2016 |VOLUME 12 | 141© ARCHIVES OF BUDO | SCIENCE OF MARTIAL ARTS
This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (http://creativecommons.org/licenses/by-nc/4.0),
which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non-commercial and is otherwise in compliance with the license.
Predictors of children´s successful defence against
adult attacker
Michal VítABCDE
, Zdenko ReguliABDE
, Martin SeberaACD
, Jitka ČihounkováABD
,
Martin BugalaABD
Masaryk University, Faculty of Sports Studies, Brno, Czech Republic
Received: 17 March 2016; Accepted: 12 May 2016; Published online: 31 May 2016
AoBID: 11016
Abstract
	 Background&StudyAim:	 The paper is based on the presumption that the probability of successful defence of a child against an adult attacker
is influenced by diversity of variables with different predictive values. The aim of the study are the best
predictors and their presumably impact on children’s chance to defend themselves.
	 Material&Methods: 	 The research sample consisted of 48 students (40 female, 8 male) from three secondary schools. The average
age was 16.6 years. Six self-defence experts performed ex-post evaluation of each video recorded scenario.
Spearman’s rank correlation coefficient, classification trees C&RT and logistic regression were used for analysis.
	 Results:	 Correlation between increasing personal score obtained by evaluation of selected criteria and the probability of
a successful defence was confirmed by the high correlation rs 
= 0.735 significance level of p<0.05. Active defence,
escape and technical means respectively, were found the best predictors out of the total number of six evaluation
criteria. Communication and safe distance keeping varied in the fifth position depending on the selected statistical
method. Guard position was found the weakest predictor.
	 Conclusions:	 There are 13.88 times higher odds of successful defence when children are dealing with an adult attacker actively.
The activity should be aimed at looking for an escape route as there are 7.69 times higher odds of successful
defence when the child is trying to escape. Finally, there are 3.75 times higher odds of successful defence
when the child uses appropriate technical means to distract attacker’s attention.
	 Keywords:	 combatives • defensive action • fighting skills • physical education • self-defence • self-protection • protective
strategy • victimology
	Copyright: 	 © 2016 the Authors. Published by Archives of Budo
	 Conflictofinterest: 	 Authors have declared that no competing interest exists
	 Ethicalapproval:	 The study was approved by the Local Ethics Committee
	 Provenance&peerreview:	 Not commissioned; externally peer reviewed
	 Sourceofsupport:	 Departmental sources
	 Author’saddress: 	 Michal Vít, Masaryk University, Faculty of Sports Studies, Kamenice 753/5, 625 00, Brno, Czech Republic,
e-mail: vit@fsps.muni.cz
Authors’ Contribution:
A	 Study Design
B	 Data Collection
C	 Statistical Analysis
D	Manuscript Preparation
E Funds Collection
 ORIGINAL ARTICLE
142 | VOLUME 12 | 2016 www.archbudo.com
Original Article
Introduction
Children’s self-defence is widely discussed concern
[1-9] as children are by nature more vulnerable
than adult people [10]. According to the legislation
of European Union and Czech Republic,
children are considered particularly vulnerable
victims [11, 12] and in case of victimization they
should be provided with specialist support and
legal protection (e.g. interrogation in a special
room by a person specifically trained, preventing
re-encounter with the offender etc.). Parents are
primarily responsible for their children but they
cannot accompany them all day. Thus children
must be responsible for their well-being and in
case of risk situation use their own resources to
defend themselves [2, 5]. There is much jeopardy
in the environment nowadays and it is naturally
not possible that one person cope with that all at
the same time.
The aim of children’s self-defence is not to create
an ultimate protective shield but to develop
personal strategy for each child based on their
individual conditions and build habits for safe
behaviour without any fear. Achieving a sense
of comfort, irreplaceable by other feelings, concurrently
with optimal psycho-physical readiness
are the ultimate objectives of self-defence
training [13]. Teaching of combative games (fun
forms of martial arts [14]), combat sports, martial
arts and self-defence should be considered a tool
for developing risk management among chil-
dren [15]. Learning by doing principle is an optimal
approach for children’s self-defence classes
allowing them to develop their innate capabilities
for self-reliance and an appropriate response
in dangerous situations [1]. Children can benefit
from self-defence teaching by establishing
awareness, increasing self-confidence and self-discipline,
developing mental strategies, improving
communication skills, physical skills, fitness and
the ability to fight back if necessary [5].
Taxonomy of combatives consists of three levels;
the first level contains preparatory combatives, the
second one combative sports and the third one
self-defence, which is recognised as an application
level of combatives in the defence of necessity by
law [16-18]. Theory and didactics of combatives
as well as self-defence are largely described in the
literature with regards to psychological, technical,
psychomotor, tactical, social, and legal aspects
[13, 19-30]. The main objective of self-defence
training is to increase participant´s defensive
skills [31]. Of course, the motor competence is
noticeably emphasized during the training process
[23, 31], although from the praxeology point
of view both physical and intellectual factors are
influential on effect of any action [32].
Considering conditions in self-defence situations,
we have to take into account that the physical
defence is the last resort of self-preservation as
the probability to win the real confrontation is
unreliable. This presumption is especially valid
when the defender is weaker than the offender
(common modus operandi). From the tactical
point of view the didactics of children’s selfdefence
have to be divided into three separate
types of risk situations according to the physical
superiority of the opponent: defence against
peers (e.g. an opponent of the same age and probably
around similar physical mass), defence against
youngsters considerably older and stronger than
the defender (e.g. child at the age of 8 against
a 15 years old bully) and defence against an adult
assailant (e.g. a kidnapper, a mugger, or a rapist
etc.). These levels are certainly simplified since
big differences in the struggle between opponents
can occur (e.g. an attacker of the same age but
considerably more mature and stronger). On the
other hand, previous research showed that physical
fitness and body mass are not condition for
the results of a hand-to-hand struggle; at least
in children of the same age and adequate psychomotor
competence they appear to be the prerequisite
for winning a hand-to-hand struggle and
self-defence [21].
There are many private companies offering selfdefence
courses for children where development
of practical skills is expected and required [6-9].
Inclusion of combatives and self-defence in physical
education has also been a topic for scholars
[5, 19, 22, 24, 33-38] since many years. Our study
following from the current situation in the Czech
Republic presents self-defence not just as a commercial
and a theoretical topic but as a regular
part of physical education curricula of elementary
and secondary schools [3, 4, 24, 38]. Although
current curricula have been in force since 2007,
proper academic research into evaluation of learning
outcomes from self-defence classes is missing.
The present study is a follow up to our previous
publications [39-42] where conceptualization
of children self-defence in PE classes and
use of scenario training were given. Coming out
from the results of scenario training in our High
Combatives - combatives are
physical exercises aimed at
physical defeat of a partner.
Combatives comprise specific
exercises which prepare
a participant to overcome
a partner by physical contact.
Defend actions – the act of
defending yourself or someone or
something from attack [45].
Defensive action – is the
appropriate movement of the
armed arm with weapon with
intent to prevent the attack of
opponent [45].
Particularly vulnerable victim
- persons who are particularly
vulnerable or who find themselves
in situations that expose them to
a particularly high risk of harm,
e.g. persons subjected to repeat
violence in close relationships,
victims of gender-based violence,
children, disabled people etc.
Self-defence - open system of
any precautions and actions
preventing victimization,
deflecting acute threat using
tactical, psychological, verbal or
physical tools including use of
reasonable force in accordance
with the law in the defence of
necessity and mitigation of the
negative effects of assault.
Scenario training - the method
used in the self-defence teaching
both for training and evaluation
purposes consisted in the preprepared
course of a situation
similar to a theatre script with
fixed roles and activity of the
actors where the roles of the
actors (assailants) and the
defender (a tested person) are
already given.
Victim - a person who is actually
and directly affected by an act
or omission that is incompatible
with the European Convention
on Human Rights, or a person
who is at risk of being directly
affected [46].
Victimology - an area of
criminology which studies
the victim of crime and their
relationship with offenders [46].
Victimization - the process in
which one becomes a victim of
crime.
Vít M et al. – Predictors of children´s successful defence...
© ARCHIVES OF BUDO | SCIENCE OF MARTIAL ARTS 2016 |VOLUME 12 | 143
School Self-defence Project (HSSP), we focused
on the importance of all evaluation criteria and
their predictive value for youngsters’ successful
defence against an adult assailant.
The aim of the study are the best predictors and
their presumably impact on children’s chance to
defend themselves.
Material and methods
The study includes data collected during scenario
training evaluation which was carried out as a final
test of HSSP in 2011. Three secondary schools
from Czech Republic were included in the project,
where three PE teachers (2 female, 1 male)
provided students with 12 self-defence classes
according to HSSP methodology. Thereafter,
scenario training (a simulation attack by an adult
assailant) and evaluation of students’ reactions
were provided.
Participants
Students from three secondary schools (A, B, C)
participated in the research. There were 19 female
A school (Business Academy and Nursing School)
participants,12 female and 8 male students B
school (Nursing School) participants and 8 female
C school (Sport High School) participants. Since
out of the whole sample just 8 persons were male,
we did not search for differences between tested
persons by sex during the analysis. The total sample
of tested persons was 48 (40 female, 8 male)
at the average age of 16.6 years.
Methods
Scenario training
Scenario training is a frequently used method in
the self-defence teaching both for training and
evaluation purposes. The scenario is understood
as a pre-prepared course of situation in which the
roles of the actors are already given. To a certain
extent, scenario training resembles a theatre script
with fixed roles and activity of actors [43, 44]. In
our research the scenario training was used for
assault simulation by an adult person on teenagers.
Each situation took place in a well-lit room
which contained various objects (obstacles) [42].
Figurant´s role
The attacker´s task was to attack the tested person
always first verbally and then physically.
The attack was aimed with proportional intensity
catch the tested person unawares and stress
them, to make physical and psychical stress leading
to the victory represented by knocking down
and final defeat [42].
Tested person´s role
The task of the tested person was to enter the
room and defend themselves against a possible
attack of the figurant and escape. The tested person
knew the room in advance but did not know
the obstacles. The tested person knew in advance
that they would become part of scenario training
during which they would (probably) be verbally
and physically attacked. The tested person
also knew in advance that there would be a trainer
and a researcher (both already known) and a third
unknown person (figurant) with a mask covering
the face.
Evaluation criteria
During the 12 self-defence classes of HSSP project
students were taught to defend themselves
both verbally and physically against a dangerous
person by following the tactical plan: to occupy an
appropriate position and posture for the defence,
keep the safe distance from the dangerous person,
answer verbally to attacker´s offensive activity,
observe environment and look for escape route,
repulse possible acute physical attack by appropriate
technical means (e.g. push away, strike,
kick etc.), and escape to safe distance or place.
Mnemotechnical clue of traffic light was used for
youngsters, where red light stood for first visual or
verbal contact with the dangerous person enhancing
the awareness, yellow light stood for verbal or
physical defence and preparation of escape route
and finally, green light stood for start and running
away. Based on the above strategy, the elements
listed in the Table 1 were selected as evaluation
criteria of children´s correct behaviour in the selfdefence
situation [42].
These criteria were considered representative
attributes of appropriate behaviour in youngsters´
self-defence which may enhance the defenders´
odds of successful defence. These evaluation criteria
give evidence of the quality of defensive action
of a tested person. The last and definitive evaluation
criterion (C7) was the result assessment of
scenario training. Tested persons´ behaviour was
evaluated pursuant to successful or not successful
defence of the tested person at the end of scenario.
This evaluation criterion gives evidence of
the result of defensive action of the tested person.
144 | VOLUME 12 | 2016 www.archbudo.com
Original Article
Experts’assessment
Six self-defence experts evaluated each video
recorded scenario situation ex-post. Behaviour of
each tested person was evaluated dichotomically
(0 or 1) in the excel file using selected C1-C7 criteria
mentioned above.
Statistical analysis
After provided experts’ assessment we proceeded
to analyse the data which consisted of dichotomical
evaluation of all criteria. Three statistical
methods were involved to analyse the data and
verify the results. Each tested person obtained
0 or 1 evaluation for each C1-C7 criterion. The
sum of evaluation of six C1-C6 criteria created
a personal score given by each of six experts.
Arithmetic mean of six personal score was set
up as a definitive value of personal score which
was used for further analysis. By evaluation of C7
criterion the ratio of experts unity was created,
defined as “defended: not defended”. This ratio
was transformed into numerical value according
to conversion table (Table 2).
Table 2. Conversion table of expert assessment of numeric
value.
Ratio of experts unity Numeric value
00:05 0
01:04 0.25
02:03 0.5
03:02 0.5
04:01 0.75
05:00 1
For determination of dependence between definitive
personal score and ratio of experts unity on
scenario result Spearman‘s rank correlation coefficient
was used. Two statistical methods were
involved for the analysis of predictors of successful
defence: C&RT classification trees and logistic
regression.
C&RT classification trees
In a standard classification tree the idea is to split
the dataset according to homogeneity of data.
The main idea of tree methods is to divide recursively
the data into smaller and smaller strata in
order to improve the fit as best as possible. The
sample space is originally split into two regions.
For each of the two created regions this process
is repeated again. The major components of the
C&RT methodology are selection and stopping
rules. The selection rule determines which stratification
to perform at each stage and the stopping
rule determines the final strata that are formed.
Once the strata have been created, the impurity
of each stratum is measured. The heterogeneity
of the outcome categories within a stratum
is referred to as “node impurity”. In each node
the variable by which we divide the data file and
borders is determined, showing where the division
was done. The method enables to arrange
the variables according to their classification
importance.
Logistic regression
Logistic regression measures the relationship
between the categorical dependent variable and
one or more independent variables by estimating
probabilities using a logistic function. Like other
forms of regression analysis, logistic regression
makes use of one or more predictor variables that
may be either continuous or categorical. The logit
of success is then fitted to the predictors using
Table 1. Evaluation criteria and their description.
Code Evaluation criteria Description
C1 Guard position
Occupies an appropriate position and posture for the defence (hidden guard,
defensive guard, offensive guard etc.).
C2 Communication with the attacker
Communicates with the attacker and the environment (uses a call, trap, calls
for help etc.).
C3 Safe distance keeping Works correctly with the distance (elongation, shortening, maintenance).
C4 Active defence against attack Actively defends him/herself against the attack.
C5 Technical means for distracting from the
attacker
Uses appropriate technical means to defend him/herself.
C6 Escape to a safe place
Course of defence heads towards breaking away from the attacker and escaping
to the safety.
Vít M et al. – Predictors of children´s successful defence...
© ARCHIVES OF BUDO | SCIENCE OF MARTIAL ARTS 2016 |VOLUME 12 | 145
linear regression analysis. The predicted value of
the logit is converted back into predicted odds
via the inverse of the natural logarithm, namely
the exponential function. Thus, although the
observed dependent variable in logistic regression
is a zero-or-one variable, the logistic regression
estimates the odds as a continuous variable
whose dependent variable is a success (a case).
This categorical prediction can be based on the
computed odds of a success, with predicted odds
above individual selected cut-off value translated
into a prediction of a success.
Results
Spearman’s rank correlation coefficient between
the tested persons´ personal score and the ratio of
experts unity on scenario result reaches the value
rs 
= 0.735. This is high correlation at a significance
level of p<0.05. The calculation shows a correlation
between the increasing personal score and the
probability of a successful defence. We interpret the
result as follows: C7 evaluation criterion (result of
scenario training) is dependent on the quality of
defence expressed by C1- C6 evaluation criteria.
In the following analysis, we concentrated on
finding out which of the C1-C6 evaluation criteria
are the best predictors for successful defence.
C&RT classification trees show the best model
of predictors (Figure 1, Table 3).
Figure 1 and Table 3 display the tree with
schema of variables classification. Each node is
designated by its ID and includes the number
of cases and frequency rate of a dependant variable.
There is a branch with information about
rule from each node which divides the node into
2 subnodes (binary tree). Out of total number
of n = 288 measurements in the ID = 1 node
the number of n = 131 cases were not successful
in defence and n = 157 cases were successful.
This group is best divided by escape variable into
two subnodes: ID = 2 and ID = 3. The ID = 2
node incorporates the sum of 115 cases which did
not meet the escape criterion. Out of this node
n = 94 cases were not successful in defence and
n = 21 cases succeeded. There is a total number
of n = 173 cases in the ID = 3 node which met
the escape criterion. Out of this node n = 37 cases
did not succeed in defence and n = 136 cases were
personal score and the probability of a successful defence. We interpret the result as follows:
C7 evaluation criterion (result of scenario training) is dependent on the quality of defence
expressed by C1- C6 evaluation criteria.
In the following analysis, we concentrated on finding out which of the C1-C6 evaluation criteria
are the best predictors for successful defence. C&RT classification trees show the best model
of predictors (Figure 1, Table 3).
Figure 1. Binary tree with classification schema.Figure 1. Binary tree with classification schema.
146 | VOLUME 12 | 2016 www.archbudo.com
Original Article
successful. This node is further best divided by
technical means variable into 2 subnodes: ID = 18
and ID = 19. Out of the total number of n = 128
cases in the node ID = 18, which used technical
means, n = 17 were not successful and n = 111
were successful. The node ID = 19 consists of
n = 45 cases which met technical means criterion.
In that node there was almost a balanced
number of cases connected with unsuccessful
defence (n = 20) and cases with successful defence
(n = 25). This node is further best divided by
active defence criterion into 2 subnodes: ID = 30
and ID = 31. In the ID = 30 node cases which did
not meet active defence criterion appeared (n = 8).
Out of this node n = 7 cases were not successful
in the defence and just n = 1 was successful. On
the other hand, cases fulfilling the active defence
criterion were included in the node ID = 31.
Out of the total number of n = 37 in that node,
n = 13 cases were connected with unsuccessful and
n = 24 with successful defence.
Based on the results of C&RT analysis displayed
in the Figure 1 and Table 3, we can conclude
that the best predictors for successful defence are
escape, technical means and active defence respectively.
In other words, the greatest chance of
successful defence had those tested persons
who were trying to escape (criterion escape = 1),
while the best predictor for failure was criterion
of escape = 0. The probability of success in
defence increased with meeting criterion of technical
means = 1. Moreover, the chances of success
increased in tested persons who were vigorous
in the defence and met the criterion of active
defence = 1. Another point of view of predictors
is offered by the allocation of independent variables
(C1-C6 criteria) according to their importance
of successful defence (C7 criterion). The
Importance plot (Figure 2) shows the best predictors
which mostly affect the dependent variable
of Defended.
Based on the data in Figure 2 we conclude that
the best predictors for successful defence are
evaluation active defence (C4), escape (C6), technical
means (C5), safe distance keeping (C3), communication
(C2) and guard position (C1) criteria
respectively.
To verify the model quality, we provided the analysis
of good and bad included cases. Since there
are correct classifications (Table 4), we consider
the model successful.
Table 4. Rate of classification success in C&RT
#
Frequency table for predicted variable Predicted
Observed variable: Observed
OBSERVED PREDICTED
COUNT
Observed Predicted
1 0 0 101
2 0 1 30
3 1 0 22
4 1 1 135
Table 3. Structure of binary tree.
Node #
Tree structure 2 Dependent variable: Defended Options: Categorical response,Tree number 2
Left
branch
Right
branch
Size of
Node
N in
class 0
N in
class 1
Selected
category
Split
variable
Split
constant
Split
category
1 2 3 288 131 157 1 Escape   0
2     115 94 21 0      
3 18 19 173 37 136 1
Technical
means
  1
18     128 17 111 1      
19 30 31 45 20 25 1
Active
defence
  0
30     8 7 1 0      
31     37 13 24 1      
Vít M et al. – Predictors of children´s successful defence...
© ARCHIVES OF BUDO | SCIENCE OF MARTIAL ARTS 2016 |VOLUME 12 | 147
Logistic regression was performed by the following
procedure. Variable of Defended (C7) was
determined as a dependent variable and C1-C6
evaluation criteria of guard position, communication,
safe distance keeping, active defence, technical
means and escape as independent variables. Most
relevant results of logistic regression are displayed
in the Table 5.
Based on the data presented in the Table 5,
we determined the best predictors for successful
defence in the following order: 1) active
defence, 2) escape, 3) technical means, 4) communication.
Explicitly, the tested person with active
defence = 0 evaluation has 13.88 times higher odds
of unsuccessful defence (Defended = 0) than the
tested person with active defence = 1 evaluation.
The tested person with escape = 0 evaluation has
7.69 times higher odds of unsuccessful defence
(Defended  =  0) than the tested person with
escape = 1evaluation. Tested person with technical
means = 0 evaluation has 3.75 times higher odds
of unsuccessful defence (Defended = 0) than the
tested person with technical means = 1evaluation.
Tested person with communication = 0 evaluation
has 3.03 times higher odds of unsuccessful defence
(Defended = 0) than the tested person with com-
munication = 1 evaluation. For verification of our
who were trying to escape (criterion escape = 1), while the best predictor for failure was
criterion of escape = 0. The probability of success in defence increased with meeting criterion
of technical means = 1. Moreover, the chances of success increased in tested persons who were
vigorous in the defence and met the criterion of active defence = 1. Another point of view of
predictors is offered by the allocation of independent variables (C1-C6 criteria) according to
their importance of successful defence (C7 criterion). The Importance plot (Figure 2) shows the
best predictors which mostly affect the dependent variable of Defended.
Figure 2. Importance plot of evaluation criteria
Based on the data in Figure 2 we conclude that the best predictors for successful defence are
evaluation active defence (C4), escape (C6), technical means (C5), safe distance keeping (C3),
communication (C2) and guard position (C1) criteria respectively.
To verify the model quality, we provided the analysis of good and bad included cases. Since
there are %9.81
288
236
 correct classifications (Table 4), we consider the model successful.
Figure 2. Importance plot of evaluation criteria
Table 5. Odds Ratios for successful defence
Effect
Defended - Odds Ratios Distribution : BINOMIAL, Link function: LOGIT
Modelled probability that Defended = 0
Level of
effect
Column Odds Ratio
Lower CL
95.0%
Upper CL
95.0%
p
Intercept   1        
Guard position 0 2 0.596 0.277 1.283 0.186
Communication 0 3 3.033 1.376 6.687 0.006
Safe distance keeping 0 4 2.069 0.947 4.521 0.068
Active defence 0 5 13.882 2.683 71.833 0.002
Technical means 0 6 3.746 1.837 7.640 0.000
Escape 0 7 7.691 3.786 15.626 0.000
148 | VOLUME 12 | 2016 www.archbudo.com
Original Article
results we tested the model quality (Table 6).
Since the p-value = 0.798951 for deviance goodness
of fit test and p-value = 0.808195 for Pearson
chi-square goodness of fit test, we conclude
hypothesis that the logistic regression response
function is appropriate. The R2 
measures of predictive
power of 0.45 and 0.6 confirm the appropriateness
of the model. The total success rate of
classification (Table 7) in our model is .
The model quality was also confirmed by ROC curve
(Receiver Operating Characteristic, Figure 3). It is
a tool for assessing and optimizing the binary classification
system (test) that shows the relationship
between the specificity and sensitivity of the test or
the detector for all permissible threshold value. The
vertical axis of the ROC curve is the relative frequency
of true positive cases i.e. the probability that
as a correct one will be evaluated positive case. The
horizontal axis is the relative frequency of false positive
cases, i.e. the probability that as a correct one
will be evaluated negative case. The best model is the
one with the largest area under the curve which can
theoretically reach the value of 1.00. In our analysis
the value equals 0.8992.
Discussion
Although we found quite precisely the best predictors
for successful defence of children against
an adult attacker using different statistical methods
and verifying quality of the models, we have
to consider following limitations of our study.
Firstly, we did not analyse real self-defence cases
but confrontation between an adult assailant and
youngsters in a simulated scenario training. Thus
our study is limited by all limitations typical of scenario
training in general. These limitations include
for instance knowing the tested persons, awareness
of the fact that the attack is simulated and no real
threat is present. Furthermore, it is the absence of
other citizens who can witness the real situation in
the public environment etc. In addition, all possible
factors which could be present in the real eco-system
(multiple attackers, use of weapons etc.), could
play additional role in the real situation. Due to
this the number of factors is limited by the design
of scenario training which focuses on one tangible
situation. On the other hand, the attack of one
person in a confined space was realistic enough and
children experienced it very emotionally. We consider
the course of simulation well-structured and
adequate for the research purposes.
Table 6. Model quality
Indicator
Defended - Statistics of goodness of fit Distribution:
BINOMIAL, Link function:
LOGIT Modelled probability that Defended = 0 (Analysis sample)
Df Stat. Stat/Df
Deviance 281 224.505 0.799
Pearson Chi2 281 227.103 0.808
AIC   238.505  
AICC   238.905  
BIC   264.146  
Cox-Snell R2   0.450  
Nagelkerke R2   0.602  
Table 7. Rate of classification success in logistic regression
 
Classification of cases Odds ratio: 25.965714
Log odds ratio: 3.256777
Predicted:0 Predicted:1 Percent correct
Observed: 0 96 35 73.282
Observed: 1 15 142 90.446
Vít M et al. – Predictors of children´s successful defence...
© ARCHIVES OF BUDO | SCIENCE OF MARTIAL ARTS 2016 |VOLUME 12 | 149
Finally, our study focuses on predictors which
were previously considered influential in selfdefence
situations. Although we regard these six
predictors as representative attributes of appropriate
behaviour of children facing violence from an
adult attacker, there could be other influencing
factors which we did not include in the analysis.
On the other hand, we consider the high correlation
(rs 
= 0.735, significance level of p<0.05)
between personal score of tested person, which is
created from selected evaluation criteria, and the
result of scenario to be the evidence of valid selection
of these criteria.
Conclusions
Predictors of children´s successful defence against
an adult attacker were explored by two statistical
methods. Both C&RT classification trees and
logistic regression methods confirmed that from
the total six evaluated criteria the best predictors
are active defence, escape and technical means
respectively. Additionally, communication and
safe distance keeping varied in the fourth and fifth
position dependently on the selected statistical
method. Guard position emerged as the weakest
predictor. Both for the didactics and evaluation
of children´s self-defence training following practical
implications are relevant. The activity itself
is the most important predictor for self-defence
situation solving.
According to the logistic regression there are
13.88 times higher odds of successful defence
when the child behaves actively against an adult
attacker. The activity should be aimed at looking
for escape route as there are 7.69 times higher
odds of successful defence when the child is trying
to escape to the safe place. There are 3.75 times
higher odds of successful defence when the child
uses appropriate technical means to distract the
attacker, which enables them to escape. Finally,
when the child communicates with the attacker
correctly, there are 3.03 times higher odds of success
in the self-defence situation. Although safe
distance keeping and guard position appeared the
weakest predictors, we have to realize that these
factors are significant tactical elements especially
in the pre-conflict phase. Having regard to the
foregoing limitations, further research with wider
scope of predictors is recommended for more
complex understanding of the matter.
Figure 3. ROC curve
The model quality was also confirmed by ROC curve (Receiver Operating Characteristic,
Figure 3). It is a tool for assessing and optimizing the binary classification system (test) that
shows the relationship between the specificity and sensitivity of the test or the detector for all
permissible threshold value. The vertical axis of the ROC curve is the relative frequency of true
positive cases i.e. the probability that as a correct one will be evaluated positive case. The
horizontal axis is the relative frequency of false positive cases, i.e. the probability that as a
correct one will be evaluated negative case. The best model is the one with the largest area
under the curve which can theoretically reach the value of 1.00. In our analysis the value equals
0.8992.
DISCUSSION
Although we found quite precisely the best predictors for successful defence of children against
an adult attacker using different statistical methods and verifying quality of the models, we have
to consider following limitations of our study.
Firstly, we did not analyse real self-defence cases but confrontation between an adult assailant
and youngsters in a simulated scenario training. Thus our study is limited by all limitations
typical of scenario training in general. These limitations include for instance knowing the tested
persons, awareness of the fact that the attack is simulated and no real threat is present.
Figure 3. ROC curve
150 | VOLUME 12 | 2016 www.archbudo.com
Original Article
References
1.	 Bishop B, Thomas M. Protecting children from
danger: building self-reliance and emergency
skills without fear: a learning by doing book for
parents and educators. Berkeley, Calif: North
Atlantic Books; 1993: 185
2.	 Kraizer SK. The safe child book: a common
sense approach to protecting children and teaching
children to protect themselves. New York:
Fireside; 1996: 176
3.	 Research Institute of Education (VÚP) in Prague.
Framework Education Programme for Elementary
Education (FEPEE) [Internet]. 2007 [accessed
2016 Feb 16]. Available from: URL:http://www.
vuppraha.cz/wp-content/uploads/2009/12/RVP_
ZV_EN_final.pdf
4.	 Research Institute of Education (VÚP) in Prague.
Framework Education Programme for Secondary
General Education (Grammar Schools). 2007
[accessed 2016 Feb 16]. Available from: URL:
http://www.nuv.cz/file/161_1_1
5.	 Potenza GM, Konukman F, Yu J-H et al. Teaching
Self-Defense to Middle School Students in
Physical Education. J Phys Educ Recreat Dance
2014; 85(1): 47-50
6.	 Fast Defense Global. Kid Course – FAST Defense
Global, Inc [Internet]. 2016 [accessed 2016 Feb
16]. Available from: URL:https://fastdefense-
global.org/course-content/kids-courses
7.	 Katz M. How Krav Maga Kids Will Benefit Your
Child [Internet]. 2016 [accessed 2016 Feb 16].
Available from: URL:http://www.your-krav-
maga-expert.com/krav-maga-kids.html
8.	 Krav Maga Worldwide. km-X Youth Program
[Internet]. 2016 [accessed 2016 Feb 17]. Available
from: URL:http://kravmaga-kids.com
9.	 Radkids. The national leader in children’s safety
[Internet]. 2016 [accessed 2016 Feb 16]. Available
from: URL:http://www.radkids.org
10.	Čírtková L. Viktimologie pro forenzní praxi.
Praha: Portál; 2014
11.	The European Parliament and the Council of
the European Union. Directive 2012/29/EU
of the European Parliament and of the Council
of 25 October 2012 [Internet]. 2012 [accessed
2016 Jan 6]. Available from: URL:http://eur-
lex.europa.eu/LexUriServ/LexUriServ.do?uri=
OJ:L:2012:315:0057:0073:EN:PDF
12.	Zákon o obětech trestných činů č. 45/2013 Sb.
[Internet]. 2013 [accessed 2016 Jan 6]. Available
from: URL:http://www.zakonyprolidi.cz/cs/2013-
45 [in Czech]
13.	Harasymowicz J, Kalina R. Training of psychomotor
adaptation – a key factor in teaching selfdefence.
Arch Budo 2005; 1: 19-26
14.	Jagiełło W, Kalina RM, Klimczak J et al. Fun
forms of martial arts in positive enhancement
of all dimensions of health and survival abilities.
In: Kalina RM (ed.) Proceedings of the 1st
World Congress on Health and Martial Arts in
Interdisciplinary Approach, HMA 2015, 17–19
September 2015, Czestochowa, Poland. Warsaw:
Archives of Budo; 2015: 32–39
15.	Lykkegaard M. The connection between martial
arts and rough & tumble play (RTP). 10th
International Conference on Kinanthropology;
2015 Nov 18-20; Brno, Czech Republic. Brno:
Masaryk University; 2015
16.	Reguli Z. Combative activities. Brno: Paido; 2004
17.	Reguli Z. Úpolové sporty: distanční studijní
text. Brno: Masarykova univerzita v Brně; 2005:
133 [in Czech]
18.	Reguli Z. Taxonomy of combatives as it is seen
from Tyrš tradition in the Czech Republic. Ido
Mov Cult 2009; 9: 38-43
19.	Ďurech M, Mlsna B, Čuperka F et al. Úpoly.
Bratislava: Univerzita Komenského; 2000 [in
Czech]
20.	Ashkinazi S, Jagiełło W, Kalina RM et al. The
importance of hand-to-hand fights for determining
psychomotor competence of antiterrorists.
Arch Budo 2005; 1: 19-26
21.	Kalina RM, Chodala A, Dadelo S et al. Empirical
basis for predicting success in combat sports and
self-defence. Kinesiology 2005; 37(1): 64-73
22.	Bartík P, Sližik M, Reguli Z. Teória a didaktika
úpolov a bojových umení. Banská Bystrica:
Univerzita Mateja Bela v Banskej Bystrici; 2007
[in Czech]
23.	Kalina RM, Jagiełło W, Wiktorek P. Motor competence
in self-defence of students of a detectives’
school during their course of studies. Arch
Budo 2007; 3: 1-6
24.	Reguli Z, Ďurech M, Vít M. Teorie a didaktika
úpolů ve školní tělesné výchově. Brno: Masarykova
univerzita; 2007 [in Czech]
25.	Adamčák Š, Bartík P. Názory žiakov na vyučovanie
úpolov v telesnej a športovej výchove na 2. stupni
základnej školy. Banská Bystrica: Univerzita
Mateja Bela v Banskej Bystrici; 2011 [in Czech]
26.	Cynarski WJ. Martial arts phenomenon - research
and multidisciplinary interpretation. Rzeszów:
Wydawnictwo Uniwersytetu Rzeszowskiego; 2012
27.	Pavelka R, Stich J. Vývoj bojových sportů. Praha:
Karolinum; 2012 [in Czech]
28.	Vít M, Reguli Z, Chvátalová J. Základy osobní
sebeobrany [Internet]. 2012 [accessed 2016 Jan
13]. Available from: URL:http://is.muni.cz/
do/rect/el/estud/fsps/ps11/sebeob/web/index.
html [in Czech]
29.	Cynarski W. The meaning of self-defence: an
expert definition. A contribution to the theory
of self-defence and combat. 10th International
Conference on Kinanthropology; 2015 Nov
18-20; Brno, Czech Republic. Brno: Masaryk
University; 2015
30.	Kohoutková J, Čihounková J, Skotáková A et
al. Self-defence for people with visual impairments.
Ido Mov Cult J Martial Arts Anthropol
2015; 15(2): 33-36
31.	Staller M. Bridging the Gap: Investigating
Effectiveness in Self-Defence. Cardiff: Cardiff
University; 2015
32.	Kalina RM, Barczyński BJ. From “physical fitness”
through “motor competence” to the “possibility
of action”. Arch Budo 2008; 4: 106-109
33.	Fojtík I. Džúdó, karatedó, aikidó, sebeobrana.
Praha: NS Svoboda; 1998 [in Czech]
34.	Brown D, Johnson A. The social practice of selfdefense
martial arts: Applications for physical
education. Quest 2000; 52(3): 246-259
35.	Sasaki T. The meaning and role of budo (the
martial arts) in school education in Japan. Arch
Budo 2006; 2: 11-14
36.	Jagiello W, Dornowski M. Martial arts in the
opinions of students at the Faculty of Physical
Education. Arch Budo 2011; 7(2): 55-59
37.	Svobodová L, Vaculíková P, Hlavoňová Z et
al. Trendy v realizaci pohybové aktivity dětí
mateřských škol a 1. stupně základních škol.
Brno: Masarykova univerzita; 2015 [in Czech]
38.	Vít M, Reguli Z. The role of combatives teaching
in physical education. Brno: Masaryk University;
2015: 128
39.	Vít M, Reguli Z. Koncepce výuky sebeobrany na
středních školách. Sport a kvalita života. Brno:
Masarykova Univerzita; 2008: 168-168 [in Czech]
40.	Vít M, Reguli Z. Modelové sebeobranné situace
ve školní tělesné výchově. Šport a zdravie.
Nitra: Univerzita Konštantína Filozofa; 2008:
97-103 [in Czech]
41.	Vít M. Sebeobrana a její mezioborové přesahy
v Rámcovém vzdělávacím programu pro gymnázia.
Sport a kvalita života Sborník abstraktů
mezinárodní konference konané; 2009 Nov 5-6;
Brno, Czech Republic. Brno: Masarykova univerzita;
2009 [in Czech]
42.	Vít M, Reguli Z, Chvátalová J. Use of scenario
training in high school self-defence teaching.
Stud Sport 2011; 5(3): 339-344
43.	Wagner J. Reality-based personal protection.
Valencia, CA: Black Belt Communications;
2005: 377
44.	Vít M, Kohoutková J, Bugala M et al. Evaluation
of stress conditions in self-defence scenario training.
3rd World Scientific Congress of Combat
Sports and Martial Arts. Rzeszów: University
of Rzeszów; 2015
45.	Dadelo S, Veršinskas R, Piwowarski J et al. The
most commonly used arrest and self-defence
actions arsenal by different officers of internal
services. Arch Budo 2015; 11: 285-291
46.	Gooch G, Williams M. A dictionary of law
enforcement. Oxford, New York: Oxford
University Press; 2007
Cite this article as: Vít M, Zdenko Reguli Z, Sebera M et al. Predictors of children´s successful defence against adult attacker. Arch Budo 2016; 12: 141-150