Key words
match analysis - soccer - football
Introduction
The main duty of soccer referees is to enforce the laws of the game and sanction any
infringements of these laws [1]. Decision-making is therefore an essential component of a soccer referee’s match
performance, because each decision may have a direct or indirect impact on the match
result. A referee makes about 140 observable decisions during a soccer match [2] and calls 25–41 foul play infringements [2]
[3]
[4]. Expert panels assessing the infringements retrospectively by video analysis find
that the proportion of incorrect decisions ranges from 14–36% [4]
[5]
[6]. The referees’ level of expertise influences offside decision-making skills and
recall accuracy [4]
[5]
[7]
[8]. In addition, stake size, attendance crowd, and the crowd’s distance to the field
may also affect the referees’ decision-making processes [9].
In addition to decision-making skills, a high level of physical fitness is required
for referees to be able to keep up with the play and get an unobstructed view of potential
infringements. During top-level national and international matches, the field referees
usually cover 10–12 km, with high-intensity running (HIR) accounting for 600–1 200 m
[10]
[11]
[12]. The referees must position themselves based on where infringements might occur
while maintaining an overview of the players and the play, but without interfering
with the players or ball. Thus, the field referees cannot be too close or too far
away from the infringements. The distance from the referee to the infringement may
be related to decision-making accuracy [13] and the lowest proportion of decision-making errors is achieved when the fouls are
called from between 11 and 15 m [4]. If the referee becomes fatigued during the match, the distance to infringements
may increase because the referee cannot keep up with the play [11]. Also, if a referee realises he is positioned too far from potential infringements,
he is likely to move faster for a better view rather than simply adjust his position
if he is closer to the action. To our knowledge, however, the association between
position and running speed in relation to match period has not been assessed in soccer
referees.
The relationship between decision-making and physical exertion has been investigated
in football players [14] as well as assistant referees [15]
[16]. Catteeuw et al. [16] found no evidence that fatigue affects decision-making in assistant referees, because
the offside-decision accuracy remained consistent throughout the match [16]. In contrast, Oudejans et al. [15] found that assistant referees made more offside-decision errors when running compared
to when they were standing still, walking, or jogging. More recently, a study investigating
Australian football field officials concluded that 300 m running time did not affect
performance in a video-based decision-making test performed immediately after running
[17]. However, higher relative running speed immediately prior to a decision increased
the amount of decision-making errors made by Australian football field officials during
match play [18].
The majority of studies investigating the relationship between fatigue and decision-making
in sport officials are performed on assistant football referees and on Australian
football field officials. Thus, how factors related to physical performance influence
decision-making in field referees is unknown. Therefore, the primary aim of the present
study is to assess if running performance prior to a called foul or positioning affects
decision-making accuracy in soccer field referees. Secondly, we aim to assess if the
referees’ movements, positioning, and time of the infringements are associated.
Materials and Methods
Participants
Eleven male field referees weighing 74.9±6.5 kg, 182.5±7.3 cm tall, and aged 35.4±7.0
years with 9.8±5.2 years of experience as top level referees in Norway participated
in the study. Data were collected during all 15 top league matches (four referees
officiated two matches) on one arena during the 2016 season. The Norwegian Data Protection
Official for Research (NSD) approved the study and it meets the ethical standards
of the International Journal of Sports Medicine [19]. All referees were informed orally and in writing of the procedures and provided
their written consent before they were included in the study.
Video analysis
Videos of the matches were produced for commercial broadcasting by Norwegian TV2 (Bergen,
Norway) and accessed via the online publishing tool, Mediabanken (www.mediabank.me).
A qualified referee viewed video footage of all 15 matches and created separate MP4
video files of 10–20 s’ duration of all foul play infringements (including replay
when available) where the referee awarded a free kick, penalty, yellow or red card.
An expert panel consisting of two active referees (one national and former FIFA international,
and one regional) independently assessed each infringement. Fifty-five infringements
were reported where at least one of the expert referees classified the decision as
“not entirely correct” or “difficult to assess”. Subsequently, the two experts met
to reach a conclusion regarding the 55 infringements. The infringements were classified
according to the following pre-defined categories: 1) correct decision; 2) correctly
awarded penalty/free kick, but a card should also been awarded; 3) correctly awarded
penalty/free kick, but the card should not have been awarded; 4) incorrect decision,
no foul play was committed; 5) incorrect decision, the other team committed foul play;
6) unknown, due to poor video quality; 7) unknown, due to the complexity in the situation.
The expert panel could replay each infringement as many times as necessary. When decision-making
accuracy was analysed, categories 1 through 3 were combined into “right decisions”
and compared to the combined “incorrect decision” categories 4 and 5. To increase
the number of cases, a secondary analysis compared category 1 to the “somewhat incorrect”
categories 2 through 5.
Movement analysis
The movements were captured by a fully automatic tracking system based on radio waves
(ZXY Technology Ecosystem, Chyronhego, Trondheim, Norway) previously described in
detail [20]. In brief, the referees’ positions were monitored by transponders, placed in lightweight
belts worn around the hip, which sent radio waves (20 Hz) to four calibrated, stationary
sensors, each mounted in one of the light masts surrounding the football field (ZXY
RadioEyeTM Positioning Sensor, Chyronhego). By integrating information from the four
receivers in an advanced vector-based process, the referees’ positions and movements
were determined. The data were compressed and filtrated by a Linux server using Ubuntu
14.04 and stored in a structured query language (SQL) database before being transferred
to a software program (Microsoft Excel, 2013). The test-retest reliability of the
system had been assessed in a previous study reporting an interclass correlation (ICC)
coefficient of 1.0, 0.999, and 0.999 (P=0.001) for x- and y- and total distance [21]. All matches were monitored live and controlled by two researchers using the ZXY
Sport Tracking software program.
Distance covered by the referees was classified into total distance (TD), running
distance (RD), and HIR distance in accordance with previous, similar studies [21]. Running and HIR were defined as movement at a speed>14.4 and≤19.8 km·h−1 and>19.8 km· h−1, respectively. The TD, RD, and HIR distances accumulated during the 10, 30, and 60 s
prior to the infringements were defined as TD10, TD30, and TD60; RD10, RD30, and RD60;
and HIR10, HIR30, and HIR60.
Combined analysis
The home team players wore identical transponders as the referees, enabling measurement
of the distance between the referee and the infringements using a semiautomatic feature
in ZXY Sport Tracking. At the time of the infringement, the home team player involved
in the infringement and the referee were identified and the software computed the
distance between them. This distance was defined as the referee’s distance to the
infringement. This distance was classified into different categories (0–5, 6–10, 11–15,
16–20, 21–25, and>25 m) similar to previous studies [4]
[13]. Simultaneously with the distances, the running speed of the referee was recorded
with ZXY Sport Tracking. The infringements were classified according to the area where
they were awarded and in accordance with the recommendations of the FIFA Referee Department
and a previous study [4]. Lateral areas were defined as the areas between the sideline and an imaginary line
from the intersection between the sideline and the middle line and the middle point
of the goalmouth, representing the influence zone of each assistant referee. The remaining
area of the field was defined as the central area.
Statistical analysis
Normal distribution was visually inspected, and normally distributed data (referee’s
distance to infringement, running speed of the referee at the time of the infringement,
TD30, TD60) are presented as mean plus standard deviation (SD) or 95% confidence interval
(95% CI) when appropriate. Not normally distributed data (TD10, RD10, RD30, RD60,
HIR10, HIR30, HIR60) are presented as median with interquartile range (IQR). The associations
between binary and continuous variables were analysed using binary logistic regression
with decision-making accuracy or field area as the dependent variables. The independent
variables were the running speed of the referee at the time of the infringement, TD30,
distance from the field referee to the foul play infringement, and match time of the
infringement. The correlations between normally distributed variables were assessed
by the Pearson product-moment correlation (r), and Spearman’s rho test was used if
one or more variables were not normally distributed. The difference between infringements
in the central and lateral zone was analysed by a one-sample binomial test. To investigate
the effect of match period on the referee’s distance to the infringement and running
speed of the referee at the time of the infringement, we used a one-way analysis of
variance (ANOVA). The distribution of infringements over the match periods was assessed
by one-sample chi-square test.
Results
The present study consists of 347 called foul play infringements (15 matches) with
a mean of 23.1 (±3.6) infringements per match. The referees made the correct decision
in 98% of all assessed fouls ([Table 1]).
Table 1 Video-based classification of the correctness of the decision was penalties or free
kicks awarded.
|
n
|
%
|
|
Correct decision
|
328
|
94.5
|
|
Correct decision , but a card should have also been awarded
|
6
|
1.7
|
|
Correct decision , but the card should not have been awarded
|
3
|
0.9
|
|
Incorrect decision, no foul play was committed
|
5
|
1.4
|
|
Incorrect decision, the other team committed foul play
|
1
|
0.3
|
|
Unknown, due to poor video
|
3
|
0.9
|
|
Unknown, due to the complexity of the situation
|
1
|
0.3
|
Associations between movement/position and decision-making
There were no associations between the correctness of called fouls and running speed
at the time of the infringement (β=0.19, p=0.09); total distance in the last 30 s
(β=0.01, p=0.55); distance to the infringement (β=0.02, p=0.82); time of infringement
(β<0.00, p=0.92); or field zone of infringement (chi square=1.17 p=0.28). None of
the incorrect decisions were made for infringements in lateral zones, and 2.1% of
the calls made on infringements in the central zone were incorrect. The same analyses
were performed including situations with incorrectly awarded cards in the incorrect
decision category, and the results were similar (p>0.08). [Figure 1a–d] is stratified by decision correctness and displays running speed at the time of
the infringement, total distance in the last 30 s, distance from the field referee
to the foul play infringement, and time of infringement.
Fig. 1 The field referee's; speed of movement when he calls fouls a, distance from the infringements when he calls fouls b, total distance accumulated during the 30 seconds prior to the called foul c, the match time of the infringements d stratified by correctness of the decision for all infringements in one area over
the entire season. Data are presented as mean with 95% confidence intervals.
Associations between movement, position, and time of infringement
The running speed of the referee and the referees’ distance to the infringement were
weakly correlated (r=0.31, p<0.001). The correlation between distance covered prior
to the infringement and running speed at the time of the infringement or the referee’s
distance from the infringement is presented in [Table 2].
Table 2 Correlation (Spearman rho) between referee running speed at the time of the infringement
or the referee’s distance from the infringement, and distance covered prior to the
infringement assessed over a period of 10, 30, and 60 s.
|
|
TD10s
|
TD30s
|
TD60s
|
RD10s
|
RD30s
|
RD60s
|
HIR10s
|
HIR30s
|
HIR60s
|
|
Speed at the time of infringement
|
Rho
|
0.159
|
0.135
|
0.096
|
0.097
|
0.092
|
0.044
|
−0.044
|
−0.011
|
−0.011
|
|
p
|
0.004
|
0.017
|
0.088
|
0.086
|
0.103
|
0.432
|
0.432
|
0.849
|
0.847
|
|
Distance from infringement
|
Rho
|
0.472
|
0.308
|
0.286
|
0.436
|
0.268
|
0.247
|
0.211
|
0.114
|
0.102
|
|
p
|
<0.0001
|
<0.0001
|
<0.0001
|
<0.0001
|
<0.0001
|
<0.0001
|
<0.0001
|
0.039
|
0.067
|
TD, total distance; RD, running distance (accumulated distance at speeds>14.4 and
≤19.8 km · h−1); HIR, high-intensity running (accumulated distance at speeds>19.8 km · h−1).
Eighty-four percent of the infringements (292) were committed in the central area
and 16% (55) in the lateral areas (p<0.001). The distances between the referee and
the infringements were 13.4 m (95% CI 12.8–14.0) and 21.9 m (95% CI 20.4–23.4) in
the central and lateral areas, respectively (β=0.77, p<0.001).
The number of infringements was unevenly distributed (p<0.001) in the different distance
categories, with most infringements being called from a distance between 10 and 14.9
m. There were no differences in number of infringements (p=0.94), distance from infringement
(p=0.16), referee running speed (p=0.78), or total distance in the last 30 s (p=0.10)
between the 15-minute match periods.
Discussion
Neither the field referees’ movements prior to infringements, running speed at the
time of the infringements, nor distance to the infringement affected decision-making
accuracy. However, there was no statistically significant tendency towards more decision-making
errors in the central area of the field. The distance between the referee and the
called infringements was also greater in the lateral areas of the field. The running
speed of the referee increased with increasing distance to the infringement, and the
running speed at the time of the infringement was also positively associated with
the distance covered prior to the infringement.
In the present study, only six free kicks/penalties (1.7%) were awarded to players
not performing foul play. This is substantially lower than what was previously reported
from international matches (14.2%) [4] and national matches in New Zealand (19.7%) [5]. Even when we included incorrectly awarded yellow or red cards and failure to award
cards when warranted, the error percentage remained very low (4%). The Norwegian top
referees were practically always right when they called a foul play. The explanation
for this low error rate remains unclear. In the present study and other studies investigating
decision-making accuracy in soccer referees [4]
[5], situations in which the referees called fouls were analysed. From a comparable
decision-making accuracy study in Australian football officials, unwarranted sanctioned
foul play was shown in 3.5% of the decisions and missed foul play in 9.5% of the decisions
[18]. Based on the results from Australian football, one might speculate that the proportion
of incorrect decisions would increase if situations not sanctioned had been included
in the present findings. In the best leagues [22] and at international matches [23], the locomotion is greater, and the crowd size and stake size are larger. These
factors may all contribute to decision-making errors [9] and possibly explain the low proportion of incorrect decisions in the present study
compared to the other studies also investigating called fouls only [4]
[5]. Even though the error percentage was low in the present study, the number of infringements
per match was comparable to previous studies [4]
[5] or lower [2]
[3].
Associations between movement/position and decision-making
The present study suggests that neither the distance covered during the 30 s prior
to the infringement nor running speed at the time of the infringement affects decision-making
accuracy. This confirms the findings from previous studies on soccer field referees
during national matches in New Zealand [5] and Australian football field officials [18], but it contradicts findings in offside judgements by assistant soccer referees
[15]. Offside judgments may be more sensitive to the movements of the referee compared
to free-kick judgments, because the assistant referees must focus on the position
of the attacking and defending players in relation to each other when the ball is
played. An assistant referee seeks to be in line with the last defender, who may be
moving at high speed. In contrast to the assistant referees, a field referee does
not have to be in line with the last defender and may run where he chooses. This enables
the field referee to focus on the acts committed by the players and on the interpretation
of the laws of the game. The field referee may also face forward, whereas the assistant
referee must often shuttle sideways. Thus, free-kick decision-making by the field
referee may be less sensitive to referee movements compared to offside decisions made
by the assistant referee [18].
The distance between the referee and the infringements did not affect decision-making
accuracy. This supports the findings from international soccer matches [4] and national Australian football matches [18]. This indicates that referees in general are able to maintain an appropriate distance
to the play. In the present study, the mean distance was slightly shorter (14.8 m)
compared to the study by Mallo et al. [4] (16.7 m). Furthermore, Krustrup et al. [11] reported the distance to the infringements to be 12 m in the central zone and 16
m in the attacking zones of the field. Fixed situations such as throw-ins and free
kicks where the attacking team plays the ball towards players deep into the attacking
zone often result in physical contact between many players within a small space. In
these situations foul play may occur, but it is demanding for the referee because
the view may be blocked by the concentration of players on a small part of the field.
Thus the probability of incorrect decisions may increase in such situations even if
the referee does not have to move rapidly ahead of the situation or have time to choose
a position ahead of the throw-in/free-kick. Therefore, fixed offensive situations
may confound a possible relationship between running speed or distance to infringements
and decision-making accuracy.
The larger distance between the referees and the infringement in the lateral areas
compared to the central areas of the field confirms the results from previous studies
[3]
[4]. In the present study, there was no difference in decision-making errors between
infringements in the central and lateral areas. This contradicts the findings of Mallo
et al. [4], who reported a larger proportion of incorrect decisions for infringements in the
lateral areas of the field, even though Mallo et al. [4] report distances to the infringements in lateral and central areas of the field
similar to the present study. It seems plausible that the field referees make fewer
incorrect calls in the lateral areas, where the assistant referees are supposed to
assist the field referees in calling foul plays and are able to view potential infringements
from another perspective. The lack of association between decision-making accuracy
and any of our possible proxies for fatigue (total distance prior to infringement,
speed at infringement, time of infringement, distance to infringement) indicates that
Norwegian top referees possess the necessary fitness to keep up with the play and
not call unwarranted fouls.
Associations between movement, position and time of infringement
The running speed of the referee was positively related to the distance to the infringements.
This may indicate that the referee anticipates possible infringements and seeks to
be closer to the situation by increasing running speed when the distances are large.
In the present study and the study by Elsworthy et al. [18], most infringements are called when the distance to the infringement is between
11 and 15 m, a distance previously reported to be the distance with the lowest proportion
of decision-making errors [4]. However, in contrast to the present study, Mallo et al. [4] report that most incidents are called from a distance of more than 25 m.
The number of infringements, distance from the infringement, running speed at the
time of the infringement, and total distance in the last 30 s before the infringement
were evenly distributed between the 15-minute periods. This may indicate that the
referees are able to maintain their running and decision-making performance throughout
the match. This builds on the evidence from a previous study reporting consistent
amounts of high-intensity running and accelerations in Norwegian soccer referees throughout
the matches [20]. Our findings are also partly in agreement with two studies by Krustrup et al. [11]
[24] which found increased distance between the referee and infringements in the attacking
zones at the end of the matches, whereas the distance remained stable in the central
areas of the field.
Strengths and limitations
The present study is strengthened by high-quality data from radio-based tracking of
the referees and infringements filmed by high-quality video produced for commercial
broadcast. A relatively large sample size and the assessments of the decisions by
external referees, a method which previously demonstrated good validity within referees
[4], between referees [3]
[4] and between referees and coaches [25], also strengthen the study. The major limitation of the study is low statistical
power for detecting differences between incorrect and correct decisions due to the
low number of incorrect decisions. However, post hoc analysis including incorrect
card decisions in the incorrect decision category and between the 55 situations where
at least one expert referee initially classified the decision as not entirely correct
or difficult to assess against the other situations did not find any difference between
the groups either. The assistant referees are supposed to assist the field referees’
decision-making, especially in the lateral areas of the field, and lack of data from
the communication between the referees can be regarded as a limitation when assessing
factors related to decision-making errors. Another weakness is that we assessed only
situations where the referee sanctioned foul play, whereas potential situations where
the referee failed to sanction foul play were not included in the present study. The
occurrence of such decision-making errors is unknown, yet in New Zealand soccer it
has been reported to be as high as 41% of all decision-making errors [5]. However, the proportion of unwarranted calls of foul play was almost 10 times greater
in the latter study compared to the present study. Nevertheless, field referees are
reported to make about 140 observable decisions during a soccer match [2], thus the decisions investigated in the present study make up only a small proportion
of the decisions the referees made during the matches. The referees in the present
study seldom called unwarranted fouls and the amount of missed fouls was not investigated.
Conclusion
Based on the present study, it seems that the Norwegian field referees have the necessary
physical fitness and positional ability to maintain their decision-making ability
and to keep up with play throughout the match. The Norwegian referees are practically
always correct when they call a foul. Future research within this field should focus
on infringements of the laws of the game that are not sanctioned by the referees and
factors related to these “missed” infringements.
