Comparison of Body Image Perceptions for Female Competitive Dancers, Fitness Cohort, and Non-Dancers in a College Population

Submitted by Lorraine Killion, Ed.D. & Dean Culpepper, Ph.D.

ABSTRACT
Body image is a complex synthesis of psychophysical elements that are perpetual, emotional, cognitive, and kinesthetic (1). The desire to achieve and maintain an ideal weight is a prevalent goal among females. The purpose of this study was to examine a female population of competitive dancers, control, and fitness cohorts’ body image and eating characteristics. A total of 51 (29 dancers, 12 control, and 10 fitness) subjects completed the MBSRQ-AS, EAT-26, a Physical Activity Questionnaire, Stunkard Figural Silhouettes, and body fat measurements. A MANOVA was conducted to determine group differences and showed a significant relation (Wilk’s Lambda = .106, F=8.735, p<001). Post hoc tests were conducted to determine directionality and showed that the dancers scored significantly higher on the Appearance Orientation subscale (p=.034) with no difference between the control and fitness cohort. Dancers also significantly perceived themselves to be overweight (p=.048) with no difference between the other two groups. Both the dancers (p<.001) and the fitness cohort (p<.001) scored as exhibiting disordered eating patterns as rated by the EAT-26. Even though the dancers had a low percent body fat (M=17.6), they tended to place more importance on how they look. The dancers perceived themselves to be overweight and engaged in disordered eating patterns. These types of perceptions and behaviors are disturbing, but not surprising since dancers have a drive for thinness to compete (2). To fully understand the scope of the issue and the psychological factors that accompany the quest for achieving a certain appearance, future research should include other female cohorts such as elite athletes, obligatory exercisers, and sedentary females to determine any similarities and differences in the groups.

INTRODUCTION
Research has documented and quantified a shift towards a thinner ideal shape for females in the Western culture for the past 20 years (3). Body image has been shown in numerous studies to be a key issue for females. Body image has been described as a multidimensional construct that describes internal, subjective representations of physical and bodily appearance (4). The internal representations of one’s own body include both cognitive and perceptual elements (5). In addition, eating disorders have been shown to be prevalent in females with more than 90 percent of those with eating disorders are women between the ages of 12 and 25 years of age (6, 7, 8). Research indicates that both of these factors (body image and eating disorders) are present among elite performers of certain sports or physical activities, ballet dancers, and professional dancers (8). Yet little has been reported on dance team participants (9, 10, 11).

Dance team is difficult to research due to the paucity of literature available and the complexity of terminology. Also, dance team is a nebulous term to define. Research demonstrates common referrals to spirit teams, spirit squads, dance teams, as well as pom squads. While the confusion in labeling and current argument as to whether this is an activity or a sport still looms, one fact that remains constant is competitive spirit teams is one of the fastest growing areas of participation for females (12).

Among high school participants, over 96,718 females were accounted for in the 2010-2011 high school athletics participation survey conducted by the National Federation of State High School Associations, ranking competitive spirit teams ninth for female participation. At the college level, the National Collegiate Athletic Association (NCAA) reported that spirit squad has experienced the most growth for women’s sport (13, 14). A nationwide Division I study conducted during the 2001-02 academic school year investigated the prevalence of dance and cheerleading programs and reported 89% of the institutions contacted indicated they sponsored competitive dance (12).

The current emerging phenomenon of dance teams has witnessed the rise in visibility of participants at sporting events and are known for their pre-game and half-time routines. Dance teams are comprised of competitive dancers who are required to practice for long hours in movements, choreography, and synchronicity among dancers. Participants are also required to incorporate specific choreography (i.e., contemporary, hip-hop, or jazz) and technical skills (jumps, kicks, and other gymnastic-type skills) into the routine. It is highly competitive and requires hours of rehearsal to master precise movements in harmony with other members of the team.

The increasing number of females participating in dance team competition is prevalent. Long rehearsal hours, use of mirrors, and dance outfits, place dance team participants at risk of body image concerns (15, 16, 17, 18). Of additional concern is the presence of wearing dance outfits which possibly place them as subjects of objectification, or being evaluated by gazing or being observed or “checked out” on the basis of their appearance (17, 19, 10).

With the growing number of females participating in dance team competition, a further examination of the psychosocial factors that accompany this new sport warrants investigation including the importance of assessing potential body image disturbance. This study was designed to examine the perceptions of dance team participants, fitness participants, and non-dancers in a college population.

METHODS
Upon Internal Review Board (IRB) approval, fifty one subjects were recruited from two university campuses. Informed consent was obtained prior to the study through an information letter that was administered to participants in dance and physical fitness classes.

Participants
Participants were female students enrolled in university classes and dance teams. Two university campuses were involved in the study and yielded a total of 51 participants. The study was comprised of 29 dancers, 10 fitness students, and 12 control subjects. The mean age and standard deviation for the participants were: dancers (M = 20.69, SD = 2.25), fitness (M = 25.40, SD = 8.67), and control (M = 20.42, SD = 0.996). The dancers were from university dance teams, the fitness participants were enrolled in fitness classes, and the participants in the control group were randomly selected from general university courses.

Instruments
Each subject completed questionnaires assessing participant demographics, physical activity involvement using the NASA Physical Activity Scale and body image perceptions using the Stunkard Figural Rating Silhouettes. Eating behavior patterns were assessed utilizing the Eating Attitudes Test (EAT-26) and attitudes concerning body image were assessed with the Multi- dimensional Body-Self Relations Questionnaire (MBSRQ). Anthropometric measurements (height and weight) were then taken. Weight was taken using a Tanita WB-110A Digital Scale and height was taken using a using a Seca 420 measuring stadiometer. Body fat measurements were taken on each participant using an Omron Fat Loss Monitor, Model HBF-306C. The Fat Loss Monitor (Omron Fat Loss Monitor, Model HBF-306C) displays the estimated value of body fat percentage by bioelectrical impedance method and indicates the Body Mass Index (BMI). The bioelectrical impedance, skinfold, and hydrostatic weighing methods have all been shown to be reliable measures of body composition (r = .957-.987). (23)

Eating Attitudes Test (EAT-26)
The Eating Attitudes Test (EAT-26) was used to differentiate participants with anorexia nervosa, bulimia nervosa, binge-eating, and those without disordered eating characteristics. It is a 26-item measurement consisting of three subscales: 1) dieting, 2) bulimia and food perception, and 3) oral control. Scoring for this instrument was a Likert scale of six possible answers (always, usually, often, sometimes, rarely, never). Scores ranged from zero to three for each question and a total score greater than 20 indicates excessive body image concern that may identify an eating disorder (20, 21). The EAT-26 has been proven to be a reliable (r =.88) measurement. (7)

Figural Rating Silhouettes
Body size judgments were obtained using the Stunkard Figure Rating Scale (see figure 1). This scale consists of a nine-figure scale of numbered silhouettes that increase gradually in size from very thin (a value of 1) to very obese (a value of 9). (22) Two body size perception variables were included in the current study. “Self-perceived body size” is the number of the figure selected by participants in response to the prompt “Choose the figure that reflects how you think you currently look.” “Ideal body size” is the number of the figure chosen in response to the prompt “Choose your ideal figure.” This scale has good test-retest reliability and adequate validity (23, 24). Following the methods of other investigators, we defined body size satisfaction as the difference between self-perceived body size and ideal body size (25, 26, 27, 28). A body size discrepancy index variable was created for each participant by subtracting the number of the figure selected as the ideal body size from the number of the figure selected as the self-perceived current body size (28). A high body size discrepancy value signifies low satisfaction with body size, and a low value signifies greater satisfaction with body size.

Multidimensional Body-Self Relations Questionnaire
The Multidimensional Body-Self Relations Questionnaire (MBSRQ) is a 69 item self-report inventory for the assessment of self-attitudinal aspects of the body image construct. The MBSRQ measures satisfaction and orientation with body appearance, fitness, and health. In addition to seven subscales (Appearance Evaluation and Orientation, Fitness Evaluation and Orientation, Health Evaluation and Orientation, and Illness Orientation), the MBSRQ has three special multi-item subscales: (1) The Body Areas Satisfaction Scale (BASS) approaches body image evaluation as dissatisfaction-satisfaction with body areas and attributes; 2) The Overweight Preoccupation Scale assesses fat anxiety, weight vigilance, dieting, and eating restraint; and 3) The Self-Classified Weight Scale assesses self-appraisals of weight from “very underweight” to “very overweight.” Internal consistency for MBSRQ subscales range from .74 -.91. This questionnaire has been studied and used extensively in the college population. Internal consistency for the subscales of the MBSRQ ranged from .67 to .85 for males and .71 to .86 for females (9).

Physical Activity Scale
Level of physical activity was obtained by self-report with the NASA Activity Scale (NAS) (29, 30). The scale enables subjects to rate their general activity behavior over the previous 30 days. The scale range is from 0 to 10, which is based on the total weekly minutes spent in exercise or the total weekly miles run or walked. A NAS of 0-1 represents very low activity. A rating of 2-3 represents regular recreation or work of modest effort in such activities as golf or yard work for a weekly total of between 30 min to 2 h. Ratings of 4-10 represent regular participation in aerobic exercise ranging from light to heavy exercise.

Procedures
The participants were instructed by a trained individual to fill out the information packets provided on clipboards. First, the participants completed a personal identification and demographic sheet that contained general information such as age and dance or sport category. The participants then completed the MBSRQ-AS, the EAT-26, Physical Activity Questionnaire, and the Stunkard Figural Rating Scale (31, 20, 29, 22). As the participants completed the written component of the study, another trained individual took height and weight measures of the participants and recorded the body mass index (BMI) from a hand-held BIA analyzer. Weight was taken using a Tanita WB-110A Digital Scale and height was taken using a using a Seca 420 measuring stadiometer. A test/retest method was utilized for both measures to offset measurement error. In the measure of weight, the individual’s weight was recorded, the participant stepped off the digital scale and the scale was returned to “zero”. The measure was then taken again and recorded. In the measure of height, the same procedure of test/retest was used. When all measures were taken, the average of the two measures was then recorded. The measures were then taken by the researchers and converted using the formula (BMI = weight/height M2). BMI was then calculated and recorded for all participants. When the information was completed, the participants returned the packets to the trained administrator. Data sheets were collected and kept in a locked file cabinet for confidentiality.
A total of 51 participants completed the MBSRQ-AS, EAT-26, a Physical Activity Questionnaire, Stunkard Figural Silhouettes, and body fat measurements. Descriptive statistics are presented in Table 1. The Dancers and the Fitness group were significantly lower in body fat and higher in physical activity and the on the EAT-26. A MANOVA was conducted to determine group differences among the different measures and the subscales.

Table 1 – Figure Rating Means for each Group (dancer, fitness, & control)
Screen Shot 2014-01-24 at 4.14.52 PM

RESULTS
The MANOVA indicated a significant relation (Wilk’s Lambda = .106, F = 8.735, p<.001). Post hoc tests were conducted and analyses were examined to determine directionality. Results showed that the dancers scored significantly higher on the Appearance Orientation subscale (p=.034) with no difference between the control and fitness cohort. Dancers also significantly perceived themselves to be overweight (p=.048) with no difference between the other two groups. Both the dancers (p<.001) and the fitness cohort (p<.001) scored as exhibiting disordered eating patterns as rated by the EAT-26 (see Table 2).
Even though the dancers had a low percent body fat (M=17.6), they tended to place more importance on how they look. Body dissatisfaction measures often focus on body build and are operationalized as the difference between ideal and self-perceived current figure as selected from a group of drawings (32, 33, 34). Measures of body dissatisfaction were computed by subtracting participants’ ratings of their Current Body Size (CBS) from their Ideal Body Size (IBS) to create a discrepancy index (DI). (28) The DI’s for each group were calculated with means and standard deviations recorded: Dancers (-.59/1.11), Fitness Group (-1.04/.966), and Control (-1.55/.85). The dancers in this study were dissatisfied with their bodies and wanted a thinner body as described in the discrepancy index, indicating a higher level of importance on their appearance (p=.045).

Table 2-Percent Fat and Eat-26 Totals for Subjects
Screen Shot 2014-01-24 at 4.15.08 PM

DISCUSSION
The primary focus of this investigation was to examine collegiate dance team participants to see if they exhibited body image distortions and disordered eating habits as exhibited in other female performers. Even though the dancers had a low percent body fat (M = 17.6), they tended to place more importance on how they look. The dancers perceived themselves to be overweight and engaged in disordered eating patterns. These types of perceptions and behaviors are disturbing, but not surprising since dancers have exhibited a drive for thinness to compete (2).

The findings of the data for this study are consistent with previous studies regarding body image in females (6, 35, 36). The females in this study perceived their current figure as heavier than their ideal figure. Although literature available on dancers exists, many of the studies have focused on ballet dancers and other professional dancer types. Future research should examine dance team participants to see if the pressures are similar (i.e., rehearsing with mirrors and being viewed during their performance by an audience). To fully understand the scope of the issue and the psychological factors that accompany the quest for achieving a certain appearance, future research should include other female cohorts such as elite athletes, obligatory exercisers, and sedentary females to determine any similarities and differences in the groups.

These results indicate that dancers had higher incidence of negative body image disturbances as compared with the controls. Dancers are usually expected to be slim, well-proportioned, and toned and are placed under a great deal of pressure to maintain these features. Often, the various aspects of a dance class can potentially lead to a negative body image (37). The pressures of being thin may present negative body images for dance team members (38). A national survey conducted reported that body image concerns continue to be prevalent among American women (39). Levels of body dissatisfaction may also foster negative affect because appearance is a central dimension for women in our culture (40).

While the dangers of distorted body image are present in the dance world, measures to minimize their impact should include coaches who focus on performance rather than personal appearance. Taking an active interest in how their dancers view themselves is critical to a more comprehensive understanding of the causes of body image concern. By further addressing this issue, researchers can also help minimize health risks in female participants as well as reduce body image dissatisfaction.

Limitations & Implications
Limitations to this study include the sample size. In addition, this study investigated indicators of disordered eating attitudes and behaviors rather than clinical diagnoses of eating disorders. Other variables that are contributing factors to the prevalence of disordered eating were not investigated. The results of the EAT-26 test were not intended to diagnose nor suggest an eating or life-threatening disorder; however, the EAT-26 was used because it has proven to be an effective screening tool in identifying eating disorder symptomology and allows for further investigation for treatment.

APPLICATIONS IN SPORT
Body image has been the subject of much research conducted in recent years. As a result, body image is now recognized a multidimensional construct with complex aspects, particularly perceptual. The majority of the existing data indicates that body image concerns are prevalent among American females. With the recent phenomenal growth of dance team participation and the increasing number of female participants; a closer examination is warranted. Yet, there is a paucity of research available on dance team participants and their perceptions of their body appearance. Because dance team members wear a designated uniform/outfit, dance to a learned synchronized routine, and perform in front of an audience, they are subjected to visual scrutinization of fans/viewers. The uniqueness of the stressors and demands placed on the dancers complicates this issue. Additional knowledge of how dance team members perceive how they look and what the audience thinks of them in regards to abilities and their physical appearance deserves further investigation. Dealing with such information will not only benefit dance team members body image and self-esteem, but assist coaches and directors in ways to assist young women in resulting body image dissatisfaction.

ACKNOWLEDGMENTS
None

REFERENCES
1. Thompson, K. (1999). Exacting beauty: Theory, assessment, and treatment of body image disturbance. Washington, DC: American Psychological Association.
2. Wood, K. C., Becker, J. A., & Thompson, J. K. (1996). Body image dissatisfaction in preadolescent children. Journal of Applied Developmental Psychology, 17, 85-100.
3. Garner, D. M., Garfinkel, P. E., Schwartz, D., and Thompson, M. (1980). Cultural
expectations of thinness in women. Psychological Reports, 47, 483-491.
4. Cash, T. F., & Pruzinsky, T. (1990). Body images: development, deviance, and
change. New York: NY, Guilford Press.
5. Cash, T. F., & Brown, T. A. (1987). Body image in anorexia nervosa and bulimia
nervosa: A review of the literature. Behavioral Modification, 11, 487-521.
6. Fallon, A. E. & Rozin, P. (1985). Sex differences in perceptions of desirable body
shape. Journal of Abnormal Psychology, 94, 102-105.
7. Garner, D. M. & Garfinkel, P. E. (1979). The Eating Attitudes Test: An index of
the symptoms of anorexia nervosa. Psychological Medicine, 9, 273-279.
8. National Alliance for the Mentally Ill (2003). Fact sheet,
http://www.ncsacw.samhsa.gov/files/
9. Sundgot-Borgen, J. (1993). Prevalence of eating disorders in elite female athletes.
International Journal of Sport Nutrition, 3, 29–40.
10. Tiggemann, M., and Slater, A. (2001). A test of objectivity theory in former
dancers and non-dancers. Psychology of Women Quarterly, 25, 1, 57-64.
11. Pierce, E. F., & Daleng, M. L. (1998). Distortion of body image among elite
female dancers. Perceptual and Motor Skills, 87, 3, 769-770.
12. Sowder, K., Hennefer, A., Pemberton, C., & Easterly, D. (2004). Defining
“Sport”. Athletic Management, 16.02, February/March.
13. National Federation of State High School Associations (http://www.nfhs.org/)
14. National Collegiate Athletic Association (NCAA) (http://www.ncaa.org/)
15. Carman, J. (2011). Passing on the Magic. Dance Magazine, 85, 12, 50-54.
16. Radell, S. A., Adame, D.D., & Cole, S.P. (2002). The effect of teaching with
mirrors on body image and locus of control in women college dancers: A pretest-
posttest study. Research Quarterly for Exercise and Sport, 74, 1, A-3.
17. Schneider, D. J. (1974). Effects of dress on self-presentation. Psychological
Reports, 35, 1, 167-170.
18. Fredrickson, B. L., Roberts, T., Noll, S. M., Quinn, D.M., & Twenge, J.M.
(1998). That swimsuit becomes you: Sex differences in self-objectification,
restrained eating, and math performance. Journal of Personality and Social
Psychology, 75, 1, 269-284.
19. Price, B. R., & Pettijohn, T. F. (2006). The effect of ballet dance attire on body
and self-perceptions of female dancers. Social Behavior and Personality, 34,
8, 991-998.
20. Garner, D. M., Olmsted, M. P., Bohr, Y., & Garfinkel, P. (1982). The eating
attitudes test: Psychometric features and clinical correlates. Psychological
Medicine, 12, 871 878.
21. Williamson, D. A., Davis, C. J., Goreczny, A. J., & Blouin, D. C. (1989). Body
image disturbances in bulimia nervosa: Influences of actual body size. Journal of
Abnormal Psychology, 98, 97-99.
22. Stunkard, A., Sorensen, T. & Schulsinger, F. (1983). Use of the Danish adoption
register for the study of obesity and thinness. In S. Kety (Ed.), The genetics of
neurological and psychiatric disorders (pp. 115-120). New York: Raven Press.
23. Thompson, J. K. & Altabe, M. N. (1991). Psychometric qualities of the figure
rating scale. International Journal of Eating Disorders, 10, 5, 615-619.
24. Smith, D. E., Thompson, J. K., Raczynski, J.M., and Hilner, J. E. (1999). Body
image among men and women in a biracial cohort: The CARDIA study.
International Journal of Eating Disorders, 25, 1, 71–82.
25. Garner, D. M., Garfinkel, P. E., and O’Shaughnessy, M. (1985). The validity of
the distinction between bulimia with and without anorexia nervosa. American
Journal of Psychiatry, 142, 581-587.
26. Flynn, K., & Fitzgibbon, M. (1996). Body images and obesity risk among Black
females: A review of the literature. Annals of Behavioral Medicine, 20, 1, 13-24.
27. Furnham, A., Badmin, N., & Sneade, I. (2002). Body image dissatisfaction:
Gender differences in eating attitudes, self-esteem, and reasons for exercise. The
Journal of Psychology, 136, 6, 581-596.
28. Thompson, K. (1996). Assessing body image disturbance: Measures,
methodology, and implementation. In J.K. Thompson (Ed.), Body image, eating
disorders, and obesity (pp. 49-81). Washington, DC: American Psychological
Association.
29. Jackson, A.S., et al. (1996). Changes in aerobic power of women, ages 20 to 64
years. Medicine and Science in Sports and Exercise, 28: 884-891.
30. Jackson, A.S., et al. (1990). Prediction of functional aerobic capacity without
exercise testing. Medicine and Science in Sports and Exercise, 22:863-870.
31. Cash, T. F. (1990). Body images: Development, deviance, and change. Cash,
Thomas F. (Ed.); Pruzinsky, Thomas (Ed.); New York, NY, US: Guilford Press,
1990. pp. 51-79.
32. Thompson, J. K., and Smolak, L. (2001). Body Image, Eating Disorders, and
Obesity in Youth: Assessment, Prevention, and Treatment. 2ndEd., Washington,
DC, US: American Psychological Association, 389. pp. 54-55.
33. Candy, C. M. & Fee, V.E. (1998). Underlying Dimensions and Psychometric
Properties of the Eating Behaviors and Body Image Test for Preadolescent Girls.
Journal of Clinical Child Psychology, 27, 117-127.
34. Collins, M. (1991). Body figure perceptions and preferences among preadolescent
children. International Journal of Eating Disorders, 10, 100-108.
35. Tiggemann, M. (1992). Body-size dissatisfaction: Individual differences in age
and gender, and relationship with self-esteem. Personality and Individual
Differences, 13, 39-43.
36. Demarest, J. & Langer, E. (1996). Perception of body shape by underweight,
average, and overweight men and women. Perceptual and Motor Skills, 83, 569-
570.
37. Oliver, W. (2008). Body image in the dance class. Journal of Physical Education,
Recreation & Dance (JOPERD), 79, 5, 18-25.
38. Irving, L. (1990). Mirror images: Effects of the standard of beauty on the self- and
body-esteem of women exhibiting varying levels of bulimic symptoms. Journal of
Social and Clinical Psychology, 9, 230-242.
39. Cash, F. & Henry, E. (1995). Women’s Body Images: The results of a national
survey in the USA. Sex Roles, 33, 19-28.
40. Thompson, K. & Stice, E. (2001). Thin-Ideal Internalization: Mounting Evidence
for a New Risk Factor for Body-Image Disturbance and Eating Pathology.
Current Directions in Psychological Science, 10, 5, 181-183.

Assessing Value of the Draft Positions in Major League Soccer’s SuperDraft

 

Abstract

This paper assigns various performance  measures to players who have been drafted in the SuperDraft of Major League Soccer. These measures are then used to assess the value of draft position. As a by-product of the analysis, we provide an estimate of the trajectory of player performance in a soccer career. Both the valuation of draft position and the career trajectory analysis may be important tools for general managers in Major League Soccer.

INTRODUCTION

By all measures, soccer (or football as it is more commonly known) is the most popular sport in the world (The Economist 2011). In the United States and Canada, the highest level of soccer is played in Major League Soccer (MLS). As of 2012, MLS consisted of 19 clubs divided into the Eastern Conference (10 teams) and the Western Conference (9 teams). Unlike the traditional European fall-to-spring schedule, the MLS season begins in March and ends in November. The MLS team with the best regular  season record is the recipient of the Supporters Shield and the winner of the MLS playoffs receives the MLS Cup.

There are various ways that  MLS teams acquire players for their rosters. One such mechanism is the annual SuperDraft  whereby players are placed on the draft-eligible list via nomination from MLS clubs. Most of these players are US college players. Players are then drafted by MLS teams in the inverse order of team performance from the previous  season. The ordering is an attempt to improve competitive balance within the league. The number of players drafted has varied since the inception of the SuperDraft in 2000. In the most recent draft of 2012, 38 players were drafted (two rounds involving the 19 MLS clubs).

The primary problem considered in this paper is the valuation of draft order in the SuperDraft. This is a fundamental problem for general managers in MLS. Consider the following hypothetical scenarios where the knowledge of draft value would be useful:

Your team has a glaring weakness at fullback and you have the 20th pick in the draft. What are the chances that the draftee can satisfy your needs at fullback? Would it be a better strategy to fill this position via a trade or via a designated player?

Your team is facing salary cap constraints and you have the 10th pick in the draft. What is the salary expectation for the player?

Your team has the draft rights for both the 10th and 11th picks. Would it be in your interest to trade both of these picks for the first draft pick?

The valuation of draft order was first considered in the context of the annual draft of the National Football League (NFL). Armed with a draft value chart, Jimmy Johnson (coach of the Dallas Cowboys) made informed trades on draft day in 1991, acquiring a whopping 19 players for the Dallas Cowboys. These informed trades eventually turned the team’s fortunes around as Dallas won three Superbowls within the next five years. Moskowitz and Wertheim (2011) provide an entertaining account of the use of draft value charts in the NFL.

The early chart was constructed according to the value of draft picks based on actual trades that took place. Massey and Thaler (2010) investigated the chart, and using data based on future salary contracts, determined that the early chart overvalued high draft picks. For example, although the first draft pick is valuable, the early chart suggested that the first pick is more valuable than is actually the case. Shuckers (2011) corroborated the overvaluation associated with the early draft chart via alternative measures of player performance. Shuckers (2011) subsequently proposed an alternative draft value chart. In other sports with drafts, Berry (2001) has looked at the success of first round picks. However, it does not appear that there has been any published investigation of the value of draft order in the MLS SuperDraft.

One of the complicating factors in assessing the value of draft order in the SuperDraft is that many of the players who have been drafted have not yet reached their peak level of play. Although some players when drafted in their early 20’s are unable to make an immediate contribution, through training and experience, they eventually become serviceable MLS players. We therefore want to be able to assess the future value of draft picks.

In section 2, we address this problem by estimating career trajectories of soccer players. Common sense dictates that players generally improve early in their career, then peak, and finally experience a decline in performance. The age at which they peak, and the extent of their improvement/deterioration  around the peak is the focus of section 2. Careful data collection is required to assess player productivity. Although the estimation of career trajectories is a necessary step in assessing draft order value, the problem itself is one of considerable interest to general managers. For example, what length and value of a contract should be offered to a player who is two years past his peak?

In section 3, we assess the value of draft position. The primary difficulty in the exercise is the determination and collection of performance measures. Clearly, a common metric such as “goals scored” is not relevant to all positional players. For example, defenders are not expected to score goals. The end product of our analysis are graphs of value plotted against draft position. Various graphs are provided based on various performance measures. With such graphs, one can assess for example, the relative value of the 20th draft pick to the first draft pick. We conclude with a short discussion in section 4.

CAREER TRAJECTORIES IN SOCCER

Assessing performance in soccer is not a straightforward task. For example, although goals scored is a popular and important statistic, it is not relevant to all positional players, especially defenders. Also, imagine a player who changes teams and experiences a surge in productivity. This may have more to do with circumstances surrounding his new team than an improvement in his personal play. In this section we make a number of subjective and hopefully reasonable decisions with respect to assessing player performance. Yearly performance data coupled with age data will help us to evaluate career trajectories  in soccer.

We began by considering players who have played on top flight clubs for a period of at least five consecutive seasons. We have chosen 12 teams from Europe which have been traditional footballing powers. The rationale is that these clubs are of the highest quality and consistency. Therefore an observed change in performance when competing for these clubs is assumed to convey a change due to the player. We have restricted our analysis to the “modern” era of soccer beginning with the 1992/1993 season when the English Premier League was formed. We have excluded goalkeepers from the analysis as it is well known that keepers can remain competitive at more advanced ages. For example, Edwin van der Sar retired as keeper from Manchester United following the 2010/11 season at 40 years of age. Although players from the 12 chosen clubs are of exceptional quality, we assume that the average career trajectory for players at these clubs is not unusual.For example, this implies that the average peak age for players at the elite clubs should be the same as the average peak age for all professional soccer players.

From the websites www.footballdatabase.eu  and www.bdfutbol.com/en/e/e.html we identified 232 players who met our criteria. These players played a total of 1791 seasons. We were also able to collect seasonal performance data on each of the players in terms of minutes played. Although minutes played does not capture all of the elements of performance, it is clearly a sensible measure of worth. The data are summarized in Table 1.

Team League No. Players
Arsenal English Premier League 25
Chelsea English Premier League 11
Liverpool English Premier League 23
Manchester United English Premier League 25
Bayern Munich German Bundesliga 26
AC Milan AS Italian Serie A 20
Roma Italian Serie A 18
Internazionale Italian Serie A 14
Juventus Italian Serie A 18
Barcelona Spanish La Liga 20
Real Madrid Spanish La Liga 16
Valencia Spanish La Liga 16
Overall 232

 

Table 1: Summary data for players who have played at least five consecutive years with a top flight club sometime during the 1992/93 through 2011/12 seasons.

Recall that our objective in this section is the evaluation of career trajectories. Therefore, we want to assess individual player performance at different ages where player performance is measured against himself. Accordingly, let xij denote the minutes played for the ith player at age j in the season where the player was j years old on January 1. The beginning of January is roughly halfway through the traditional fall-spring soccer season. Corresponding to xij, let wij be the number of available minutes during the season where wijis the product of 90 minutes and the number of games  that his team played in the season. We then define zij= xij/wij as a performance statistic for the i the player in the given season as it represents his proportion of available minutes on the field. To provide a personal measure
of performance for the ith player at age j, we define

 

where the maximum is taken over all of his active seasons at the club. Therefore the personal performance measure ppij for the i player has a maximum value of 100% in at least one of his seasons. Note that we only considered matches played in the domestic  season. Our rationale for this choice is that other matches (e.g. friendlies, FA Cup, Europa League, Champions League, etc) may not have a consistent level of competition. Also, it is well known that players are often given rest when matches are scheduled in frequent succession.

In Figure 1, we provide a scatterplot of the personal performance measure ppij versus age j based on minutes played for all players over all seasons. The plot consists of 1791 points. Although there is considerable variability in the scatterplot, the lowess function describes the overall trend. The shape of the lowess plot corresponds to our intuition where performance improves early in a career, then peaks, and concludes with a period of decline. We observe that typical player performance peaks at roughly 24-27 years of age. Also, the drop off is minimal for roughly two years surrounding the peak period. The improvement early in a career is more rapid than the decline at the end of a career. We note that although the peak period of 24-27 years appears to be in rough agreement with common opinion (www.forum.ea.com/uk/posts/list/802432.page), we know of no quantitative study that has provided  such an interval. A miscalculation by even a couple of years can have dire consequences for general managers when deciding upon long term contracts. The parameters of the lowess function were set according to span=0.5 and degree=2 which is consistent with the analyses of Shuckers (2011). The choice of the span parameter appears reasonable as we do not want large fractions of the data to fit local segments of the trajectory curve. For example, the trajectory at young ages does not likely have much to do with longevity. The degree parameter  should also take a value exceeding 1 since we do not expect linear segments in the trajectory curve. Setting degree=2 provides more curvature.

Figure 1:  Scatterplot of the personal performance  measure ppij based on minutes played versus age. A lowess fit is included to help assess the overall trend.

Recall that each player has at least one plotted value of 100%. Accordingly, there is an inherent assumption that data are collected over a period that includes each player’s peak years of play. The assumption is valid for most players since it is unlikely that a player had five consecutive years of form that were all below peak performance, yet he managed to play on one of these top flight teams. However, to check the robustness of the assumption, we have repeated the analysis based on a restricted subset of 129 players  who played at least seven consecutive seasons with one of the 12 top flight clubs. These players played a total of 1229 seasons. A greater period of seasons improves the chance of capturing a player’s peak year. We found no meaningful difference in the resultant plot when compared to Figure 1. Related to the assumption, we encountered 9 players (e.g. Zinedine Zidane) who played at least five years on each of two teams from our list. In these cases, we only retained the years at the club where the player was 26 years of age. In the case of Zidane, we used his Juventus years instead of his years at Real Madrid.

How do we interpret the vertical scale in Figure 1? From 24-27 years of age, an average player operates at roughly 75% of top performance. Similarly, an average 34 year old operates at roughly 50% of top performance. Therefore, we might view the 35 year old as being able to contribute 50/75 two thirds of what he was able to contribute at his peak. Here, the measure of contribution can be interpreted in minutes played.

ASSESSING VALUE  OF  DRAFT POSITION

In this section we investigate the relationship between player value and draft position in the MLS SuperDraft. To facilitate the investigation, www.wikipedia.org/wiki/MLS  SuperDraft provides the entire history of the MLS SuperDraft going back to the inaugural draft on February 6, 2000. The list consists of a total of 745 players chosen in the 13 drafts from 2000 through 2012. The number of players drafted per year ranges from 72 (2001) to 38 (2012).

Whereas obtaining the draft position of each player is uncontroversial and routine, the determination of player value is far from straightforward. As in section 2, we first consider value measures based on minutes played. Another indicator of player value which we consider is yearly salary. Although players can be underpaid or overpaid in a given contract, subsequent player contracts tend to adjust to reality. A difficulty  with using salary as a proxy for performance is that there are time lags between observed performance and contract. Fortunately, for the MLS, there are good sources of data. For minutes played, we made use of http://socceroutsider.com/ and player’s personal Wikipedia sites. For salary information, data are available for the six seasons 2007 through 2012 at www.mlsplayers.org/salary info.html. We had to dig deeper for earlier seasons, referring to www.washingtonpost.com/wp-srv/sports/mls/longterm/2006/mls.salaries.html for the 2006 season, www.bigapplesoccer.com/article.php?article id=3103 for the 2005 season and http://sportsillustrated.cnn.com/2004/soccer/01/06/mls.salaries.sa/  for the 2004 season.

Before the various value metrics are introduced, we indicate some general problems that need to be addressed with respect to player valuation:

A spectacular  season is valuable to a club. There is also value in a longstanding career. How do we balance short term performance with longevity?

How do we assess a player who was drafted in the SuperDraft but went on to play in some other league? For example, Clint Dempsey was drafted 8th in the 2004 SuperDraft by the New England Revolution. However, since 2006/07, Dempsey has played in the English Premier League for both Fulham and Tottenham. Clearly, Dempsey is a valuable player although he has only limited MLS data.

In our current list of drafted players, some players are still young and have not yet reached their full potential. Should we assess their value on some combination of current performance and future performance? The career trajectory plot Figure 1 may be helpful in this regard.

With the growth of the MLS, there has been an escalation in salaries over time. How do we compare a salary from the early years to recent salaries?

To facilitate a comparison of various metrics, we define each metric on a scale of 0 to 100. We begin by considering minutes played in regular season MLS matches. The restriction to regular season matches levels the comparison amongst players since each team has the same number of games of comparable importance. For each player, we calculated his percentage of total minutes played relative to available minutes in a season. This was done for each year over all of his MLS seasons. We then defined the player performance metric y1 for a given player as his maximum percentage over his MLS career. We also defined the player performance metric y2 for a given player as his average percentage over his three best MLS seasons. Therefore, the measure y2 favours career longevity over y1.

In the case of players who have gone on to play in ldquo;superior” leagues, we arbitrarily assign a percentile rank of 90% in seasons where they played in superior leagues. We define a superior  league as any of the four famous leagues listed in Table 1. Players who have made it to one of these four top flight leagues have typically developed in the MLS before making their jump to the big time. Although there are other leagues that many football fans would agree are better than the MLS (e.g. Liga Portuguese (Portugal), Ligo Do Brasil (Brasil), Ligue One (France), Primera Liga (Argentina), Eredivisie (Holland)), very few MLS drafted players have gone on to play in these professional leagues. When an MLS draftee plays in leagues other than the MLS or one of the four top flight leagues, we assign a percentile rank of 0% for those seasons. A 0% score is also assigned to players who discontinue playing. A rationale for the 0% performance score is that it was a mistake to draft such a player as they did not contribute to the MLS team that drafted them. In our list of MLS drafted players, only 12 played in superior leagues. Of the 339 that played in other leagues, 277 played in the NASL (North American Soccer League) or the USL (United Soccer Leagues) which can truly be viewed as lower quality leagues compared to the MLS. When a player belonged to multiple leagues in a year, we used the league where he played most of his games.

To account for young players who have not yet reached their top level of performance (i.e. less than 24 years of age), we imputed values for their unobserved seasons. For example, suppose that we have a 21 year old player in the MLS who has just completed a season. We take his minutes played as a 21 year old and multiply by l(22)/l(21) to obtain his predicted minutes as a 22 year old where l(x) is the value of the lowess function at age x in Figure 1. We do not allow predicted percentages to exceed 100%. The determination of player ages was not as straightforward as we had hoped; there were at least six additional websites that we accessed to collect birthdates.

An obvious difficulty with the minutes played variables y1 and y2 is that they do not account for team strength. For example, it is easier to play extended minutes on a poorer performing club than at a stronger club. Another difficulty with these measures is that there are a number of players who represent their club in nearly all of the regular season matches. Therefore minutes played does not adequately distinguish these players in terms of performance.

Our preferred performance metrics are based on salary data. Varying team strength is less of an issue when dealing with salary data since league-wide  salary caps exist. Salary caps help to impose a realism on salaries so that players are paid what they are worth. An exception are the salaries paid to designated players who are not MLS draftees. For each player, we calculated his percentile rank for a given year based on his salary relative to all MLS players in that year. Thus the comparison sensibly involves the performance of draftees against the population of MLS players. This was done for each year over all of his MLS seasons. We then defined the player performance metric y3 for a given player as his maximum percentile rank over his MLS career. We also defined the player performance metric y4 for a given player as his average percentile rank over his three best MLS seasons. We used the same rules as above in handling players who have gone on to play at superior clubs and for young players who have not yet reached their full potential. In the case of multi-year contracts, the amount that a player received in a year is used as his yearly salary.

In Figure 2, we provide a scatterplot of the preferred y4 metric versus draft position. To assess the overall trend, a lowess plot is superimposed. We observe the anticipated pattern that early draft picks have more value. The plot decreases rapidly during the early picks and then levels off. It is interesting that there is little additional value beyond draft position 25. What this suggests is that whereas general managers have good intuition of value early in the draft, late draft picks are more or less a crapshoot. It also suggests for example that managers should value a 50th draft pick as about equal to a 25th draft pick. If this sort of information is not well understood, a savvy manager may be able to trade a 25th draft choice for a 50th draft choice plus additional assets. We observe from the variability in the plot that it is still possible to draft a productive player late in the draft but the probability of doing so is much decreased. Moreover, the variability is greater for early draft picks. This suggests that that there may be great pressure for teams drafting early as an early draft choice can turn out to be either a star or a bust. With less expected of late draft picks, there may only be upside for a general manager. How might we interpret the vertical scale of the plot? According to the salary metric, the first draft pick provides you with a player who on average ranks in the top 85.6% of players in the MLS. The first draft pick is about 1.7 times as valuable as the 10th draft pick and is about 5.3 times as valuable as the 20th draft pick in terms of salary. Finally, we note that we obtain a very similar plot if y4 is based on the best four seasons rather than the best three seasons. For ease of reference, the values of the loess curve in Figure 2 are provided in the pick value chart given in Table 2 of the Appendix.

In Figure 3, we provide a comparison of the lowess plots using each of the four proposed metrics. We observe that each plot conveys similar information and this provides assurance that the valuations are meaningful. For example, all four plots indicate that the first draft pick injects a team with a player who on average will rank roughly in the top 80% of MLS players. The level of agreement in the four curves was a bit of a surprise to us as we were aware of the flaws in assessing value based on minutes played. For the most part, we also note that y1 dominates y2 and that y3 dominates y4. This suggests that the consideration of a single season (y1 and y3) tends to inflate a player’s value when compared to their performance over multiple seasons (y2 and y4). Finally, we note that the preferred metric y4 lies amongst the middle of the four lowess curves.

Figure 2: Scatterplot of the value metric y4 versus draft position. A lowess fit is included to help assess the overall trend.

DISCUSSION

The major contribution of the paper is the construction of Figure 1 and Figure 2. In Figure 1, we have estimated the performance trajectory of soccer players (excluding  keepers). The plot may be of value to general managers in planning team rosters and offering contracts. In Figure 2, we have estimated  the value of draft picks in the MLS SuperDraft. This may also be of value to general managers when planning rosters and assessing trades.

The major difficulty in both the construction of Figure 1 and Figure 2 is the determination of player value. Value is a subjective quantity and involves many factors including player age, the assessment of the importance of longevity, positional characteristics, the confounding of individual and team characteristics, changes in league salaries and missing data. We have attempted to handle these issues sensibly and we note that various value metrics have lead to similar results (see Figure 3). Data collection and data management also proved to be a substantial exercise in obtaining our results. It may be interesting to extend the work by restricting analyses to various positions (e.g. defenders, midfielders,  forwards) although this reduces the sizes of data sets and often introduces uncertainties with respect to the categorization of players.

Figure 3: Lowess plots for each of the value metrics.

Finally, a relatively new change may affect the future of the MLS SuperDraft. In 2006, the Home Grown player criteria was established in the MLS to nurture promising young players living in the vicinity of an MLS team. The designation requires commitment by players to practice/play sufficiently under the club’s development system. Teams may then sign a player to their first professional contract if the player has trained for at least one year in the club’s youth development program and has met the league’s Home Grown player criteria. The important implication is that such players do not participate in the SuperDraft. Therefore, although the Home Grown player program is currently in flux, should it gain widespread popularity, it may eventually dilute the quality of the SuperDraft.

APPENDIX

Pick    Value Pick    Value Pick    Value Pick    Value
01       85.6
02       81.5
03       77.5
04       73.5
05       69.5
06       65.6
07       61.7
08       57.9
09       54.2
10       50.5
11       46.8
12       43.2
13       39.7
14       36.1
15       32.7
16       29.2
17       25.4
18       21.9
19      18.8
20      16.1
21      13.5
22      11.3
23      09.6
24      08.3
25      07.3
26      06.3
27      05.7
28      05.4
29      05.1
30      04.8
31      04.7
32      04.6
33      04.3
34      04.1
35      04.0
36      03.8
37      03.6
38      03.5
39      03.5
40      03.4
41      03.3
42      03.3
43      03.3
44      03.2
45      03.1
46      03.1
47      03.1
48      03.0
49      02.9
50      02.7
51      02.6
52      02.5
53      02.3
54      02.2
55      02.0
56      01.9
57      01.7
58      01.6
59      01.4
60      01.2
61      01.1
62      00.9
63      00.7
64      00.6
65      00.4
66      00.2
67      0.00
68      0.00
69      0.00
70      0.00
71      0.00
72      0.00

 

Table 2: Pick value chart corresponding to the lowess curve in Figure 2.

 

 

REFERENCES

Berry, S.M. (2001). “Do you feel a draft in here?”, In the column, “A Statistician Reads the Sports Pages”, Chance, 14(2): 53-57.
Massey, C. and Thaler, R.H. (2010). “The loser’s curse: overconfi dence vs. market efficiency in the National Football League draft”, http://ssrn.com/abstract=697121.

Moskowitz, T.J. and Wertheim, L.J. (2011). Scorecasting: The Hidden Influences Behind how Sports are Played and Games are Won, Crown Archetype: New York.
Shuckers, M. (2011). “An alternative to the NFL draft pick value chart based upon player performance”, Journal of Quantitative Analysis in Sports, 7(2): Article 10.1
The Economist (2011). Game Theory blog post under “Ranking sports’ popularity: And the silver goes to …”, www.economist.com/blogs18