with suggestions that successful possession may be more likely to
494
be maintained in wide, shallow areas of the pitch compared to central areas (Zubillaga et al.,
495
2013). Previous research investigating VEA in youth men’s soccer found significant
496
differences between scan frequency and opponent pressure, with higher amounts of opponent
497
pressure resulting in lower scan frequencies compared to when experiencing low amounts of
498
opponent pressure (Aksum et al., 2021; Pokolm et al., 2022). The disparities in findings could
499
be attributed to differences across age groups and playing positions being investigated (e.g.
500
elite youth male defenders, midfielders, and attackers vs elite women’s soccer central
501
midfield players). Future research should aim to further investigate the influence of opponent
502
pressure across different playing positions, as well as investigate a potential relationship
503
between pitch location and opponent pressure.
504
No relationships were found between situational variables (state of the game, time in
505
the game and stage of competition) and scan frequency. A possible explanation for this could
506
be attributed to all matches being played in a major senior international tournament with all
507
games being highly pressurised knock-out matches. Our data further suggests VEA is highly
508
individualised and unique to each player with regards to how a player visually explores their
509
environment. Previous research into visual search behaviours in men’s soccer has emphasised
510
the importance understanding individuals strengths and weaknesses relative to their own
511
action capabilities as this may constraint one’s ability to pick up the most important
512
information during visually guided behaviours (Button et al., 2011). Future research should
513
22
therefore consider analysing these factors not in isolation, but in the context of other variables
514
to understand the influence these factors have on how a player scans their environment.
515
Scan frequency was a significant predictor of both action result and turn with the ball.
516
Higher scan frequencies resulted in increased odds of performing a successful action with the
517
ball and turning with the ball. Applied to the context of the current study, if a player has an
518
enhanced understanding of their surroundings as a result of frequently exploring their
519
environment, they may be more likely to perform a turn with the ball in order to identify
520
potential empty space in an opposition’s defensive structure. Research investigating
521
possession tactics in UEFA Women’s Euro 2022 (O’Donoghue & Beckley, 2023) found the
522
most successful possessions were those of nine or more passes at a slower pass rate. These
523
findings highlight the importance of well-constructed build-up play where possession is
524
developed gradually resulting in more goal-scoring opportunities being created. In elite men’s
525
soccer, players scanned significantly more when possession was maintained (Jordet et al.,
526
2020) and a higher likelihood of turning with the ball was identified with a higher exploration
527
excursion and exploratory frequency (McGuckian et al., 2018). Therefore, with the current
528
study finding a significant relationship between scan frequency and action result (i.e. higher
529
scan frequencies resulting in players being more likely to maintain possession of the ball),
530
this could have important implications for elite women’s soccer.
531
Higher scan frequencies resulted in decreased odds of players choosing to dribble or
532
shoot compared to pass. This aligns with research into elite men’s soccer which has found a
533
higher likelihood of players performing a pass compared to a shot, dribble or receiving
534
(Jordet et al., 2020). Contradictory to our hypotheses, higher scan frequencies resulted in
535
decreased odds of players performing a forward pass compared to a backwards pass and
536
receiving the ball in a forward body orientation compared to a backwards orientation. This
537
accumulation of evidence contradicts that of previous research into elite youth soccer which
538
23
found higher scan frequencies have been associated with more forward passes compared to
539
backwards passes (Eldridge et al., 2013) and research into elite youth soccer identifying
540
higher scan frequencies resulted in more forwards and sideways body orientations compared
541
to backwards (Aksum et al., 2021). A potential explanation for the differences in findings
542
could be reflected by differences in developmental activities where literature has found elite
543
women’s soccer players may have spent less time in formalised training during early
544
adolescence (e.g. academies) and so may have a lower ‘training age’ compared to elite male
545
players (Ford et al., 2020). As a result, less time may have been spent developing specific
546
technical and perceptual-cognitive skills, such as decision making and visual search
547
(Pappalardo et al., 2021). Moreover, these contradictory findings can be further explained by
548
research into the technical and tactical demands of elite male and women’s soccer, which has
549
found possession is lost more frequently, and passes are performed with less accuracy in elite
550
women’s soccer compared to elite men’s soccer (Bradley et al., 2014; Garnica-Caparros &
551
Memmert, 2021). Therefore, these differences in the technical and tactical demands of the
552
game provide additional explanations for the differences in findings between the current
553
study and that found in previous research in elite men’s soccer. No relationship was identified
554
between scan frequency and pass distance, with a potential reason for this being the
555
combination of the study’s random effects structure nesting players within fixtures as well as
556
a considerably greater number of ball possessions analysed falling in the ‘0-14m’ category
557
compared to the ‘35m+’ category (565 v 52) resulting in this analysis potentially being
558
underpowered. Consolidating the above-mentioned findings, higher scan frequencies were
559
associated with a high likelihood of players performing a pass over a shot, dribble or
560
receiving as well as receiving the ball with a backwards body orientation and performing a
561
backwards pass. This collection of evidence provides an initial insight into the relationships
562
between VEA and a players performance with the ball.
563
24
Current findings can be interpreted through the lens of ecological psychology and
564
Gibson’s (1979) concept of affordances. Gibson’s (1979) ecological approach to visual
565
perception places an emphasis on the reciprocal nature of perception and action suggesting
566
how the pickup of information from the environment is as an active process which involves
567
the mobile body (see Fajen et al., 2009). Applied to the findings of the current study, if a
568
player scans their environment more frequently, they may be more likely to see a greater
569
number of opportunities for action (affordances), whilst having a better understanding of the
570
positions of teammates and opposition players. Research has suggested that a player can turn
571
their head frequently to perceive affordances in the playing environment, but their ability to
572
act upon this information remains grounded in their own action capabilities (Fajen et al.,
573
2009; Pocock et al., 2019). Therefore, from a theoretical perspective, our study reinforces the
574
coupling of perception and action, with players appearing to support performance by visually
575
exploring their environment immediately prior to engaging with the ball.
576
Based upon the study’s findings we propose some practical implications. Our results
577
revealed differences in VEA across pitch locations as well as VEA being related to a player’s
578
performance with the ball. It is recommended that coaches design practice activities where
579
central midfield players are exposed to a high volume of passes being received in central
580
defensive pitch locations with an emphasis on linking their visual exploratory activity to their
581
subsequent actions with the ball. Coaches should further strive to provide players with active
582
decision-making practices (e.g. small-sided or full sided game related practices) that involve
583
modifications placed upon the game (Eldridge et al., 2023). This may allow for players to be
584
exposed to frequently occurring in-game situations with sufficient contextual variation, for
585
example receiving the ball under varying amounts of opponent pressure with different body
586
orientations (Pokolm et al., 2022). Therefore, is strongly encouraged that practices are
587
designed to promote the coupling of perception and action whilst taking into consideration
588
25
the context and environment in which players visually explore their environment to support
589
skilled performance. It is also worth highlighting how current findings have been compared
590
and contrasted to that of VEA in elite and youth and men’s soccer. Research has highlighted
591
how professional women’s soccer has been required to adapt to the rules and regulations of
592
men’s soccer, with evidence suggesting soccer may be more demanding for female players
593
(Pedersen et al., 2019). Current findings shine a light on the challenges and difficulties of
594
comparing men’s and women’s soccer and so it is imperative to design practice environments
595
that are tailored specifically to women’s soccer. For example, based upon the current study’s
596
findings, coaches should aim to design practice activities that encourage central midfield
597
players to develop not just scan frequency, but also the timing of their scans relative to ball
598
contact. When receiving the ball in central locations of the pitch coaches are encouraged to
599
develop a player’s ability to scan their environment in the final seconds prior to ball contact
600
in order for players to identify the most up to date information from the environment.
601
The findings presented here should be considered in the context of some limitations.
602
Firstly, the data presented is from one international tournament investigating central midfield
603
players in isolation and so the results may not necessarily be representative of other
604
populations and leagues. Secondly, the number of individual ball possessions analysed is
605
relatively low (n = 1,038) in comparison to studies in men’s soccer which were conducted in
606
a similar vein (Pokolm et al., 2022; n = 5,338) and so findings must be interpreted and
607
applied with caution. Also, as a result of achieving a moderate agreement for the variable
608
‘lines broken’, this variable was not included in any statistical modelling. In future research,
609
one approach to improving the inter-rater reliability of the variable ‘lines broken’, is to
610
conduct further video familiarisation. For example, to achieve greater consistency, both
611
coders could be presented with numerous video examples of passes that broke and did not
612
break an oppositions line of defence, and the coders would then justify their decisions whilst
613
26
referring back to the operational definition. Whilst important to recognise these limitations,
614
we provide a number of recommendations for how future research can address these
615
limitations. Future investigations should aspire to further understand how both contextual and
616
situational factors influence a player’s VEA with a potentially fruitful avenue to explore VEA
617
in the context of positional differences whilst coupling this to performance with the ball.
618
Additionally, future work should investigate VEA in more in-situ and immersive
619
environments which may provide an insight into understanding the type of practice activities
620
that develop VEA. Recent advancements in technology have opened the door on participants
621
being able to be surrounded with representative match scenarios in a 360-degree setting (see
622
Honer et al., 2023; Musculus et al., 2021; Vater et al., 2019). Therefore, a logical next step
623
appears to be to apply the study’s findings in a more controlled setting by manipulating
624
variables of interest (e.g., pitch location, opponent pressure, action type). By manipulating
625
these variables in an immersive environment players could be presented with real life footage
626
from an 11v11 match and are required to visually explore their environment and make
627
decisions about their subsequent performance with the ball.
628
Conclusion
629
The primary objective of the study was to describe VEA in elite women’s soccer as
630
well as gain insight into the potential relationships that may exist between scan frequency and
631
contextual, situational, and performance with the ball factors. The study found a significant
632
relationship between pitch location and scan frequency as well as scan frequency being a
633
significant predictor of both action result and turn with the ball. More specifically, higher
634
scan frequencies were observed in central defensive midfield pitch locations, with players
635
also more likely to perform a turn with the ball and perform a successful action with the ball
636
(maintain possession) compared to an unsuccessful action (losing possession). When
637
designing representative practice environments, pitch location seems an important variable to
638
27
help further understand the contextual demands associated with how a player visually
639
explores their environment to guide subsequent actions with the ball. Therefore, the study has
640
established VEA is influenced by pitch location and related to performance with the ball in
641
elite women’s soccer. Future research is therefore required to extend and develop upon these
642
findings across different age groups (e.g. women’s youth soccer), playing positions (e.g.
643
defenders, wide players, and forwards) and skill levels (e.g. semi-professional) as well as
644
adopting more experimental research designs to further understand the influence of VEA on
645
performance with the ball.
646
647
648
Acknowledgements
649
We gratefully acknowledge the help of UEFA for their support in providing access to match footage
650
that was used for analysis. Also, a special thanks to our research assistants, Josh Power, and Ethan
651
Powell, for their valuable contributions to analysis.
652
Disclosure statement
653
Author ND is a current consultant for UEFA which enabled us access UEFA tactical footage. UEFA
654
had no involvement in the production of the manuscript.
655
Data availability statement
656
All relevant data is available online from:
657
https://osf.io/2cnh7/?view_only=613db7aaada144e1abfefb648a0b272f
658
Funding
659
There was no funding associated with this work.
660
Ethical approval
661
Ethical approval was obtained from the lead authors institution. Ethical approval number code:
662
2122_71
663
28
References
664
Aksum, K. M., Pokolm, M., Bjørndal, C. T., Rein, R., Memmert, D., & Jordet, G. (2021).
665
Scanning activity in elite youth soccer players. Journal of Sports Sciences, 39(21),
666
2401–2410. https://doi.org/10.1080/02640414.2021.1935115
667
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2014). Fitting linear mixed-effects models
668
using lme4. arXiv preprint arXiv:1406.5823.
669
Batterham, A. M., & Hopkins, W. G. (2006). Making meaningful inferences about
670
magnitudes. International Journal of Sports Physiology and Performance, 1(1), 50-57.
671
DOI: 10.1123/ijspp.1.1.50
672
Bradley, P. S., Dellal, A., Mohr, M., Castellano, J., & Wilkie, A. (2014). Gender differences
673
in match performance characteristics of soccer players competing in the UEFA
674
Champions League. Human Movement Science, 33(1), 159–171.
675
https://doi.org/10.1016/j.humov.2013.07.024
676
Bradley, P., O’Donoghue, P., Wooster, B., & Tordoff, P. (2007). The reliability of ProZone
677
MatchViewer: a video-based technical performance analysis system. International
678
Journal Of Performance Analysis In Sport, 7(3), 117-129.
679
https://doi.org/10.1080/24748668.2007.11868415
680
Brown, V. A. (2021). An introduction to linear mixed-effects modeling in R. Advances in
681
Methods and Practices in Psychological Science, 4(1), 1-19. DOI:
682
10.1177/2515245920960351
683
Button, C., Dicks, M., Haines, R., Barker, R., & Davids, K. (2011). Statistical modelling of
684
gaze behaviour as categorical time series: What you should watch to save soccer
685
29
penalties. Cognitive processing, 12, 235-244.
686
https://doi.org/10.1007/s10339-010-0384-6
687
Christensen, R. H. B. (2018). Cumulative link models for ordinal regression with the R
688
package ordinal. Submitted in J. Stat. Software, 35.
689
Cicchetti, D. (1994). Guidelines, criteria, and rules of thumb for evaluating normed and
690
standardized assessment instruments in psychology. Psychological Assessment, 6(4),
691
284-290. https://doi.org/10.1037/1040-3590.6.4.284
692
Datson, N., Hulton, A., Andersson, H., Lewis, T., Weston, M., Drust, B., & Gregson, W.
693
(2014). Applied Physiology of Female Soccer Players. Sports Medicine: Evaluations of
694
Research in Exercise Science and Sports Medicine, 44(9), 1225–1240.
695
https://doi.org/10.2165/00007256-199316030-00003
696
de Araújo, M. C., Baumgart, C., Jansen, C. T., Freiwald, J., & Hoppe, M. W. (2020). Sex
697
differences in physical capacities of German Bundesliga soccer players. The Journal of
698
Strength & Conditioning Research, 34(8), 2329-2337.
699
de Jong, L. M., Gastin, P. B., Angelova, M., Bruce, L., & Dwyer, D. B. (2020). Technical
700
determinants of success in professional women’s soccer: A wider range of variables
701
reveals new insights. PloS one, 15(10), 1-12.
702
de Jong, L. M. S., Gastin, P. B., Bruce, L., & Dwyer, D. B. (2022). Teamwork and
703
performance in professional women’s soccer: A network-based analysis. International
704
Journal of Sports Science & Coaching. https://doi.org/10.1177/17479541221092355
705
Eldridge, D., Pocock, C., Pulling, C., Kearney, P., & Dicks, M. (2023). Visual exploratory
706
activity and practice design: Perceptions of experienced coaches in professional football
707
academies. International Journal of Sports Science & Coaching, 18(2), 370-381.
708
30
https://doi.org/10.1177/17479541221122412
709
Eldridge, D., Pulling, C., & Robins, M. T. (2013). Visual exploratory activity and resultant
710
behavioural analysis of youth midfield soccer players. Journal of Human Sport and
711
Exercise, 8(3), 560-577.
712
Elff, M. (2021). Mclogit: Multinomial Logit Models, with or Without Random Effects or
713
Overdispersion. https://CRAN.R-project.org/package=mclogit.
714
Fajen, B. R., Riley, M. A., & Turvey, M. T. (2009). Information, affordances, and the control
715
of action in sport. International Journal of Sport Psychology, 40(1), 79-107.
716
Fernandes, T., Camerino, O., Garganta, J., Hileno, R., & Barreira, D. (2020). How do elite
717
soccer teams perform to ball recovery? Effects of tactical modelling and contextual
718
variables on the defensive patterns of play. Journal of Human Kinetics, 73(1), 165-179.
719
Ford, P. R., Hodges, N. J., Broadbent, D., O’Connor, D., Scott, D., Datson, N., ... &
720
Williams, A. M. (2020). The developmental and professional activities of female
721
international soccer players from five high-performing nations. Journal of Sports
722
Sciences, 38(11-12), 1432-1440. https://doi.org/10.1080/02640414.2020.1789384
723
Garnica-Caparrós, M., & Memmert, D. (2021). Understanding gender differences in
724
professional European soccer through machine learning interpretability and match
725
actions data. Scientific Reports, 11(1), 1–14.
726
https://doi.org/10.1038/s41598-021-90264-w
727
Gibson, J. J. (1979). The Ecological Approach to Visual Perception. Boston:
728
Houghton-Mifflin
729
Griffin, J., Larsen, B., Horan, S., Keogh, J., Dodd, K., Andreatta, M., & Minahan, C. (2020).
730
Women’s Soccer: An Examination of Factors That Influence Movement Patterns.
731
31
Journal of Strength and Conditioning Research, 34(8), 2384–2393.
732
https://doi.org/10.1519/JSC.0000000000003638
733
Harkness-Armstrong A, Till K, Datson N, Myhill N, Emmonds S (2022). A systematic
734
review of match-play characteristics in women’s soccer. PLoS ONE 17(6): e0268334.
735
https://doi.org/ 10.1371/journal.pone.0268334
736
Hopkins, W., Marshall, S., Batterham, A., & Hanin, J. (2009). Progressive statistics
737
for studies in sports medicine and exercise science. Medicine & Science in Sports &
738
Exercise, 41(1), 3-12. DOI: 10.1249/MSS.0b013e31818cb278
739
Ivn-Baragao, I., Maneiro, R., Losada, J.L. & Ard, A. (2022). Tactical Differences
740
Between Winning and Losing Teams in Elite Women's Football. Apunts Educacin
741
Fsica y Deportes, 147, 45-54. https://doi.org/10.5672/apunts.2014-
742
0983.es.(2022/1).147.05
743
Jordet, G., Aksum, K. M., Pedersen, D. N., Walvekar, A., Trivedi, A., McCall, A., Ivarsson,
744
A., & Priestley, D. (2020). Scanning, Contextual Factors, and Association With
745
Performance in English Premier League Soccerers: An Investigation Across a Season.
746
Frontiers in Psychology, 11, 1–16.
747
https://doi.org/10.3389/fpsyg.2020.553813
748
Jordet, G., Bloomfield, J., & Heijmerikx, J. (2013). The Hidden Foundation of Field Vision
749
in English Premier League (EPL) Soccer Players. MIT Sloan Sports Analytics
750
Conference, Boston.
751
Kubayi, A., & Larkin, P. (2020). Technical performance of soccer teams according to match
752
outcome at the 2019 FIFA Women’s World Cup. International Journal of Performance
753
Analysis in Sport, 20(5), 908–916. https://doi.org/10.1080/24748668.2020.1809320
754
32
Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical
755
data. biometrics, 33 (1), 159-174. https://doi.org/10.2307/2529310
756
Liu, H., Hopkins, W., Gmez, A., & Molinuevo, S. (2013). Inter-operator reliability of live
757
football match statistics from OPTA Sportsdata. International Journal Of Performance
758
Analysis In Sport, 13(3), 803-821. https://doi.org/10.1080/24748668.2013.11868690
759
Maneiro, R., Losada, J. L., Casal, C. A., & Ardá, A. (2020). The Influence of Match Status
760
on Ball Possession in High Performance Women’s Soccer. Frontiers in Psychology, 11,
761
1–11. https://doi.org/10.3389/fpsyg.2020.00487
762
Mara, J. K., Wheeler, K. W., & Lyons, K. (2012). Attacking strategies that lead to goal
763
scoring opportunities in high level women's football. International Journal of Sports
764
Science & Coaching, 7(3), 565-577. https://doi.org/10.1260/1747-9541.7.3.565
765
Martínez-Lagunas, V., Niessen, M., & Hartmann, U. (2014). Women’s soccer: Player
766
characteristics and demands of the game. Journal of Sport and Health Science, 3(4),
767
258–272. https://doi.org/10.1016/j.jshs.2014.10.001
768
McGuckian, T. B., Cole, M. H., Chalkley, D., Jordet, G., & Pepping, G. J. (2020).
769
Constraints on visual exploration of youth football players during 11v11 match-play:
770
The influence of playing role, pitch position and phase of play. Journal of Sports
771
Sciences, 38(6), 658-668. https://doi.org/10.1080/02640414.2020.1723375
772
McGuckian, T. B., Cole, M. H., Jordet, G., Chalkley, D., & Pepping, G. J. (2018). Don’t turn
773
blind! The relationship between exploration before ball possession and on-ball
774
performance in association soccer. Frontiers in Psychology, 9, 1–13.
775
https://doi.org/10.3389/fpsyg.2018.02520
776
Mitrotasios, M., Ródenas, J. G., Armatas, V., & Malavés, R. A. (2022). Creating goal scoring
777
33
opportunities in men and women UEFA Champions League soccer matches: Tactical
778
similarities and differences, 43, 154-161.
779
Musculus, L., Bäder, J., Sander, L., & Vogt, T. (2021). The influence of environmental
780
constraints in 360° videos on decision making in soccer. Journal of Sport and Exercise
781
Psychology, 43(5), 365-374. https://doi.org/10.1123/jsep.2020-0166
782
O’Donoghue, P., & Beckley, S. (2023). Possession tactics in the UEFA women’s EURO
783
2022 soccer tournament. International Journal of Performance Analysis in Sport, 23 (1),
784
1-17. https://doi.org/10.1080/24748668.2023.2206273
785
Okholm Kryger, K., Wang, A., Mehta, R., Impellizzeri, F. M., Massey, A., & McCall, A.
786
(2021). Research on women’s soccer: a scoping review. Science and Medicine in
787
Soccer, 1–10. https://doi.org/10.1080/24733938.2020.1868560
788
Pappalardo, L., Rossi, A., Natilli, M., & Cintia, P. (2021). Explaining the difference between
789
men’s and women’s soccer. PLoS One, 16(8), 1–17. e0255407
790
https://doi.org/10.1371/journal.pone.0255407
791
Pedersen, A. V., Aksdal, I. M., & Stalsberg, R. (2019). Scaling demands of soccer according
792
to anthropometric and physiological sex differences: a fairer comparison of men’s and
793
women’s soccer. Frontiers in psychology, 10, 1-11. doi: 10.3389/fpsyg.2019.00762
794
Phatak, A., & Gruber, M. (2019). Keep your head up—correlation between visual exploration
795
frequency, passing percentage, and turnover rate in elite soccer midfielders. Sports, 7(6),
796
1–9. https://doi.org/10.3390/sports7060139
797
Pocock, C., Dicks, M., Thelwell, R. C., Chapman, M., & Barker, J. B. (2019). Using an
798
imagery intervention to train visual exploratory activity in elite academy football
799
players. Journal of Applied Sport Psychology, 31(2), 218-234.
800
34
https://doi.org/10.1080/10413200.2017.1395929
801
Pokolm, M. (2021). Visual Scanning in Football. In Memmert, D (Ed), Match Analysis (pp.
802
87-94). Routledge.
803
Pokolm, M., Rein, R., Müller, D., Nopp, S., Kirchhain, M., Aksum, K. M., ... & Memmert,
804
D. (2022). Modeling Players’ Scanning Activity in Football. Journal of Sport and
805
Exercise Psychology, 44(4), 263-271. https://doi.org/10.1123/jsep.2020-0299
806
RStudio Team (2018). Rstudio: Integrated Development for R. RStudio, Inc., Boston, MA.
807
URL: http://www.rstudio.com/.
808
Soroka, A., & Bergier, J. (2010). Actions with the ball that determine the effectiveness of
809
play in women’s football. Journal of Human Kinetics, 26(1), 97-104.
810
UEFA (2022, August 31). Women’s Euro watched by over 365 million people globally.
811
https://www.uefa.com/insideuefa/news/0278-15ff73f066e1-c729b5099cbb-1000--
812
women-s-euro-watched-by-over-365-million-people-globally/.
813
Vater, C., Luginbhl, S., & Magnaguagno, L. (2019).Testing the functionality of peripheral
814
vision in a mixed-methods football field study, Journal of Sports Sciences, 37 (24),
815
2789-2797, DOI: 10.1080/02640414.2019.1664100
816
Vescovi, J. D., Fernandes, E., & Klas, A. (2021). Physical demands of women's soccer
817
matches: a perspective across the developmental spectrum. Frontiers in Sports and
818
Active Living, 3, 1-12.
819
Zubillaga, A., Gabbett, T. J., Fradua, L., Ruiz-Ruiz, C., Caro, Ó., & Ervilla, R. (2013).
820
Influence of ball position on playing space in Spanish elite women's football match-
821
play. International Journal of Sports Science & Coaching, 8(4), 713-722.
822
35
823
Figure 1. Hypothesised relationships between contextual (“on or around the ball”),
situational (“off the ball”), performance with the ball factors and scan frequency. Each
arrow represents a hypothesised relationship. Numbers denote references for each
example: (1) Aksum et al. (2021); (2) Jordet et al. (2020); (3) Fernandes et al. (2020); (4)
Eldridge et al. (2013); (5) Pokolm et al. (2022); (6) McGuckian et al. (2018).
824
825
826
827
828
829
830
831
832
833
834
36
Table 1. Instance Inclusion Criteria
835
Explanation
1
2
3
4
5
6
Visual explorations (scans) were measured in possession with all
head movements that occurred in the final 10 seconds prior to
receiving the ball from a teammate
If there was a turnover of possession in the final 10 seconds prior
to analysed player receiving the ball from a teammate, analysis
began the moment the analysed players team received control of
the ball. For example, if the opposition team had control of the ball
for 4 seconds and lost possession to the analysed players team, the
analysis started at 6 seconds (the moment possession was won) and
finished on the analysed players first touch of the ball.
In situations where the opposition team made contact with the ball,
however, did not have the ball under control (e.g., duelling for the
ball, clearing the ball or a deflected pass), it was deemed that the
analysed players team had not lost possession in our analysis.
For set pieces (e.g., corner kicks, throw-ins, and free kicks) the
10 second interval for analysing visual explorations was kept
consistent to enable the successful registering of scans. The
minimum of 5 seconds and maximum of 10 seconds interval was
maintained to ensure that there was consistency across all of
instances analysed. As a result of pilot testing, central midfield
players prior to the ball entering play from a set play (e.g., throw-
in) were scanning their environment for information prior to
receiving the ball. Therefore, by maintaining the 10 second
interval the aim was to minimise the chance of excluding
potentially important scans that may have informed the analysed
players subsequent action with the ball.
In instances where the analysed player received a pass from a
teammate, performed a pass to a teammate and then received the
ball again within the 10 second interval (e.g., combination play),
without the opposition gaining control of the ball, visual
explorations were analysed throughout the entire 10 second
interval and ended on the analysed players first touch of the ball.
The analysed players team were required to be in possession of the
ball for a minimum of 5 seconds prior to the analysed player
receiving the ball from a teammate (to provide sufficient time to
analyse a player’s VEA and ensure all analysed players are
performing VEA in attacking situations). Therefore, all instances
were a minimum of 5 seconds and a maximum of 10 seconds in
duration due to the analysed players team often gaining possession
37
of the ball and the analysed player then receiving a pass from a
teammate between 5-10 seconds of possession being won.
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
38
Table 2. Intra-class corelations for number of scans (continuous variable).
865
Inter-rater
Intra-rater
Variable
ICC (95% CI)
p
Strength of
Agreement
ICC (95% CI)
p
Strength of
Agreement
Number
of scans
0.899
(0.861-0.926)
<0.001
Excellent
0.912
(0.801-0.953)
<0.001
Excellent
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
39
Table 3. Cohen’s k for all categorical variables.
886
Inter-rater
Intra-rater
Variable
Kappa (κ)
p
Strength of
Agreement
Kappa (κ)
p
Strength of
Agreement
Action Type
0.871
<0.001
Almost Perfect
0.939
<0.001
Almost Perfect
Pitch Location
0.935
<0.001
Almost Perfect
0.897
<0.001
Almost Perfect
Lines Broken
0.594
<0.001
Moderate
0.842
<0.001
Almost Perfect
Action Result
0.925
<0.001
Almost Perfect
0.980
<0.001
Almost Perfect
Pass Distance
0.855
<0.001
Almost Perfect
0.928
<0.001
Almost Perfect
Pass Direction
0.906
<0.001
Almost Perfect
0.933
<0.001
Almost Perfect
Turn with the ball
Opponent Pressure
Body Orientation
0.810
0.818
0.870
<0.001
<0.001
<0.001
Almost Perfect
Almost Perfect
Almost Perfect
0.846
0.826
0.900
<0.001
<0.001
<0.001
Almost Perfect
Almost Perfect
Almost Perfect
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
40
0.00
0.10
0.20
0.30
0.40
0.50
Ball Contact -
1 second 1.01 - 2
seconds 2.01 - 3
seconds 3.01 - 4
seconds 4.01 - 5
seconds
Scan Frequency (scans/second)
Seconds prior to ball contact
905
Figure 2. Mean (± SE) scan frequency (scans/s) during the final five seconds prior to ball
contact.
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
41
Table 4. Mean Scan frequency and number of instances for all analysed
variables.
Scan Frequency (scans/s)
Number of Instances
Variable (s)
M (SD)
n
Contextual Factors
Opponent Pressure
0-3 m
4-6 m
7-9 m
10-32 m
Situational Factors
Game State
Winning
Drawing
0.34 (0.17)
0.37 (0.18)
0.40 (0.18)
0.39 (0.27)
0.35 (0.15)
0.36 (0.17)
645
295
82
16
159
693
Losing
Time in the game
0-15 min
16-30 min
31-45 min
45-60 min
61-75 min
76-90 min
90-105 min
105-120 min
Stage of competition
Quarter Final
Semi Final
Final
Performance with the ball
Action Type
Pass
Shot
Dribble
Receiving
Action Result
Successful
Unsuccessful
Pass Direction
Backwards
Sideways
Forwards
No Pass
Pass Distance
0-14 m
15-34 m
35 m +
0.31 (0.17)
0.37 (0.15)
0.37 (0.17)
0.35 (0.17)
0.37 (0.18)
0.35 (0.17)
0.33 (0.20)
0.31 (0.16)
0.29 (0.15)
0.36 (0.17)
0.35 (0.17)
0.34 (0.18)
0.36 (0.17)
0.29 (0.16)
0.26 (0.18)
0.32 (0.18)
0.36 (0.17)
0.32 (0.18)
0.38 (0.16)
0.35 (0.18)
0.35 (0.17)
0.30 (0.17)
0.35 (0.17)
0.37 (0.17)
0.38 (0.19)
186
184
138
142
162
154
141
68
49
616
277
145
904
44
20
70
776
262
326
135
443
134
565
287
52
42
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
No Pass
Body Orientation
Backwards
Sideways
Forwards
Turn with the ball
Turn with the ball
No turn with the ball
0.30 (0.17)
0.38 (0.17)
0.33 (0.17)
0.35 (0.18)
0.39 (0.18)
0.35 (0.17)
134
243
477
318
148
890
43
956
957
958
959
960
961
962
963
964
965
966
967
968
969
Figure 3. Mean scan frequency (scans/second) presented in 12 pitch locations (attacking
direction from left to right) with standard deviation values and the number of instances (n).
Note. Only pitch location zones with a minimum of 5 instances included in the figure. In all
defensive zones n < 5.
44
970
Figure 4. Effect Sizes for differences in estimated mean and statistical significance for pitch location.
Statistical difference (p < 0.05*, p < 0.001**).
45
Table 5. Estimated means ± SE for Scan Frequency for all contextual (“on or around the
ball”) and situational (“off the ball”) factors.
Scan Frequency (scans/s)
Variable (s)
Estimated Means (SE)
Contextual Factors
Pitch Location
Defensive Right (DR)
Defensive Centre Left (DCL)
Defensive Left (DL)
Defensive Midfield Right (DMR)
Defensive Midfield Centre Right (DMCR)
Defensive Midfield Centre Left (DMCL)
Defensive Midfield Left (DML)
Attacking Midfield Right (AMR)
Attacking Midfield Centre Right (AMCR)
Attacking Midfield Centre Left (AMCL)
Attacking Midfield Left (AML)
Attacking Right (AR)
Attacking Centre Right (ACR)
Attacking Centre Left (ACL)
Attacking Left (AL)
Opponent Pressure
0-3 m
4-6 m
7-9 m
10-32 m
Situational Factors
Game State
Winning
Drawing
Losing
Time in the game
0-15 min
16-30 min
31-45 min
45-60 min
61-75 min
76-90 min
90-105 min
105-120 min
Stage of Competition
Quarter Final
Semi Final
Final
0.26 (0.17)
0.41 (0.12)
0.26 (0.12)
0.39 (0.02)
0.39 (0.02)
0.39 (0.02)
0.37 (0.03)
0.34 (0.02)
0.35 (0.02)
0.31 (0.02)
0.35 (0.02)
0.27 (0.04)
0.27 (0.04)
0.26 (0.04)
0.31 (0.04)
0.31 (0.02)
0.33 (0.02)
0.35 (0.03)
0.34 (0.04)
0.33 (0.03)
0.34 (0.02)
0.31 (0.02)
0.34 (0.02)
0.35 (0.03)
0.33 (0.02)
0.35 (0.02)
0.34 (0.02)
0.32 (0.02)
0.30 (0.03)
0.32 (0.03)
0.34 (0.02)
0.33 (0.02)
0.32 (0.03)
46
