JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2019, 9999, 1–11 NUMBER 9999 ()

Application of the matching law to Mixed Martial Arts

HOLLY A. SENIUK AND JANIE P. VU
UNIVERSITY OF NEW BRUNSWICK, FREDERICTON

MELISSA R. NOSIK
BEHAVIOR ANALYST CERTIFICATION BOARD

In the contemporary behavior-analytic literature, athletic performance (e.g.,choice of shot or

movement) across multiple team sports has been found to correspond with predictions of
the generalized matching equation. However, the research in this area has focused primarily on
team sports. In the current study the Generalized Matching Equation (GME) was applied to
Mixed Martial Arts (MMA) performance by examining strike selection as a function of landing
significant strikes among fighters from various weight classes in the Ultimate Fighting Champion-
ship. The results suggest that the GME is a good descriptor of strike selection in MMA, an indi-
vidual sport that is dynamic and fast paced where responding results in immediate feedback from
an opponent.
Key words: choice, generalized matching equation, matching law, mixed martial arts, sports

In 2005 Kofi Annan (Former United Nations’ Muise, 2011). Because of the significant influence
Secretary General) described sports as “…a univer- of sports in society, it is important to consider the
sal language that can bring people together, no role that behavior analysis could play in this
matter what their origin, background, religious industry.
beliefs or economic status” (United Nations, A recent development in behavior-analytic
2005, p. 1). In 2011 it was estimated that approx- research pertaining to sports is the application of
imately 117 million people watched the Green the matching law in athletic performance. The
Bay Packers play the Pittsburgh Steelers in the matching law states that when an organism is pres-
Super Bowl (Bauder, 2011), while an average of ented with two or more response options they allo-
only 59.2 million viewed the 2012 US Presiden- cate their responses proportionally to the amount
tial debates (Neilsenwire, 2012). In the same year, of reinforcement obtained for each alternative
the Province of Quebec French Canadian debate (Herrnstein, 1961). The Generalized Matching
was to air during the same time as the Montreal Equation (GME; Baum, 1974), while algebraically
Canadiens were scheduled to play their rivals, the equivalent to the original matching law equation,
Boston Bruins, in National Hockey League uses logarithmic transformations of the ratios of
playoffs. Given concerns about the lack of viewers responses and reinforcers to permit the use of linear
for the debate, political leaders moved the debate regression lines. Given that “perfect” matching
to avoid overlap with the playoff game (Cohen & (i.e., slope is equal to 1) is rarely observed in
research, this allows for an analysis of data that
Holly A. Seniuk is now at Behavior Analyst Certifica- is easier to interpret (Reed & Kaplan, 2011).
tion Board. Janie P. Vu is now at University of Guelph. Researchers have demonstrated applications of
The authors would like to thank Dr. David Cox and
Joshua Wright for their assistance conducting the analyses the GME in team sports such as baseball (Cox,
for this article. Sosine, & Dallery, 2017; Poling, Weeden,
Address correspondence to: Holly A. Seniuk, Behavior Redner, & Foster, 2011), basketball (Alferink,
Analyst Certification Board, 7950 Shaffer Pkwy., Littleton,
CO 80127. Email: [email protected], [email protected] Critchfield, & Hitt, 2009; Romanowich,
doi: 10.1002/jaba.653 Bourret, & Vollmer, 2007; Schenk & Reed, 2019;
© 2019 Society for the Experimental Analysis of Behavior
1
2 HOLLY A. SENIUK et al.

Vollmer & Bourret, 2000), football (Critchfield & Research on choice in general is of interest
Stilling, 2015; Falligant, Boomhower, & Pence, to behavior analysts because of the contribution it
2016; Reed, Critchfield, & Martens, 2006; could make to our understanding of behavior
Stilling & Critchfield, 2010), and hockey (Seniuk, under certain circumstances and improve predic-
Williams, Reed, & Wright, 2015). These studies tion and control of such behavior. Sports present
have demonstrated that the GME describes choice an area where we can apply that understanding to
in athletic competition (Critchfield & Reed, 2009; make meaningful improvements in performance.
Reed & Kaplan, 2011) and have advanced Imagine a scenario where a coach is training a nov-
research on the matching law by focusing on using ice fighter to engage in more variable responding
interpretations of sensitivity (i.e., overmatching, in order to reduce their opponent’s ability to pre-
undermatching, and bias1) to explain the role of dict their next move. The study of an individual
sport where choices are made rapidly based on
specific contextual variables termed explanatory
opponent responding has the potential to inform
flexibility (Stilling & Critchfield, 2010), and how
the coach about how reinforcement influences
matching may predict athletic success (Alferink
response allocation, how choice may be influenced
et al., 2009; Seniuk et al., 2015).
by contextual variables, and whether a fighter’s
Research in this area is an important demonstra-
responding suggests a bias. The trainer could then
tion of applying behavioral evaluations to competi-
use this information to guide their intervention
tive group performance. However, it is a challenge
selection to target increasing variable responding.
to translate the basic research to application in this
See Harding, Wacker, Berg, Rick, and Lee (2004)
area, and the literature on matching law applica-
for an example of a behavior-analytic intervention
tions in the realm of sports has primarily focused
to promote response variability in martial arts
on team sport contexts that are more discrete-trial
training.
in nature (e.g., baseball and football). For example,
Mixed martial arts (MMA) is a uniquely quali-
in baseball a turn at bat resembles a discrete trial in
fied individual sport for a GME application because
that the antecedent of a pitcher throwing a pitch
of the dynamic nature of the sport wherein behavior
evokes the behavior of the batter swinging (or not)
is primarily evoked and reinforced directly by the
at the ball which is then followed by the conse-
behavior of the opponent, and could change pat-
quence of either hitting or missing the ball. Each
terns in responding very quickly. Specifically,
pitch is an opportunity to swing and has the
MMA2 is a diverse form of combat sport that com-
potential to produce a measureable outcome
bines various fighting styles, most notably jiu-jitsu,
(Poling et al., 2011). A few studies have applied
judo, kickboxing, and wrestling (La Bounty,
the GME to individual athlete behavior (Alferink
Campbell, Galvan, Cooke, & Antonio, 2011).
et al., 2009; Poling et al., 2011; Seniuk et al.,
Fighting techniques used in MMA are typically
2015) in the context of a team sport. However,
classified under one of three categories: kicks,
no prior studies have examined whether the
punches, and grappling (Kim, 2010). In the current
matching law can account for response allocation
study, strike selection (i.e., location on the
in an individual sport.
2
Rounds in an MMA match are 5 min in duration,
1
Overmatching refers to when the slope is greater than each match consists of 1 to 5 rounds. Matches end when
1 indicating that the organism is allocating more responses 1) the allotted amount of time has elapsed, 2) the referee
than necessary to obtain reinforcement. Undermatching stops the match (e.g., if a competitor can no longer
refers to when the slope is less than 1, indicating the defend himself), 3) a verbal submission or tap out, 4) a
organism is allocating fewer responses than necessary. Bias medical doctor is required, or 5) a knockout. When a
is a deviation from matching where the organism shows a fight is ended due to the allotted amount of time elapsing
preference for one alternative over the other(s) (Reed & victory is decided based on a panel of three judges
Kaplan, 2011). (combatsportslaw.com)
 MATCHING LAW AND MIXED MARTIAL ARTS 3

opponent’s body) is examined as a function of land- moved downward from there, eliminating
ing significant strikes among fighters from various fighters from the analysis who did not meet the
weight classes in the Ultimate Fighting Champion- conditions above. Fighters with at least five fights
ship (UFC). were selected to avoid misleading results due to
The study of martial arts is not new to limited data. Data were collapsed across multiple
behavior analysis. In 2004, Harding et al. dem- fights for each fighter (average of eight fights per
onstrated that differential reinforcement and fighter).
extinction were effective in promoting both The eight weight classes studied were: Heavy-
response variability and stimulus generalization weight (over 84 kg), Middleweight (over 77 up
in martial arts fighters. Additionally, Benitez to 84 kg), Welterweight (over 70 up to 77 kg),
Santiago and Miltenberger (2016) used video- Lightweight (over 66 up to 70 kg), Feather-
feedback to improve capoeira (i.e., an Afro- weight (over 61 up to 66 kg), Bantamweight
Brazilian form of martial arts) skills. The purpose (over 57 up to 61 kg), and Flyweight (over
of the present study was to expand the literature 52 up to 57 kg).
on the matching law by applying it to an individ-
ual sport and improve our understanding of
choice in a dynamic and fast-paced free-operant Data Collection
context (i.e., during any given fight the behavior All data on fighter rankings were obtained
of striking can be emitted at any time, and over a from the UFC website (ufc.com). Individual
wide range of response rates; Cooper, Heron, & fight data for each fighter were retrieved from
Heward, 2007). FightMetric, the official statistics system of the
UFC, which utilizes scientific methods to ana-
lyze MMA bouts across 67 statistical categories,
METHOD
encompassing all strikes, takedowns, submis-
Participants sions, and position changes (fightmetric.com).
In this retrospective study a total of 37 fighters The second author of this study input the data
(listed in Table 1 in order of rank) competing in from the FightMetric website into a Microsoft
the UFC between 2011 and 2015 were included. Excel® spreadsheet and applied Equation 1 to
Each ranked within the top 15 of their the data for each fighter. The first author then
corresponding weight class. This date range was verified that the equation was input correctly for
selected to include the largest number of fights each fighter. FightMetric does not provide their
per fighter possible for this analysis. When specific analyses for how strikes (e.g., punch or
selecting fighters for analysis, the belt holder in kick), takedowns (i.e., off-balancing an opponent
each individual weight class was selected first, to bring them to the ground), submissions
followed by at least three other fighters of varying (i.e., yielding to an opponent), and position
ranks from the UFC rankings. Selection was changes (i.e., advancing to a more dominant
based on the stipulation that fighters must have position, for example moving from a back mount
had five or more fights since the beginning of position to a full mount position) are defined
their professional MMA career that (a) went to at and scored, which limits further manipulation of
least the third round in a nontitle fight, (b) at specific data for different analyses (see Discus-
least the fifth round in a title fight, or (c) went sion). However, the archival data does provide
the distance (i.e., the maximum number of enough information for analyses that would not
rounds). When selecting fighters, the authors otherwise be possible. This analysis includes data
started with fighters ranked at the top of their on the number of significant strikes attempted
weight class (based on the UFC rankings) and (behavior), and landed (reinforcer) to each of
4 HOLLY A. SENIUK et al.

Table 1 Middleweight Division

Slope, Bias, and R2 Values When Comparing Strike Fighter a value b value R2 value
Selection (Head, Body, Legs) as a Function of Chris Weidman (G)* 0.9896 -0.0441 0.9324
Reinforcement Frequency (Landing a Significant Strike) Jacare Ronaldo Souza (S) 1.1285 -0.0069 0.9512
Vitor Belfort (G) 1.1351 0.0298 0.8321
Heavyweight Division Dan Henderson (S)* 1.1749 0.0019 0.8070
Fighter a value b value R2 value
Note. *Indicates that there is a 10 or fewer percentage
Fabricio Werdum (S) 0.8021 -0.0872 0.8386 points difference between striking and grappling accuracy.
Cain Velasquez (S) 1.1739 0.0357 0.9824
Junior Dos Santos (G) 1.1339 0.0197 0.9874
Frank Mir (S)* 1.0775 -0.0334 0.6365
Alistair Overeem (S) 1.0533 0.0075 0.9335 the head, body, and legs. FightMetric defines a
Mark Hunt (G) 1.1061 0.0089 0.9385 significant strike as any strike at distance and
Lightweight Division power strikes in the clinch and on the ground.
Fighter a value b value R2 value Short strikes in the clinch (i.e., standing up)
Rafael dos Anjos (S)* 1.2857 0.0567 0.8872 and on the ground are not included (Kuhn &
Anthony Pettis (G)* 1.1764 0.0272 0.8872
Donald Cerrone (S) 0.8999 -0.0639 0.4318 Crigger, 2013).
Khabib Nurmagomedov (S)* 0.9569 -0.0675 0.8382
Benson Henderson (E) 0.6773 -0.181 0.6513
Myles Jury (G) 0.9416 -0.0254 0.7727 Data Analysis
Featherweight Division
The equation used was:
Fighter a value b value R2 value
Jose Aldo (G) 1.1298 0.0707 0.9519    
B1 R1
Chad Mendes (G)* 1.0666 0.002 0.9096 log = s
log + logb ð1Þ
Nik Lentz (S) 1.2245 0.049 0.9647 B2 + B3 R2 + R3
Charles Oliveira (S)* 1.3777 0.0873 0.8821
Ricardo Lamas (S) 0.5416 -0.2002 0.1653
Bantamweight Division This equation has been previously used in
Fighter a value b value R2 value research where the matching law is applied to
TJ Dillashaw (S)* 1.4177 0.0878 0.9291 more than two alternatives (see Kangas, Berry,
Takeya Mizugaki (G) 1.3046 0.0612 0.9748
Urijah Faber (S)* 1.0777 -0.0119 0.9264
Cassidy, Dallery, Vaidya & Hackenberg, 2009;
Bryan Caraway (S)* 1.1843 0.0444 0.9222 Seniuk et al., 2015). In this equation, B represents
Welterweight Division the strike location (head, body, legs) and
Fighter a value b value R2 value R represents reinforcement, which was defined as
Robbie Lawler (G) 1.1128 0.0247 0.9573 landing a significant strike. In cases where there
Rory MacDonald (G)* 1.1244 0.0231 0.9440
Jake Ellenberger (G)* 1.4129 0.0892 0.9695
were no significant strikes landed during a match
Rick Story (G)* 1.3231 0.0464 0.8540 (i.e., the calculation is not possible with a denomi-
Light Heavyweight Division nator of zero) the same procedure as Ecott and
R2 Critchfield (2004), and Seniuk et al., (2015) was
Fighter a value b value value
used. This procedure involves taking the sum of
Daniel Cormier (S)* 1.2042 0.0401 0.9788 each type of strike attempted or landed across all
Anthony Johnson (G)* 0.9587 -0.0172 0.9128
Alexander Gustafson (E) 1.2574 0.0524 0.9509 fights and multiplying it by 0.01. This value
Antonio Rogerio Nogueira 1.0888 0.0092 0.8385 becomes a constant, which is added to each strike
(G)
value for each fight. Zeroes were only encountered
Flyweight Division
Fighter a value b value R2 value
in the dataset an average of one fight per fighter
analyzed.
Demetrious Johnson (G)* 1.2713 0.0645 0.7062
Joseph Benavidez (S)* 1.5677 0.1293 0.7602 The procedures used for creating the spread-
Ian McCall (S)* 1.3267 0.0507 0.5954 sheet in Microsoft Excel®, entering the data,
Brad Pickett (G) 1.2988 0.0305 0.8943
and conducting the linear regression were based
 MATCHING LAW AND MIXED MARTIAL ARTS 5

on those outlined in Reed (2009; see ERRA- RESULTS

TUM, 2011).3 In order to conduct the linear Table 1 summarizes the relation between
regression, the equation was first applied by com- strike location on opponent (i.e., head, body, leg)
paring a) head strikes to the sum of body and leg and landing significant strike ratios for all
strikes, b) body strikes to the sum of head and 37 fighters, organized by weight class. Fighters
leg strikes, and c) leg strikes to the sum of head are ordered based on their relative rank and each
and body strikes. A composite, that included the fighter is classified as either primarily striker (S),
combining of all comparisons above, was then primarily grappler (G), or equal accuracy (E) as
used for the linear regression to fit the equation identified through UFC.com. Classification was
to the data (Seniuk et al., 2015). based on their percentage of accuracy in striking
Additional analyses were conducted based on or grappling. An asterisk indicates that the per-
Cox et al. (2017). The purpose of this analysis centage point difference between striking and
was to calculate the minimum and maximum grappling accuracy was equal to or less than 10.
proportions of responses that could potentially Visual inspection of this information in relation
occur given the actual responses and reinforce- to sensitivity and bias indicates no systematic
ments observed (Herrnstein, 1970; Equations 2 trends. The results in Table 1 present the slope
and 3). (a) value, bias (b) value, and the variance (R2) for
each fighter. The GME was able to account for
RL
ð2Þ at least 80% of strike selection in 29 of 37 fighters
PL + P R included in the analysis, and more than half of
PL + PR −R R the variance for all but two fighters. The GME
ð3Þ described response allocation for all fighters with
PL + P R
78% showing overmatching (a > 1).
Typically, wider ranges indicate that the Figure 1 shows frequency distributions of
matching equation can provide valuable infor- slope and bias values. All values, taken from
mation about an organism’s response allocation Table 1, were rounded to the first decimal (a),
and smaller ranges suggest that any observed or second decimal (b) given the wide variety in
matching relations lack empirical validity. Simi- values. Slope values ranged from 0.5 to 1.6, with
lar to Cox et al. (2017) the response with the the most frequently observed value being 1.1. It
largest number of observed events was selected, should be noted that the bias values represent
thus the proportion of all strikes to the head to the composite bias score. The most frequent
all strikes was plotted as a function of the pro- values were 0, 0.02, 0.03, 0.04, 0.05, 0.06, and
portion of reinforcement for strikes to the head 0.09, each with a frequency of three. Figure 2
to all reinforcements obtained. Based on represents the best, median, and worst fit of the
Herrnstein’s (1970) Equations 2 and 3 a range fighters to the matching equation.4
of all possible responses was calculated for each Given that there were three alternatives in the
of the head strike proportions. equation, the implications of the composite bias
provided little valuable information. Therefore,
Table 2 presents the bias values for each behavior
for each fighter. This table provides an example
of how the free parameters can be individually
3
One deviation to this procedure included that the analyzed in multiresponse situations. That is, for
equations log(B1 / SUM (B2 + B3)) and log(R1 / SUM
(R2 + R3)) were entered into Excel® to account for three 4
Data are available for all 37 fighters upon request to
alternatives instead of two. the first author.
6 HOLLY A. SENIUK et al.

14

12

10
Number of Fighters

0.5 0.7 0.8 0.9 0.96 0.99 1.1 1.2 1.3 1.4 1.6
Slope Value

4
Number of Fighters

0
-0.18
-0.09
-0.07
-0.06
-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.09
0.13
0.2

Bias Value

Figure 1. The top panel shows the frequency distribu-

tion of slope (a) values for all fighters rounded to the nearest
hundredth and the bottom panel shows composite bias (b)
values for all fighters rounded to nearest hundredth.

each fighter head shots were compared to body

and legs, body to head and legs, and legs to head
Figure 2. Highest, median, and lowest MMA fighters’
and body to get a more accurate representation of matching plots. Top panel (highest matching), middle
bias. For all fighters the b values for head strikes panel (median matching), bottom panel (lowest matching).
compared to body and leg were greater than one, Each fighter’s record at the time of data collection is also
and all b values for body versus head and leg, and included.
leg versus head and body were less than one.
These values suggest a clear bias in favor of strikes be related to the observation that the vast majority
to the head for all fighters. of fighters demonstrated overmatching. However,
Figure 3 shows the results of the application of visual inspection did not reveal distinct differences
Herrnstein’s (1970) Equation 2 and Equation 3. in this pattern between fighters that demonstrated
Similar to the results of Cox et al., (2017) there overmatching versus those who did not.
were large ranges between minimum and maxi-
mum responding for all fighters. There is one
important difference to note between these results
and those of Cox and colleagues. The results of DISCUSSION
this study show that the data points for the ratio The purpose of the present study was to expand
between the proportion of head strikes to all the literature on the matching law by applying it
strikes and the proportion of reinforcement for to an individual sport and improve our under-
head strikes to reinforcement for all strikes tend to standing of choice in a dynamic and fast-paced
map on to the upper limit of the range. This may free-operant context. The results suggest that the
 MATCHING LAW AND MIXED MARTIAL ARTS 7

GME was a good descriptor of strike selection for to the head) based on the reinforcer most valu-
approximately 78% of the fighters. Eight (22%) of able or most likely to lead to the desired out-
the 37 fighters displayed R2 values below 0.80. An come (e.g., movement of the opponent’s hands
analysis of primary skill (i.e., striking vs. grappling) toward the face to block exposing the body).
and ranking showed no relationship between skill However, this bias may change later in the fight
or ranking and R2. Therefore there is no clear as the fighter observes that the opponent is weak
explanation for this finding relating to skill level or at blocking strikes to the legs. This type of anal-
fighting style. However, given the complex and ysis would be particularly interesting if compar-
dynamic nonlaboratory setting through which ing striking to takedown techniques given the
these data were collected this finding is not neces- difference in topography of the two behaviors.
sarily surprising. Matching experiments conducted The results for this study show a general bias for
with animals in controlled laboratory settings are most fighters, however the data available do not
convenient in the fact that animal behavior in allow for more fine-grained analyses that may
choice situations is often orderly and predictable provide more insight into how bias is or is not
(Mazur, 2013). Given the multitude of potential influenced by variables such as the opponent or
confounding variables in studies of human behav- the results of a previous fight.
ior in natural settings some exceptions to the over- The dataset available resulted in a number of
all findings are likely. limitations to aspects of the analyses in this
A notable finding is that a large proportion study that could be important. Although archival
of the sample displayed overmatching. While datasets are often large (i.e., a benefit), the data
examination of skills and ranking suggested no required for more nuanced analyses may not be
relationship between these variables and over- available (i.e., drawback). Coaches can benefit
matching, there are a few possible explanations. from the analysis of archival data to inform train-
One explanation might be the quality of the ing. For example, a coach may recommend that a
reinforcer used in this study (i.e., a significant fighter focus on using a particular strike more
strike landed), which was selected based on the often based on how frequently they land that
available data. However, in an actual fight situ- type of strike. However, the collection of live
ation the function of striking may change data may lead to more in-depth analyses that
within or between rounds based on fighter strat- would further our understanding of choice in dif-
egy. For example, in some situations landing a ferent contexts and lead to the development of
strike may be preferred, while in other situations applications of matching that could be easily
the function of a particular strike may be to “fake implemented by coaches and trainers. Baum
out” the opponent to obtain access to another, (2010) suggests various analyses that could meet
possibly more sensitive, area of the body. Another this objective. For example, how behavior and
example includes instances where the function of reinforcement change as a function of the feed-
the strike is to gather information about the back from the opponent. A possible application
topography or other features of the opponent’s of this type of analysis would involve a round-by-
response that may be valuable. round analysis to specifically identify the out-
Changes in reinforcer type or quality may also come associated with each strike. As mentioned
influence bias throughout a fight. The results of previously, the data from FightMetric only pro-
this study suggest that fighters generally showed vided data on landing a strike. Data from direct
a bias toward strikes to the head as shown observation would allow for other reinforcers to
through the breakdown of free parameters by be considered. For example, whether the
response type in Table 2. A fighter may exhibit response of the opponent opens the opportunity
a bias toward a particular alternative (e.g., strike for a different more effective strike, or evokes a
8 HOLLY A. SENIUK et al.

Table 2 Middleweight Division

Isolated Bias Values for Each Fighter Comparing Each Fighter Head Body Legs
Type of Strike to the Alternatives Chris Weidman (G)* 0.4495 -0.4566 -0.3135
Jacare Ronaldo Souza (S) 0.3243 -0.3271 -0.4564
Heavyweight Division Vitor Belfort (G) 0.2575 -0.313 -0.5125
Fighter Head Body Legs Dan Henderson (S)* 0.4515 -0.5203 -0.4077
Fabricio Werdum (S) 0.238 -0.1973 -0.2336 Note. *Indicates that there is a 10 or fewer percentage
Cain Velasquez (S) 0.2505 -0.2754 -0.2396
Junior Dos Santos (G) 0.2302 -0.2486 -0.2169
points difference between striking and grappling accuracy.
Frank Mir (S)* 0.5355 -0.4412 -0.5948
Alistair Overeem (S) 0.1608 -0.1662 -0.1975
Mark Hunt (G) 0.6186 -0.8142 -0.3163
Lightweight Division response from the opponent that provides infor-
Fighter Head Body Legs mation about defensive weaknesses.
Rafael dos Anjos (S)* 0.318 -0.2758 -0.3151 The use of direct observation data would open
Anthony Pettis (G)* 0.4004 -0.4416 -0.4692 the doors for other, more detailed, analyses such
Donald Cerrone (S) 0.3659 -0.1677 -0.512
Khabib Nurmagomedov (S)* 0.6544 -0.9225 -0.7434 as strike type and sequencing. The MMA sport is
Benson Henderson (E) 0.3646 -0.4672 -0.4064 an eclectic form of martial arts where competitors
Myles Jury (G) 0.3817 -0.471 -0.4715
draw from their training in a variety of combat
Featherweight Division sports (e.g., jiu-jitsu, boxing, wrestling, muay
Fighter Head Body Legs
thai). This means that the topographies of strikes
Jose Aldo (G) 0.2069 -0.3068 -0.1883
Chad Mendes (G)* 0.284 -0.272 -0.2946 and takedown tactics vary both within and across
Nik Lentz (S) 0.2709 -0.4823 -0.2887 fighters. An interesting analysis would be the
Charles Oliveira (S)* 0.2949 -0.22 -0.4176
Ricardo Lamas (S) 0.308 -0.5427 -0.4666
investigation of how strike topography may
change within a fight (round by round), or over
Bantamweight Division
Fighter Head Body Legs the course of a fighter’s season or career. More-
TJ Dillashaw (S)* 0.4059 -0.602 -0.3297
over, the available data did not allow for an analy-
Takeya Mizugaki (G) 0.3215 -0.2798 -0.0124 sis of strike magnitude. Similar to how strike
Urijah Faber (S)* 0.3732 -0.2886 -0.785 topography may change as a function of reinforce-
Bryan Caraway (S)* 0.6065 -0.9115 -0.5709
ment, strike magnitude may also change as a func-
Welterweight Division
Fighter Head Body Legs tion of reinforcement and/or the context of the
Robbie Lawler (G) 0.3013 -0.2833 -0.5185
situation. MMA is unique from other sports
Rory MacDonald (G)* 0.3126 -0.313 -0.4836 examined in matching law studies in that ranking
Jake Ellenberger (G)* 0.3164 -0.4278 -0.417 has a subjective component. For example, fighters
Rick Story (G)* 0.419 -0.3889 -0.4044
who win by knockouts and submissions are con-
Light Heavyweight Division
Fighter Head Body Legs sidered more dominant and tend to climb the
rankings faster. Therefore, a fighter’s responding
Daniel Cormier (S)* 0.2792 -0.2389 -0.3035
Anthony Johnson (G)* 0.4469 -0.5718 -0.583 may be biased toward higher magnitude strikes if
Alexander Gustafson (E) 0.4222 -0.4861 -0.5067 their goal is to increase their ranking through
Antonio Rogerio Nogueira (G) 0.4972 -0.5924 -0.9939
knockouts. The application of Baum’s (2010)
Flyweight Division
Fighter Head Body Legs
analysis of preference pulses may provide insight
in this area.
Demetrious Johnson (G)* 0.2374 -0.2209 -0.2633
Joseph Benavidez (S)* 0.4268 -0.5562 -0.7688 Another limitation of this study is that only
Ian McCall (S)* 0.4018 -0.4225 -0.4611 strikes were analyzed and takedowns were not
Brad Pickett (G) 0.5369 -0.5315 -0.5563
included. However, the focus for this study was
 MATCHING LAW AND MIXED MARTIAL ARTS 9

1 1
Takeya Mizugaki Jose Aldo
0.8 0.8
Matching
relation
0.6 between head 0.6
strikes and all
strikes
0.4 0.4

0.2 0.2
B1/B1+B2+B3

Min. and max.

possible strikes
0 0
0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1

1 1
Vitor Belfort
Anthony Pettis
0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
R1/R1+R2+R3

Figure 3. A representative sample of the results of the application of Herrnstein’s (1970) Equation 2 and Equation 3.
Horizontal black lines represent the minimum and maximum number of stikes that a fighter could emit during a fight.
Closed black circles represent the proportion of head strikes to all strikes in relation to the proportion of reinforcement
for head strikes to reinforcement for all strikes for each fight. The dashed black line depicts “perfect matching” where
sensitivity is equal to 1 and bias is equal to zero.

on strikes as they tend to occur much more fre- “the same empirical and conceptual tools apply
quently within a fight and the data available equally to all levels of analysis, from laboratory
indicate various body targets for strikes. An to field” (Critchfield & Reed, 2009, pg. 1),
interesting comparison might be an analysis of which has led many behavior analysts to
strikes versus takedowns. Similarly, the data emphasize the importance of quantitative ana-
used for the present study did not allow for an lyses in both basic and applied behavior analy-
analysis of strike topography. Again, data col- sis (see Critchfield & Reed, 2009; Nevin,
lected through direct observation may allow for 2008; Waltz & Follette, 2009). If quantitative
such analyses to be conducted. analyses as the mainstream of basic behavior
Although there clearly are limitations to these analysis (Nevin, 2008) are to inform applied
data, because the source was not based on direct research, their generality to everyday behaviors
observation of each fight, this study does dem- is of great importance.
onstrate that the matching law is highly applica- In conclusion, the present study represents an
ble to MMA fighting. Our hope is that it serves application of the matching law to an individual
as a springboard for other researchers interested sport. The data are orderly at the level of indi-
in pursuing work in this area. In his translational vidual findings, which is a significant finding.
writings, Skinner (e.g. 1953, 1957) argued that Behavior analysis in the context of sports is a
10 HOLLY A. SENIUK et al.

small but growing area of practice and research and. Practice, 15, 112–121. https://doi.org/10.1037/
in our profession—and behavior analysts have bar0000011.
Ecott, C. L., & Critchfield, T. S. (2004). Noncontingent
the tools to make an impact in this area. By reinforcement, alternative reinforcement, and the
understanding the complex contingencies that matching law: A laboratory demonstration. Journal of
are operating during athletic performance at the Applied Behavior Analysis, 37, 249–265. https://doi.
org/10.1901/jaba.2004.37-249.
individual level we may identify behavior change ERRATUM. (2011). Journal of Applied Behavior Analysis,
strategies that could be used to improve athletic 44, 132. https://doi.org/10.1901/jaba.2011.44-132.
performance. Given the high value that sports Falligant, J. M., Boomhower, S. R., & Pence, S. T.
(2016). Application of the generalized matching law
play in our society, such behavior-analytic appli- to point-after-touchdown conversions and kicker selec-
cations could have broad-reaching effects. tion in college football. Psychology of Sport and Exercise,
26, 149–153. https://doi.org/10.1016/j.psychsport.
2016.07.006.
REFERENCES Fightmetric.com. Retrieved November 23, 2015, from
fightmetric.com website: http://fightmetric.com/
Alferink, L. A., Critchfield, T. S., & Hitt, J. L. (2009). statistics/events/completed
Generality of the matching law as a descriptor of shot Harding, J. W., Wacker, D. P., Berg, W. K., Rick, G., &
selection in basketball. Journal of Applied Behavior Lee, J. F. (2004). Promoting response variability and
Analysis, 42, 595–608. https://doi.org/10.1901/jaba. stimulus generalization in martial arts training. Journal
2009.42-595. of Applied Behavior Analysis, 37, 185–195. https://doi.
Bauder, D. (2011). Super Bowl 2011 is most watched program org/10.1901/jaba.2004.37-185.
ever. Retrieved from: http://www.huffingtonpost.com/ Herrnstein, R. J. (1961). Relative and absolute strength of
2011/02/07/super-bowl-2011-ratings-s_n_819559.html response as a function of frequency of reinforcement.
Baum, W. H. (1974). On two types of deviation from Journal of the Experimental Analysis of Behavior, 4,
the matching law: Bias and undermatching. Journal 267–272. https://doi.org/10.1901/jeab.1961.4-267.
of the Experimental Analysis of Behavior, 22, 231–242. Herrnstein, R. J. (1970). On the law of effect. Journal of
https://doi.org/10.1901.jeab.1974.22-231. the Experimental Analysis of Behavior, 13, 243–266.
Baum, W. H. (2010). Dynamics of choice: A tutorial. https://doi.org/10.1901.jeab.1970.13-243.
Journal of the Experimental Analysis of Behavior, 94, Kangas, B. D., Berry, M. S., Cassidy, R. N., Dallery, J.,
161–194. https://doi.org/10.1901/jeab.2010.94-161. Vaidya, M., & Hackenberg, T. D. (2009). Concur-
Benitez Santiago, A., & Miltenberger, R. G. (2016). rent performance in a three-alternative choice situation:
Using videofeedback to improve martial arts perfor- Response allocation in a Rock/Paper/Scissors game.
mance. Behavioral Interventions, 31, 12–27. https:// Behavioral Processes, 82(2), 164–172. https://doi.org/,
doi.org/10.1002/bin.1424. https://doi.org/10.1016/j.beproc.2009.06.004.
Cohen, T., & Muise, M. (2011). French debate rescheduled Kim, M. (2010). Mixed martial arts: the evolution of a com-
to avoid Habs playoff game. Retrieved from: https:// bat sport and its laws and regulations. Sports Lawyers
nationalpost.com/sports/nhl/french-debate-rescheduled- Journal, 17, 49–70.
to-avoid-habs-playoff-gamecombatsportslaw.com. Kuhn, R., & Crigger, K. (2013). Fightnomics: The hidden
Retrieved July 10, 2019, from combatsportslaw.com numbers in mixed martial arts and why there’s no such
website: https://combatsportslaw.com/2018/11/02/the- thing as a fair fight. Cedar Rapids, IA: Graybeard
current-version-of-the-abcs-unified-rules-of-mma/ Publishing Co.
Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). La Bounty, P., Campbell, B. I., Galvan, E.,
Applied behavior analysis (2nd ed.). Upper Saddle Cooke, M., & Antonio, J. (2011). Strength and con-
River, NJ: Pearson Education Inc. ditioning considerations for mixed martial arts.
Cox, D. J., Sosine, J., & Dallery, J. (2017). Application Strength and Conditioning Journal, 33, 56–67. https://
of the matching law to pitch selection in professional doi.org/10.1519/SSC.0b013e3182044304.
baseball. Journal of Applied Behavior Analysis, 50, Mazur, J. E. (2013). Learning and behavior (7th ed.).
393–406. https://doi.org/doi, https://doi.org/10. New York, NY: Routledge.
1002/jaba.381. Neilsenwire (2012). Final presidential debate draws 59.2
Critchfield, T. S., & Reed, D. D. (2009). What are we million viewers. Retrieved from http://blog.nielsen.
doing when we translate from quantitative models? com/nielsenwire/politics/final-presidential-debate-
The Behavior Analyst, 2, 339–362. https://doi.org/10. draws-59-2-million-viewers/
1007/bf03392197. Nevin, J. A. (2008). Control, prediction, order, and the
Critchfield, T. S., & Stilling, S. T. (2015). A matching joys of research. Journal of the Experimental Analysis
law analysis of risk tolerance and gain-loss framing in of Behavior, 89, 119–123. https://doi.org/10.1901/
football play selection. Behavior Analysis: Research jeab.2008.89-119.
 MATCHING LAW AND MIXED MARTIAL ARTS 11

Poling, A., Weeden, M. A., Redner, R., & Foster, M. Analysis: Research and Practice, 15(3-4), 152–160.
(2011). Switch hitting in baseball: Apparent rule-follow- https://doi.org/10.1037/bar0000019.
ing, not matching. Journal of the Experimental Analysis of Skinner, B. F. (1953). Science and human behavior.
Behavior, 96, 283–289. https://doi.org/10.1901/jeab. New York, NY: Macmillan.
2011.96-283. Skinner, B. F. (1957). Verbal behavior. New York, NY:
Reed, D. D. (2009). Using Microsoft Excel® 2007 to Appleton-Century-Crofts.
conduct generalized matching analyses. Journal of Stilling, S. T., & Critchfield, T. S. (2010). The matching
Applied Behavior Analysis, 4, 867–875. https://doi. relation and situation-specific bias modulation in pro-
org/10.1901/jaba.2009.42-867. fessional football play selection. Journal of the Experi-
Reed, D. D., Critchfield, T. S., & Martens, B. K. (2006). mental Analysis of Behavior, 93, 435–454. https://doi.
The generalized matching law in elite sport competi- org/10.1901/jeab.2010.93-435ufc.com. Retrieved
tion: Football play calling as operant choice. Journal November 23, 2015, from, ufc.com website, http://
of Applied Behavior Analysis, 39, 281–297. https:// www.ufc.com/rankings.
doi.org/10.1901/jaba.2006.146-05.
United Nations (2005). International year of sport and
Reed, D. D., & Kaplan, B. A. (2011). The matching law:
physical education. Retrieved from http://www.un.org/
A tutorial for practitioners. Behavior Analysis in Prac-
sport2005/
tice, 4(2), 15–24. http://10.1007/BF03391780.
Romanowich, P., Bourret, J., & Vollmer, T. R. (2007). Vollmer, T. R., & Bourret, J. (2000). An application of
Further analysis of the matching law to describe the matching law to evaluate the allocation of two-
two- and three-point shot allocation by professional and three-point shots by college basketball players.
basketball players. Journal of Applied Behavior Journal of Applied Behavior Analysis, 33, 137–150.
Analysis, 40, 311–315. http://dx/doi.org/10.1901/ https://doi.org/10.1901/jaba.2000.33-137.
jaba2007.119-05. Waltz, T. J., & Follette, W. C. (2009). Molar functional
Schenk, M. J., & Reed, D. D. (2019). Experimental eval- relations and clinical behavior analysis: Implications
uation of matching via a commercially available bas- for assessment and treatment. The Behavior Analyst,
ketball video game. Journal of Applied Behavior 32, 51–68. https://doi.org/10.1007/bf03392175.
Analysis. https://doi.org/10.1002/jaba.551.
Seniuk, H., Williams, W. L., Reed, D. D., & Wright, J. W. Received September 7, 2018
(2015). An examination of matching with multiple Final acceptance July 19, 2019
response alternatives in professional hockey. Behavior Action Editor, Derek Reed