May the best-sighted win? The relationship between visual function and performance in Para judo

Abstract

This study aimed to establish the optimal cut-off point(s) between classes in a new, evidence-based classification system for VI judo. We collected results from international VI judo competitions held between 2012 and 2018. Data on visual acuity (VA) and visual field (VF) measured during classification were obtained. Performance was determined by calculating a win ratio for each athlete. VA was significantly associated with judo performance (r = −.33, p <.001), VF was not (r =.30, p =.15). Decision tree analysis suggested to split the data into two groups with a VA cut-off of 2.5 logMAR units. Stability assessment using bootstrap sampling suggested a split into two groups, but showed considerable variability in the cut-off point between 2.0 and 3.5 logMAR. We conclude that to minimise the impact of impairment on the outcome of competition, VI judo should be split into two sport classes to separate partially sighted from functionally blind athletes. To establish an exact cut-off point and to decide if other measures of visual function need to be included, we argue for continued research efforts together with careful evaluation of research results from a multidisciplinary perspective.

Full text

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rjsp20
Journal of Sports Sciences
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rjsp20
May the best-sighted win? The relationship
between visual function and performance in Para
judo
Kai Krabben , Evgeny Mashkovskiy , H. J. C. (Rianne) Ravensbergen & David L.
Mann
To cite this article: Kai Krabben , Evgeny Mashkovskiy , H. J. C. (Rianne) Ravensbergen &
David L. Mann (2020): May the best-sighted win? The relationship between visual function and
performance in Para judo, Journal of Sports Sciences
To link to this article: https://doi.org/10.1080/02640414.2020.1851899
© 2020 The Author(s). Published by Informa
UK Limited, trading as Taylor & Francis
Group.
Published online: 27 Nov 2020.
Submit your article to this journal
View related articles
View Crossmark data

May the best-sighted win? The relationship between visual function and
performance in Para judo
Kai Krabben
a
, Evgeny Mashkovskiy
b
, H. J. C. (Rianne) Ravensbergen
a
and David L. Mann
a
a
Department of Human Movement Sciences, Faculty of Behaviour and Movement Sciences, Vrije Universiteit Amsterdam, Amsterdam Movement
Sciences, Amsterdam, The Netherlands;
b
Department of Labour and Social Protection of Population of Moscow, Moscow, Russia
ABSTRACT
This study aimed to establish the optimal cut-o point(s) between classes in a new, evidence-based
classication system for VI judo. We collected results from international VI judo competitions held
between 2012 and 2018. Data on visual acuity (VA) and visual eld (VF) measured during classication
were obtained. Performance was determined by calculating a win ratio for each athlete. VA was
signicantly associated with judo performance (r = −.33, p <.001), VF was not (r =.30, p =.15). Decision
tree analysis suggested to split the data into two groups with a VA cut-o of 2.5 logMAR units. Stability
assessment using bootstrap sampling suggested a split into two groups, but showed considerable
variability in the cut-o point between 2.0 and 3.5 logMAR. We conclude that to minimise the impact
of impairment on the outcome of competition, VI judo should be split into two sport classes to separate
partially sighted from functionally blind athletes. To establish an exact cut-o point and to decide if other
measures of visual function need to be included, we argue for continued research eorts together with
careful evaluation of research results from a multidisciplinary perspective.
ARTICLE HISTORY
Accepted 12 November 2020
KEYWORDS
Paralympic sports; judo;
vision impairment; evidence-
based classification; decision
tree analysis
Introduction
In contact sports such as judo or wrestling, kinaesthetic informa-
tion is considered equally or more important than vision (and
other senses), making those sports particularly suitable for indi-
viduals with vision impairment (Kuznetsova & Barabanshchikova,
2006; Starosta, 2013). In Paralympic judo, athletes with dierent
degrees of vision impairment (ranging from partially sighted to
fully blind) even compete against each other within the same
competitive class, whereas in most Paralympic sports, athletes
are allocated to dierent sport classes to compete against others
with a similar activity limitation (Mann & Ravensbergen, 2018;
Tweedy & Vanlandewijck, 2011). The rules of judo for athletes
with vision impairment (VI judo) are adapted to better accom-
modate athletes with low vision. Rather than starting a match
a few metres apart, as is done in able-sighted judo, a bout in VI
judo starts after the two combatants have taken a standardised
two-handed grip on each other’s jacket. This adaptation is
designed to make the sport more suitable for those with vision
impairment by removing what is presumably the most visually
demanding aspect of a judo bout, i.e. obtaining an appropriate
grip on the opponent (Piras et al., 2014).
Despite the adaptation to the VI judo rules, it remains ques-
tionable whether judo athletes (judokas) with dierent degrees
of vision impairment can compete equitably against each other
(Jones & Howe, 2005). Experts within VI judo (coaches, athletes,
administrators) expressed the opinion that blind judokas do
not stand an equal chance of winning when ghting against
partially sighted opponents (Krabben et al., 2019). This expert
opinion is in agreement with statistical analyses of results in
elite-level VI judo, showing blind judokas are less successful
than partially sighted judokas. Even though VI judo eectively
only holds one competitive class for all eligible athletes, ath-
letes are allocated to one of three sport classes based on an
assessment of their visual function (Table 1). This class alloca-
tion has allowed for comparisons of competitive success
between athletes of dierent sport classes, showing function-
ally blind (class B1) athletes perform worse than their partially
sighted (class B2/B3) opponents (Kons et al., 2019; Krabben
et al., 2018; Mashkovskiy et al., 2019). Yet VI judokas with
most residual vision (class B3) do not win more often when
ghting opponents with less residual vision (class B2)
(Mashkovskiy et al., 2019). These ndings suggest that as long
as both judokas have some residual functional vision (i.e. they
are not blind or limited to their ability to see only light or
rudimentary motion), the better-sighted athlete does not hold
an advantage over the other on the basis of their vision alone.
To further examine the impact of vision on VI judo perfor-
mance, Krabben et al. (2018) experimentally compared able-
sighted judokas ghting with and without blindfolds under VI
judo rules. Although blindfolded athletes were still able to
maintain reasonable levels of performance, they were signi-
cantly disadvantaged when ghting against non-blindfolded
opponents. These ndings seem to indicate that the current VI
judo regulations do not full the aim of Paralympic classica-
tion, which is to “minimise the impact of impairment on the
outcome of competition” (Tweedy & Vanlandewijck, 2011).
Although expert consensus as well as empirical evidence
suggest a need to change the way visually impaired judokas
are grouped for competition, additional information is
CONTACT Kai Krabben k.j.krabben@vu.nl Department of Human Movement Sciences, Faculty of Behaviour and Movement Sciences, Vrije Universiteit
Amsterdam, Amsterdam Movement Sciences, Van der Boechorststraat 9, 1081BT Amsterdam, The Netherlands
JOURNAL OF SPORTS SCIENCES
https://doi.org/10.1080/02640414.2020.1851899
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.

needed to establish new and more legitimate classication
criteria. Based on the abovementioned comparisons
between current classes, it may seem straightforward to
divide competition into two separate classes, one for the
current class B1 athletes, and another for the current class
B2 and B3 athletes. Yet these analyses do not account for
the potential impact of dierences in visual function
between athletes within the same sport class; two athletes
may currently fall into the same competitive class, but hold
a dierent degree of impairment which might impact their
performance dierently. For instance, even within the B1
class, athletes dier in their degree of visual function:
some may still perceive light or even hand motion, whereas
others have complete loss of vision. It remains possible that
those with some remaining vision retain an advantage over
those who are completely blind. Similarly, it is possible that
athletes with severely limited vision who are now classied
in the B2 class could be at a disadvantage when ghting
against B2 athletes with better vision, and so may need to
be allocated to a sport class separate to those other B2
athletes (i.e. in their own class or joined with the current
B1 judokas). To overcome the limitations of comparisons
between current classes, the International Paralympic
Committee (IPC) mandated that new classication criteria
should be based on evidence relating sport performance
to direct measures of impairment (Tweedy et al., 2016).
New classication criteria for VI judo should thus be based
on research relating direct measures of visual function to
measures of judo performance. To date this has not yet
occurred.
A direct assessment of the visual function of all VI athletes is
obtained during classication, a process that aims to determine
the eligibility of athletes to compete in Para sports and to allocate
eligible athletes to sport classes (Mann & Ravensbergen, 2018). VI
classication is performed by certied classiers who have a back-
ground in optometry or ophthalmology (International Blind
Sports Federation, 2018). Most VI athletes are classied on the
basis of their impaired visual acuity (VA), which is a measure of
their sharpness of vision. Alternatively, some athletes may be
classied eligible to compete on the basis of visual eld (VF)
loss. VF is a measure of the area of peripheral vision with which
an individual can see (i.e. without moving their eyes). The data on
VA and VF of judokas obtained during classication hold promise
for a more direct analysis of the impact of vision impairment on
judo performance, yet previous studies did not have access to this
information and were therefore constricted to comparisons
between sport classes.
Another issue not addressed within earlier work is
whether the impact of vision impairment diers across the
dierent gender and weight categories within VI judo. VI
judo has seven weight classes for men (−60 kg, −66 kg,
−73 kg, −81 kg, −90 kg, −100 kg and +100 kg) and six for
women (−48 kg, −52 kg, −57 kg, −63 kg, −70 kg and
+70 kg). Experts within the VI judo community speculated
that the visual demands of judo may be higher for light-
weight compared to heavyweight athletes because of the
faster nature of the bouts, and therefore vision impairment
might impact performance more within the lighter weight
categories (Krabben et al., 2019). If the impact of impair-
ment on performance would indeed dier between gender
and weight categories, this might warrant the development
of gender and weight-specic classication criteria for VI
judo.
The aim of this study was to establish the optimal cut-o
point(s) between sport classes in a new, evidence-based system
of classication for VI judo. We examined the relationship
between direct measures of visual function collected during clas-
sication and the results of international VI judo competitions
between 2012 and 2018. For the purposes of this study, the
International Blind Sports Federation (IBSA) allowed us to access
classication data stored in their database, the IBSA Sports
Administration System (ISAS). Based on expert opinions and ear-
lier empirical work, we hypothesised that (1) functionally blind
judokas would be less successful than those with some residual
functional vision, and (2) the impact of impairment on perfor-
mance would be lower in the heavier compared to the lighter
weight categories. We expected to nd a single cut-o point
between two ideal classes separating athletes with and without
functional vision.
Method
Data sample
We included all available competition results from Paralympic
Games, World Championships, World Cups and Continental
Championships held between 2012 and 2018.
1
Most of these
data were collected from the IBSA website.
2
Additional
archived competition results were provided to us by IBSA.
Only competitions for which the full results were available
were included. Data on the visual function of athletes were
collected from ISAS. IBSA granted permission to access this
information for the current study. On the advice of our institu-
tional ethical committee, no ethical approval was required for
this study because all data were obtained in secondary form
and anonymised before being further analysed.
Table 1. Current classification criteria for visually impaired (judo) athletes.
Athletes can be classified based on tests of visual acuity, a measure of the
sharpness or clarity of vision, or visual field, a measure of the area of peripheral
vision with which an individual can see (i.e. without moving their eyes). Only one
of the two criteria (visual acuity or visual field) needs to be met in order to be
allocated to a sport class.
Class
Visual
Acuity
(LogMAR) Visual Field (radius) Description
B3 1.0 to 1.4 Less than 20
degrees
Limited visual acuity and/or visual
field in both eyes.
B2 1.5 to 2.6 Less than 5 degrees Severely limited visual acuity and/or
visual field in both eyes.
B1 Poorer
than 2.6
Cannot be B1 with
only loss of visual
field
An athlete can distinguish only light
from dark, or is not able to
perceive light.
1
As classification procedures for VI sports underwent significant changes in the lead up to the Paralympic Games of London 2012, only data obtained from 2012
onwards was included.
2
http://www.ibsasport.org/sports/judo/results/
2K. KRABBEN ET AL.

Procedure
For each of the athletes who competed in one or more of
the included competitions, we collected all classication
data between 2012 and 2018 through ISAS. VA is measured
during classication using the Berkeley Rudimentary Vision
Test (BRVT; Bailey et al., 2012). The BRVT was developed
specically to measure visual function in those with (very)
poor vision. VA is measured in logMAR units, with higher VA
values representing poorer vision. Young adults without
vision impairment are expected to have a VA of approxi-
mately 0.0 logMAR units. The current minimum impairment
criteria (MIC) for VA to be eligible to compete in VI sports is
set at 1.0 logMAR. With the BRVT, VA can be measured up
to 2.9 logMAR. For athletes with VA worse than 2.9 logMAR,
VA can be classied as either white eld discrimination
(WFD), black white discrimination (BWD), light perception
(LP) or no light perception (NLP) (Bailey et al., 2012). For
analysis purposes, a numeric value was assigned to these
acuity levels: WFD = 3.2 logMAR, BWD = 3.5 logMAR,
LP = 3.7 logMAR, NLP = 4.0 logMAR. The values assigned
to WFD and BWD are suggested by the BRVT. They are
written on the card pairs that are used to perform the test
and calculated from the size of the presented target when
observing these targets at the designated testing distance
of 25 cm. The value of 4.0 logMAR for NLP corresponds with
previous studies (Aaberg et al., 1998; Allen et al., 2019;
Jacobs et al., 2011) and represents the logMAR equivalent
to an object that subtends 180 degrees of visual eld (i.e.
essentially the whole visual eld). We therefore considered
4.0 logMAR to be a theoretical equivalent of complete
vision loss.
3
A value of 3.7 logMAR was assigned to LP
because this value falls in between the values assigned to
WFP and NLP and like other studies, retains a 0.3 logMAR
dierence between LP and NLP (Ikeda & Kishi, 2010; Jackson
et al., 2016; Moshfeghi et al., 2003). VF is measured in
degrees radius and is assessed during classication using
specic visual eld machines. The Goldmann Visual Field
Perimeter is preferred, the Humphrey Field Analyser or
Octopus Interzeag are also accepted. For descriptions of
and comparisons between these machines we refer to
Bevers et al. (2019). All measures of visual function data
were collected as recorded by the classiers on the ocial
classication sheets stored in ISAS.
Some VI athletes have progressive medical conditions caus-
ing their visual function to change over time. Therefore, most
athletes who undergo classication need to be re-evaluated to
establish whether their visual function has changed. Based on
the athlete’s condition, classiers may decide that athletes
need to be reclassied within either the next one, two or four
years. In case the classier believes an athlete’s condition is
highly unlikely to progress, the athlete may obtain a conrmed
status; they do not need to be re-evaluated in classication (this
is mainly for athletes with severe vision loss, i.e. LP or NLP).
Data analysis
In case multiple classications of the same athlete were avail-
able, we selected the classication under which the athlete
competed most for further analysis, to ensure independence
of observations. In any cases where an athlete turned 18
between 2012 and 2018, we included in our analyses the data
only from those competitions where the athlete was 18 years or
older at the time of the competition. Performance was assessed
using a win ratio, calculated as the number of ghts won
divided by the total number of ghts the athlete competed in
within the analysed time frame (i.e. when the selected classi-
cation was valid). Most international VI judo competitions are
organised using (dierent variations of) a knock-out system,
which means losing competitors are eliminated as the compe-
tition progresses. Within an elimination system, the average
win ratio across all competitors is therefore expected to be
lower than 50%.
The relationship between visual function and perfor-
mance was analysed through calculation of Pearson’s corre-
lation coecient. Decision tree analyses were used to
determine whether or not the data supported splitting VI
judo into more than one sport class, and if so, what the
ideal cut-o point(s) between these classes should be. We
applied the unbiased recursive partitioning algorithm
(Hothorn et al., 2006), which recursively aims to perform
univariate splits in the input variables as long as these are
signicantly associated with the response variable. The
results of recursive partitioning are known to be potentially
unstable, as small changes in the data sample may lead to
substantially dierent decision trees being built (Strobl
et al., 2009). To assess the stability of the decision tree,
we examined the variability in cut-o selection by boot-
strapping of 10,000 random resamples of our data, using
the toolkit for stability assessment of tree-based learners
(Philipp et al., 2016). 10,000 samples were randomly drawn
with replacement from the original dataset and had the
same size as the original dataset. For each of these 10,000
samples, a separate decision tree was built. We summarised
the number of splits and the values of the split points over
all decision trees to estimate the optimal number of classes
and cut-o point(s).
To assess whether the impact of impairment on perfor-
mance diered across weight categories, we repeated the ana-
lyses described above when grouping the three lightest and
the three heaviest weight categories for men and women (we
did not analyse each weight category separately because of low
numbers of athlete in each category, which would have ren-
dered the analyses underpowered). For each group, we calcu-
lated the Pearson’s correlation coecient between visual
function and win ratio. Additionally, we also used decision
tree analyses to assess whether the data for each gender and
weight group supported splitting competition into dierent
sport classes.
3
logMAR stands for the logarithm of the minimum angle of resolution, i.e. the minimum visual angle at which a person can resolve details, measured in minutes of arc
(1 minute of arc is 1/60
th
degree). A logMAR score of 4.0 corresponds to a visual field of 10800 minutes of arc or 180 degrees. Any value above 4.0 logMAR would refer
to a minimal angle of resolution larger than a full visual field.
JOURNAL OF SPORTS SCIENCES 3

Results
Judo performance
Match data for eighteen dierent tournaments were col-
lected (Table 2), spanning 3101 individual ghts. A total of
617 dierent ISAS-registered athletes competed in one or
more of these ghts. On average athletes competed in
3.7 ± 2.9 of the included tournaments, where they obtained
an average win ratio of 0.34 ± 0.29. A boxplot showed that
athletes who competed more often performed better
(Figure 1). Athletes who only competed once or twice
were particularly less likely to be competitive compared to
athletes who competed in at least three competitions (t
(570.75) = 10.7, p< .001, d= 0.9). Including these athletes
in our analysis might cloud the true impairment-
performance relationship, because these athletes apparently
performed worse not necessarily as a result of their degree
of impairment. Therefore we excluded 282 athletes who
only competed once or twice, leaving 335 athletes for
further analyses.
Visual function
For 39 out of the 335 athletes, no classication data were
available. For 294 out of the 296 remaining athletes (99.3%),
a measure of VA could be retrieved from the collected classi-
cation data. For only 25 athletes (8.4%), VF was measured
during classication. VA was bimodally distributed (Figure 2
(a)), with most athletes having either a VA between 1.0 and
2.5 logMAR (79.1% of all athletes), or worse than 3.5 logMAR
(12.8%; these were athletes with either LP or NLP). Only 5.1% all
athletes had a VA between 2.6 and 3.5 logMAR. Seven athletes
(2.4%) had a VA better (i.e. lower) than the current MIC of 1.0
logMAR; these athletes all qualied to compete on the basis of
an impaired VF.
Relationship between visual acuity and judo performance
We found a signicant, moderate correlation between VA and
win ratio (r= −.33, p< .001; Figure 2(b)), indicating that athletes
with better vision won a higher percentage of their ghts. The
decision tree algorithm found a binary split in the data at a VA
of 2.5 logMAR (Figure 3). Athletes with a VA of 2.5 logMAR or
better (n= 243), had an average win ratio of 0.51 ± 0.25 (M± SD,
95%CI [0.49,0.52]). Athletes with vision worse than 2.5 logMAR
(n= 53) had an average win ratio of 0.25 ± 0.24 (M± SD, 95%CI
[0.22,0.29]). The dierence between these two groups was
signicant and showed a large eect size (t(78.9) = 6.8,
p< .001, d= 1.0). No additional split was made by the algorithm,
suggesting the win ratio of athletes could not be better pre-
dicted by further splitting the data on the basis of VA. Analyses
of the impairment-performance relationships after splitting the
data at a VA of 2.5 logMAR, clearly showed visual function is not
related to performance within the rst subgroup of athletes
with VA of 2.5 logMAR or better (r= .04, p= .49). Yet within the
subgroup of athletes with vision worse than 2.5 logMAR,
a trend for a negative relationship between impairment and
performance remained present (r= −.20, p= .15).
Bootstrapping results conrmed that VA could be used as
a variable to split the data into groups with dierent perfor-
mance levels. For each of the 10,000 bootstrap samples, the
decision tree split the data into multiple groups on the basis of
VA. The majority of the bootstrap trees (65.8%) stopped after
Table 2. Tournaments included within the dataset.
Competition Place Start Date
2012 Paralympic Games London, United
Kingdom
30-8-2012
2013 European Championships Eger, Hungary 4-12-2013
2014 World Championships Colorado Springs, USA 4-9-2014
2014 Asia Games Incheon, South Korea 20-10-2014
2015 World Cup Eger Eger, Hungary 20-2-2015
2015 World Championships Seoul, South Korea 13-5-2015
2015 Parapan American Games Toronto, Canada 12-8-2015
2015 European Championships Odivelas, Portugal 27-11-2015
2016 Paralympic Games Rio de Janeiro, Brazil 8-9-2016
2017 Asian and Oceanian
Championships
Tashkent, Uzbekistan 4-8-2017
2017 European Championships Walsall, United
Kingdom
26-8-2017
2017 American Judo Championships Sao Paulo, Brazil 23-8-2017
2017 World Cup Tashkent Tashkent, Uzbekistan 9-10-2017
2018 World Cup Antalya Antalya, Turkey 22-4-2018
2018 Pan-American Championships Calgary, Canada 21-5-2018
2018 World Cup Atyrau Atyrau, Kazakhstan 6-9-2018
2018 Asia Para Games Jakarta, Indonesia 8-10-2018
2018 World Championships Odivelas, Portugal 16-11-2018
Figure 1. Box plot of win ratio by number of tournaments an athlete competed in. Athletes who competed more often in international competitions performed better.
A t-test confirmed that athletes who only competed in one or two competitions performed worse than the others (t(570.75) = 10.7, p<.001, d= 0.9).
4K. KRABBEN ET AL.

a single split. For 25.9% of the bootstrap samples, a second split
on VA was made and 8.3% of the trees made even three or
more splits on the basis of VA. The histogram of the rst cut-o
points selected for VA shows a large spread, with 96.5% of the
rst splits made between 2.0 and 3.5 logMAR (Figure 4). This
indicates that the normal variability in data which can be
expected from sampling might cause the selected value for
the split point to vary substantially within this range. The
most frequently selected cut-o point was 2.5 logMAR (17.5%
of the 10,000 cases). The next most frequently selected VA cut-
o points were 3.5 logMAR (17.1%), 2.9 logMAR (13.0%) and 2.0
logMAR (11.7%).
Relationship between visual eld and judo performance
We found no signicant correlation between VF and win ratio
(r= .30, p= .15; Figure 5). No split could be made in the data on
the basis of VF. This might indicate VF is not related to perfor-
mance in VI judo, yet with only 25 athletes included we should
acknowledge this analysis is likely to be underpowered.
Impact of impairment across weight categories
Similar trends in the relationship between impairment and
performance were observed across the dierent gender and
weight groups (Figure 6). A signicant negative correlation was
found between VA and win ratio in all groups (men light
weights: r= −.29, p= .003; men heavy weights: r= −.24,
p= .021; women light weights: r= −.43, p = .001; women
heavy weights: r= −.47, p= .001). Decision tree analyses sug-
gested to split competition into two classes in all four groups,
with the suggested cut-o points ranging from 2.3 to 3.2
logMAR.
Discussion
The aim of this study was to establish the optimal cut-o point-
(s) between sport classes in a new, evidence-based classica-
tion system for VI judo. We collected seven years of
international competition results and related these data to
the visual function of the athletes measured during classica-
tion. We did not achieve our aim of nding a specic optimal
cut-o point. Still, the results add weight to earlier ndings
showing the most severely impaired athletes in VI judo are
less successful than their better-sighted opponents (Kons
et al., 2019; Krabben et al., 2018; Mashkovskiy et al., 2019). Yet
where earlier work compared existing groups based on current
classes, in the current study we directly related visual function
to performance. This approach allowed us to provide more rm
support for the development of new, evidence-based classes
for VI judo and to help further establish the cut-o between
classes (Tweedy et al., 2016).
Figure 2. Visualisations of impairment and performance data. A) Distribution of the analysed athletes’ visual acuity. B) Relationship between visual acuity and win ratio.
C) Same data as Figure B presented as a boxplot where each box represents an equal number of athletes.
JOURNAL OF SPORTS SCIENCES 5

Figure 4. Histogram of the first split points for VA using 10,000 bootstrapped samples. Each of the 10,000 samples was randomly drawn with replacement from the
original dataset and for each sample a separate decision tree was built. All of these 10,000 decision trees split the data into multiple groups. 2.5 logMAR was most
frequently selected as the first split point (in 17.5% of all cases); 96.5% of all cases selected a first split point between 2.0 and 3.5 logMAR.
Figure 3. Relationship between visual acuity and win ratio. The vertical black line represents the suggested cut-off point. The red line represents the linear fit between
visual function and performance over all athletes. The black dotted lines represent the linear fits between visual function and performance for each of the two
subgroups created by the split.
Figure 5. Relationship between visual field and win ratio. The red line represents the linear fit between visual function and performance.
6K. KRABBEN ET AL.

We found that the current one-class system in VI judo does
not full the Paralympic aim to “minimise the impact of impair-
ment on the outcome of competition” (Tweedy &
Vanlandewijck, 2011). Decision tree analysis suggested to split
competition into two classes with substantially dierent per-
formance levels, one class for athletes with VA up to 2.5 logMAR
and another class for athletes with VA worse than 2.5 logMAR. It
is worth noting that this cut-o point is close to that currently
used to distinguish those with most severe impairment from all
other athletes in most VI sports (2.6 logMAR, see Table 1). The
current ndings support the expert opinion that competition
should be split into separate classes for functionally blind and
partially sighted athletes. However, we did not nd support for
the development of gender or weight specic classication
criteria hypothesised by VI judo experts (Krabben et al., 2019).
It might seem counterintuitive to suggest a split into multi-
ple sport classes when the overall correlation between VA and
win ratio was only weak. Yet a strong correlation between
impairment and performance is not necessary to justify a split
in the data. Actually, for a two-class system, rather than a strong
correlation, we should ideally nd two clusters in the data: one
with better performance for those with less impairment, and
another with worse performance with more impairment. This
pattern is largely consistent with what we nd in our data
(Figure 3). Indeed the overall correlation between impairment
and performance was only weak, largely because there is no
association between vision and judo performance in the VA
range from 1.0 to 2.5 logMAR. Crucially, we interpret this to be
a good outcome for the purposes of classication because it
means that that group should compete fairly. If there were to
be a strong correlation within that group then it would be
necessary to split the group into further classes. In other
words, a strong correlation between impairment and perfor-
mance will in all likelihood indicate the need for many classes.
The antithesis is not true: a weak correlation doesn’t mean that
no split is required.
Although the bootstrap analysis could not identify one spe-
cic optimal cut-o point, it further supported the conclusion
to split VI judo into multiple sport classes. Stability assessment
of the main decision tree over 10,000 bootstrap samples sug-
gested in all cases to split competition into more than one sport
class on the basis of VA. Yet the range of split points varied
substantially between 2.0 up to 3.5 logMAR, indicating that the
normal variability in the data which can be expected from
sampling might cause the selected value for the split point to
vary substantially within this range. These ndings supported
the need for a split into (at least) two sport classes, to ensure
the most severely impaired athletes (those with only LP or NLP)
do not compete against those with relatively better visual
function (better than 2.0 logMAR). Each split point between
2.0 and 3.5 logMAR will achieve this aim and would therefore
be a considerable improvement compared to the current one-
class system. Yet the results were inconclusive whether any
split point between 2.0 and 3.5 logMAR is more optimal than
any other.
Analyses of the impairment-performance relationships after
applying a split in the data at 2.5 logMAR showed a trend for
impairment to impact performance for athletes with vision
worse than 2.5 logMAR (Figure 3). Even when applying a split
into two classes, additional measures might still be needed to
Figure 6. Impairment-performance relationships for separate gender and weight groups. Red lines represents the linear fit between visual function and performance
over all athletes within the group. Blue vertical lines show suggested cut-off values from decision tree analyses. The blue dotted lines represent the linear fits between
visual function and performance for each of the two subgroups created by the split.
JOURNAL OF SPORTS SCIENCES 7

equalise chances for athletes in a new class meant for the most
severely impaired athletes. One option might be to blindfold all
competitors in this class. Blindfolding is generally not consid-
ered appropriate in VI sports, although experts in VI sports
expressed there might be some situations where the use of
blindfolds is appropriate (Ravensbergen et al., 2016). In VI
swimming for instance, blackened swimming goggles are con-
sidered appropriate only for athletes within the most severe VI
class (Ravensbergen et al., 2018).
We did not achieve our aim to establish specic criteria for
new sport classes for VI judo. Instead of a specic cut-o point,
we could only identify a range of VA values. Several challenges
exist in further narrowing down to an exact cut-o point within
the suggested range of VA values. First of all, VA values within
this range were underrepresented in our data sample, with VA
values between 2.5 and 3.5 logMAR especially rare (Figure 2(a)).
It may be that these VA values are generally uncommon
amongst the visually impaired population. Yet it might also
be that classiers currently do not accurately measure VA
when it is worse than 2.6 logMAR, which is the current cut-o
point for the B1 class. Classiers often have limited time and
might therefore opt to simply classify an athlete with vision
worse than 2.6 logMAR as having either LP or NLP rather than
continuing to test with the BRVT.
4
The implication is that the
substantially lower performance levels we have found for the
group of athletes with only LP or NLP might actually concern
a broader group of athletes with VA worse than 2.6 logMAR. If
so, more accurate assessment of the visual function of these
athletes would have likely reduced the range of possible VA
cut-o values we have found, but it does not alter the funda-
mental conclusion that the group with VA worse than 2.5
logMAR performs more poorly than the group with VA 2.5
logMAR or better. Nonetheless, we recommend careful and
accurate assessment of VA in those with severe vision impair-
ment in combination with continued monitoring of their per-
formance in future VI judo competitions.
A second challenge in establishing the cut-o point is the
large degree of variability in performance across the whole
range of VA values in the athletes included in our study. Even
though the most severely impaired athletes performed on
average signicantly worse than those with some residual
vision, many (functionally) blind judokas achieved high levels
of performance, and many partially sighted athletes did not
perform well at all. This is in agreement with earlier work
showing even without vision, judokas are still capable of throw-
ing sighted opponents when ghting under VI judo rules
(Krabben et al., 2018). A considerable component of judo per-
formance thus seems determined by factors other than vision,
presumably factors related to talent and training. The joint IPC-
IBSA position stand on VI classication therefore encourages
researchers to collect background information on confounding
factors such as practice volume to control for these factors
while establishing the impairment-performance relationship
(Mann & Ravensbergen, 2018). Moreover, classication cur-
rently only considers tests of VA and VF. Yet other aspects of
vision (e.g. the ability to perceive contrast or motion) might be
more strongly associated with judo performance and would
therefore be more suitable for use in classication. Indeed,
a panel of VI judo experts identied six additional measures
of visual function besides VA and VF which might be important
enough for VI judo to be included in classication (motion
perception, dynamic visual acuity, light sensitivity, ocular coor-
dination, depth perception, and contrast sensitivity; Krabben
et al., 2019). Future research should aim to evaluate whether
the inclusion of those additional tests of visual function would
increase the amount of variability in performance explained by
VI, in which case those tests could be included in classication.
Alternately, it is possible that at least one of those measures
could be a better predictor of performance in judo. In VI shoot-
ing for example, the evaluation of additional tests of visual
function led to the recommendation to include a test of con-
trast sensitivity in classication (Allen et al., 2018). Finally, we
should acknowledge that our analysis of the relationship
between VF and performance was limited. VF was assessed
during classication in only 8.4% of all athletes in our data
sample, and just 2.7% qualied to compete in VI judo on the
basis of an impaired VF. This makes it challenging to establish
evidence-based classication criteria for VF on the basis of
existing data. Considering that a panel of VI judo experts unan-
imously agreed that VF should remain included in classication
for VI judo (Krabben et al., 2019), additional research eorts into
the relationship between VF and judo performance would
seem warranted.
Yet even with additional research results available, it
remains plausible that research data may at best provide
a range of values rather than an exact cut-o point to distin-
guish sport classes. It may be that there is a more gradual
decrease in judo performance in the range of visual acuity
from 2.0 to 3.5 logMAR, rather than judo performance decreas-
ing drastically at a single level of impairment to visual acuity. If
so, decisions on the establishment of new classication criteria
would need to be taken on other than purely data-driven
grounds. One consideration might be the risk of setting the cut-
o too high (i.e. at a too severe level of impairment within the
range provided by research) or too low (i.e. too mild). On the
one hand, athletes with a level of visual function within the
provided range might still benet from their limited vision,
meaning they would hold an unfair advantage when being
allocated to a class with functionally blind athletes. Yet their
level of visual function might also not be sucient anymore to
support their performance, in which case they would be dis-
advantaged when being allocated to a class with better-sighted
athletes. Depending on which of these scenarios would be
judged more harmful to the legitimacy of VI judo competition,
the cut-o point between classes might be set at a more or less
severe level of impairment. Alternatively, the cut-o might be
set at a more conceptual border between “partially sighted”
and “blind”. One option might be at the current cut-o for the
B1 class of 2.6 logMAR, which is based on the World Health
Organisation’s denition for blindness (World Health
4
On some of the classification sheets we accessed during data collection for this study, we even found VA values reported as “>2.6” rather than specific values. These
classifications were excluded from the research data as we could not extract an exact VA value, but it suggests that indeed classifiers might be more concerned with
determining the correct sport class of an athlete than the exact level of VA.
8K. KRABBEN ET AL.

Organization, 2004). Another option might be 2.9 logMAR,
which is the highest numeric VA value measurable by the
BRVT (Bailey et al., 2012). The decision where to “draw the
line” on the basis of research data might therefore not be
straightforward and/or objectively possible. To bridge the gap
between research ndings and the establishment of new clas-
sication criteria, experts from other elds might need to be
consulted such as philosophers or legal experts. As argued by
McNamee (2017, p. 207):
“[W]hat would be required is a classication system and committee
that found space for philosophers and social scientists, not merely
scientic and clinically trained evaluators of structure and function
for classication eligibility issues.”
Besides continued research eorts, careful evaluation of research
results from a multidisciplinary perspective may prove critical to
establish the most legitimate way to structure VI judo competition.
Conclusion
Vision impairment is signicantly associated with performance
when applying the current classication system for VI judo. This
means judokas with less vision impairment have a competitive
advantage over those with more severe impairment. Results of
the current study suggest to split VI judo into two sport classes
on the basis of VA, with a suggested cut-o point between 2.0
and 3.5 logMAR. To further narrow down to an exact cut-o
point and to decide if other measures of visual function need to
be included, we argue for continued research eorts together
with careful evaluation of research results and philosophical
considerations from a multidisciplinary perspective.Notes
Disclosure statement
No potential conict of interest was reported by the authors.
References
Aaberg, T. M., Flynn, H. W., Schiman, J., & Newton, J. (1998). Nosocomial
acute-onset postoperative endophthalmitis survey: A 10-year review of
incidence and outcomes. Ophthalmology, 105(6), 1004–1010. https://doi.
org/10.1016/S0161-6420(98)96000-6
Allen, P. M., Latham, K., Ravensbergen, R. H. J. C., Myint, J., & Mann, D. L. (2019).
Rie shooting for athletes with vision impairment: does one class t all?
Frontiers in Psychology, 10(1727). https://doi.org/10.3389/fpsyg.2019.01727
Allen, P. M., Ravensbergen, H. J. C., Latham, K., Rose, A., Myint, J., &
Mann, D. L. (2018). Contrast sensitivity is a signicant predictor of
performance in rie shooting for athletes with vision impairment.
Frontiers in Psychology, 9(950). https://doi.org/10.3389/fpsyg.2018.00950
Bailey, I. L., Jackson, A. J., Minto, H., Greer, R. B., & Chu, M. A. (2012). The
berkeley rudimentary vision test. Optometry and Vision Science, 89(9),
1257–1264. https://doi.org/10.1097/OPX.0b013e318264e85a
Bevers, C., Blanckaert, G., Van Keer, K., Fils, J.-F., Vandewalle, E., & Stalmans, I.
(2019). Semi-automated kinetic perimetry: Comparison of the octopus 900
and humphrey visual eld analyzer 3 versus Goldmann perimetry. Acta
ophthalmologica, 97(4), e499–e505. https://doi.org/10.1111/aos.13940
Hothorn, T., Hornik, K., & Zeileis, A. (2006). Unbiased recursive partitioning:
A conditional inference framework. Journal of Computational and
Graphical Statistics, 15(3), 651–674. https://doi.org/10.1198/
106186006X133933
Ikeda, F., & Kishi, S. (2010). Inner neural retina loss in central retinal artery
occlusion. Japanese Journal of Ophthalmology, 54(5), 423–429. https://
doi.org/10.1007/s10384-010-0841-x
International Blind Sports Federation. (2018). IBSA classication rules. http://
www.ibsasport.org/documents/les/182-1-IBSA-Classication-rules-2018.
pdf
Jackson, T. L., Johnston, R. L., Donachie, P. H. J., Williamson, T. H.,
Sparrow, J. M., & Steel, D. H. W. (2016). The royal college of ophthalmol-
ogists’ national ophthalmology database study of vitreoretinal surgery:
report 6, diabetic vitrectomy. JAMA Ophthalmology, 134(1), 79–85.
https://doi.org/10.1001/jamaophthalmol.2015.4587
Jacobs, D. J., Leng, T., & Flynn, J. H. W. (2011). Delayed-onset bleb-associated
endophthalmitis: Presentation and outcome by culture result. Clinical
Ophthalmology (Auckland, N.Z.), 5, 739–744. https://doi.org/10.2147/
OPTH.S17975
Jones, C., & Howe, P. D. (2005). The conceptual boundaries of sport for the
disabled: classication and athletic performance. Journal of the Philosophy
of Sport, 32(2), 133–146. https://doi.org/10.1080/00948705.2005.9714678
Kons, R. L., Krabben, K., Mann, D. L., Fischer, G., & Detanico, D. (2019). The
eect of vision impairment on competitive and technical-tactical perfor-
mance in judo: Is the present system legitimate? Adapted Physical Activity
Quarterly: APAQ, 36(3), 388–398. https://doi.org/10.1123/apaq.2018-0181
Krabben, K., Ravensbergen, R. H. J. C., Nakamoto, H., & Mann, D. L. (2019).
The development of evidence-based classication of vision impairment
in judo: A delphi study. Frontiers in Psychology, 10(98). https://doi.org/10.
3389/fpsyg.2019.00098
Krabben, K., van der Kamp, J., & Mann, D. L. (2018). Fight without sight: The
contribution of vision to judo performance. Psychology of Sport and
Exercise, 37, 157–163. https://doi.org/10.1016/j.psychsport.2017.08.004
Kuznetsova, A. S., & Barabanshchikova, V. V. (2006). The eects of
self-regulation techniques on human functional states: The moderating
role of dominant sensory modality. Review of Psychology, 12(1), 45–53.
Mann, D. L., & Ravensbergen, H. J. C. (2018). International Paralympic
committee (IPC) and international blind sports federation (IBSA) joint
position stand on the sport-specic classication of athletes with vision
impairment. Sports Medicine, 48(9), 2011–2023. https://doi.org/10.1007/
s40279-018-0949-6
Mashkovskiy, E., Magomedova, A., & Achkasov, E. (2019). Degree of vision
impairment inuence the ght outcomes in the Paralympic judo: A
10-year retrospective analysis. The Journal of Sports Medicine and Physical
Fitness, 59(3), 376–379. https://doi.org/10.23736/s0022-4707.18.08232-4
McNamee, M. J. (2017). Paralympism, Paralympic values and disability sport:
A conceptual and ethical critique. Disability and Rehabilitation, 39(2),
201–209. https://doi.org/10.3109/09638288.2015.1095247
Moshfeghi, D. M., Kaiser, P. K., Scott, I. U., Sears, J. E., Benz, M.,
Sinesterra, J. P., Kaiser, R. S., Bakri, S. J., Maturi, R. K., Belmont, J.,
Beer, P. M., Murray, T. G., Quiroz-Mercado, H., & Mieler, W. F. (2003).
Acute endophthalmitis following intravitreal triamcinolone acetonide
injection. American Journal of Ophthalmology, 136(5), 791–796. https://
doi.org/10.1016/S0002-9394(03)00483-5
Philipp, M., Zeileis, A., & Strobl, C. (2016). A toolkit for stability assessment of
tree-based learners. https://EconPapers.repec.org/RePEc:inn:wpaper:2016-11
Piras, A., Pierantozzi, E., & Squatrito, S. (2014). Visual search strategy in judo
ghters during the execution of the rst grip. International Journal of
Sports Science & Coaching, 9(1), 185–198. https://doi.org/10.1260/1747-
9541.9.1.185
Ravensbergen, H. J. C., Genee, A. D., & Mann, D. L. (2018). Expert consensus
to guide the classication of paralympic swimmers with vision impair-
ment: A delphi study. Frontiers in Psychology, 9(1756). https://doi.org/10.
3389/fpsyg.2018.01756
Ravensbergen, H. J. C., Mann, D. L., & Kamper, S. J. (2016). Expert consensus
statement to guide the evidence-based classication of Paralympic ath-
letes with vision impairment: A delphi study. British Journal of Sports
Medicine, 50(7), 386–391. https://doi.org/10.1136/bjsports-2015-095434
Starosta, W. (2013). Kinesthetic sense and awareness in wrestling: the
structure, conditions and development of an “opponent’s feeling”.
International Journal of Wrestling Science, 3(2), 29–50. https://doi.org/
10.1080/21615667.2013.10878987
Strobl, C., Malley, J., & Tutz, G. (2009). An introduction to recursive
partitioning: Rationale, application, and characteristics of classica-
tion and regression trees, bagging, and random forests.
Psychological Methods, 14(4), 323–348. https://doi.org/10.1037/
a0016973
JOURNAL OF SPORTS SCIENCES 9

Tweedy, S. M., Mann, D. L., & Vanlandewijck, Y. C. (2016). Research needs for
the development of evidence-based systems of classication for physi-
cal, visual and intellectual impairments. In Y. C. Vanlandewijck &
W. R. Thompson (Eds.), Training and coaching of the paralympic athlete.
IOC PressWiley Blackwell. 122-149. https://doi.org/10.1002/
9781119045144.ch7
Tweedy, S. M., & Vanlandewijck, Y. C. (2011). International Paralympic
committee position stand—background and scientic principles of
classication in Paralympic sport. British Journal of Sports Medicine,
45(4), 259. https://doi.org/10.1136/bjsm.2009.065060
World Health Organization. (2004). International statistical classication of
diseases and related health problems (Vol. 1).
10 K. KRABBEN ET AL.