This article was written by Stephen Smith
This article was published in the Fall 2011 Baseball Research Journal
Introducing a model to eliminate five basic problems in talent research and applying this model to predicting whether youth athletes were selected for their leagues’ All-Star teams.
One of the most appealing reasons for talent research in sport is the hope that future talent can be predicted in “key transferable skills.”[fn]Abbott, A., Button, C., Pepping, J.G., & Collins, D. (2005). Unnatural selection: Talent identification and development in sport. Nonlinear Dynamics, Psychology, and Life Sciences, 9, 62.[/fn] Previous talent research has used a variety of measures to select talent but typically focus on only a small portion of physical skills or attributes (height, weight, strength, and speed). Some research has included skill-specific tests (slalom dribbling in handball, serving in volleyball) but the area of talent research has not investigated the dynamic nature of talent in sport. This neglect leads to several problems (see Abbott et al., 2005, for a review). In this paper we will enumerate five basic problems with talent research as it is often conducted, introduce the Dynamical Systems model to eliminate these problems, and then apply this model to predicting whether athletes in our study were selected for their leagues’ All-Star teams.
HOW IS TALENT RESEARCH USED?
Talent research studies typically test discrete physical measures such as upper body strength, vertical jump, linear speed, and the cardiovascular endurance of athletes in an attempt to predict future performance. Early identification would provide young athletes with the several years needed for developmental training to become elite-level performers.[fn]Ericsson, K.A., Krampe R.T., & Tesch-Rômer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 3, 363–406.[/fn] In baseball, specific talent research studies have indicated that vertical jump,[fn]Everett, P.W. (1952). The prediction of baseball ability. Research Quarterly, 23, 15–19.[/fn] vertical jump and medicine ball throw,[fn]Grove, J. (2001). Practical screening tests for talent identification in baseball. Applied Research in Coaching and Athletics Annual (Boston), 16, 63–77.[/fn] and Profile of Mood States (POMS)[fn]Friend, J., & LeUnes, A. (1990) Predicting baseball player performance. Journal of Sports Behavior, 13, 73–87.[/fn] are the best predictors of baseball performance. But talent research can be a confusing area and perhaps this is why it has sparked controversy.
PROBLEMS IN TALENT RESEARCH
One problem with talent-related research is a wide range of developmental levels in participant groups, making results difficult to compare or correlate. Another is the intent of how the information is used, which skews the directions of research and can color the reporting of the results. Three baseball related studies in the field studied college-level (Friend & LeUnes, 1990), professional (Grove, 2001), and high school and Little League baseball players (Watkins, Garcia, & Turek, 1994).[fn]Watkins, B., Garcia, A., & Turek, E. (1994). The relation between self efficacy and sport performance: Evidence from a sample of youth baseball players. Journal of Applied Sport Psychology, 6, 21–31.[/fn] All three articles attempted to identify measures of physiological, perceptual, and psychological variables without the goal of eliminating individuals from future participation. The goal here was to identify variables that can be improved throughout the developmental process rather than be used to exclude individuals from participation.
DYNAMICAL SYSTEMS MODEL
Abbott et al. (2005) reviewed five reasons why identifying and eliminating potential athletes based solely on physiological measures (i.e. organismic constraints) is questionable. The first reason is that it is too difficult to measure the value of a physiological variable at an immature point in time and expect that variable to remain constant and unchanged throughout an individual’s development.[fn]Abbott, A., & Collins, D. (2002). A theoretical and empirical analysis of a ‘state of the art’ talent identification model. High Ability Studies, 13, 157–178.[/fn], [fn]Aitken, D.A., & Jenkins, D.G. (1998). Anthropometric-based selection and sprint kayak training in children. Journal of Sports Sciences, 16, 539–543.[/fn], [fn]Simmons, C., & Paull, G.C. (2001). Season-of-birth bias in association football. Journal of Sports Science, 19, 677–686.[/fn] For example, height is advantageous in several sporting domains (basketball, volleyball, football, and baseball). Abbott and Collins (2002) report significant variability in height between boys and girls aged 11–12. There are many maturational changes that take place from age 11–12 through the next few years. The previous talent research conducted on baseball does not take into account individual variability, nor did it include a baseball-specific test for the skill of batting.
The second problem is that performance is constrained by several factors that are not all genetically determined (Abbott et al., 2005). The theoretical framework known as the dynamical systems perspective[fn]Newell, K.M. (1986). Constraints on the development of coordination. In M.G. Wade & H.T.A. Whiting (Eds.), Motor development in children: Aspects of coordination and control (pp. 341–360). Boston: Martinus Nijhoff.[/fn] suggests that skill acquisition arises from an interaction between constraints within the organism, in the environment, and in the task.[fn]Summers, J.J. (2005). Historical perspective on skill acquisition. In Skill Acquisition In Sport: Research Theory and Practice. (A. Mark Williams, and Nicola J. Hodges Eds.) (p. 15). London, UK: Routledge (Taylor & Francis Group).[/fn] An individual will have to go through periods of instability in order to reach a more stable pattern of movement or perception. This is the key tenet of dynamical systems and also appears to be a major flaw in past talent research studies. Several studies measured a limited array of physiological factors that may play a role in the performance of the skill being tested; however, these studies only tested a limited spectrum of what actually enhances skill acquisition and success in sports.[fn]Gabbett, T., Georgieff, B., & Domrow, N. (2007). The use of physiological, anthropometric, and skill data to predict selection in a talent-identified junior volleyball squad. Journal of Sports Sciences, 25, 1,337–1,344.[/fn], [fn]Lidor, R., Falk, B., Arnon, M., Cohen, Y., Segal, G., & Lander, Y. (2005). Measurement of talent in team handball: The questionable use of motor and physical tests. Journal of Strength and Conditioning Research, 19, 318–325.[/fn] Summers (2005) mentioned principles that cover metabolic, neural, biomechanical, and psychological processes. For example, if an athlete lacks motivation to practice and improve in a particular sport (a psychological variable), then he/she is highly unlikely to improve even if they are extremely talented physiologically.
The third problem associated with talent research according to Abbott et al. (2005) and Morris[fn]Morris, T. (2000). Psychological characteristics and talent identification in soccer. Journal of Sports Sciences, 18, 715–726.[/fn] is that discrete performance indicators are poor predictors of talent. Fleishman and Mumford as well as various other studies conducted by Fleishman and colleagues consistently showed particular abilities related to skill acquisition change as an individual continued to practice during development.[fn]Fleishman, E.A., & Mumford, M.D. (1989). Abilities as causes of individual differences in skill acquisition. Human Performance, 2, 201–223.[/fn], [fn]Fleishman, E.A. (1964). The structure and measurement of physical fitness. Prentice Hall: Englewood Cliffs, NJ. Fleishman, E.A., & Bartlett, C.J. (1969). Human abilities. Annual Review of Psychology, 20, 349–380.[/fn] As previously stated, variations in development result from interaction between the environment, the task, and personal (organismic) constraints (Newell, 1986). Of particular interest to this research, and a factor that seems to be missing in previous studies, are psychological variables (organismic constraints) such as motivation. Thelen and Ulrich (1991) noted certain behaviors do not emerge to support other processes (physiological and perceptual abilities) until they are ready to emerge.[fn]Thelen, E., & Ulrich, B. D. (1991). Hidden skills: A dynamic systems analysis of treadmill-elicited stepping during the first year. Monographs of the Society for Research in Child Development, 56, 1–104.[/fn] If rate limiters such as lack of motivation or muscular strength are present, then identifying talent for immediate or future development will be difficult if not impossible.[fn]Haywood, K.M., & Getchell, N. (2004). Lifespan motor development (4th ed.). Champaign, IL: Human Kinetics.[/fn]
According to Davids, Lees, and Burwitz[fn]Davids, K., Lees, A., & Burwitz, L. (2000). Understanding and measuring coordination and control in kicking skills in soccer: Implications for talent identification and skill acquisition. Journal of Sports Sciences, 18, 703–714.[/fn] and Helson, Hodges, van Winckel, and Starkes,[fn]Helson, W.F., Hodges, N.J., Van Winckel, J., & Starkes (2000). The roles of talent, physical precocity and practice in the development of soccer expertise. Journal of Sports Sciences, 18, 727–736.[/fn] the talent selection criteria used are often unstable during developmental transitions in an athlete’s career. This is the fourth problem in talent research studies. Athletes learning a new skill (regardless of age) go through transitions in performance. Based on these transitions, performance may become unstable (the athlete may even appear to regress in the skill). Using dynamical systems terms, this is called a critical fluctuation where a stable characteristic of performance becomes unstable in order for it to scale up to a more effective state (Abbott et al., 2005). Therefore, it would be increasingly difficult for an athlete to be correctly and accurately identified as talented if the athlete was observed during one of these periods of critical fluctuations in development.
The fifth problem associated with talent research studies is they typically do not take into account factors that support successful development towards acquisition of skill.[fn]Simonton, D.K. (1999). Talent and its development: An emergenic and epigenetic model. Psychological Review, 106, 435–457.[/fn] Recruiters and scouts look for individuals with the capacity to develop into athletes who will garner significant success later on. However, many talent research studies (Friend & LeUnes, 1990; Grove, 2001) measure performance at a particular moment, rather than skill acquisition potential. The major limitation with this method of testing is eliminating potential talent too early. A child’s desire to participate in a particular sport may be squashed or dissuaded by negative-seeming test results. For this reason, season-long statistics for batting average (BA) and on-base plus slugging percentage (OPS) are preferred for analysis over momentary test results.
The previous five problems do not paint a flattering picture but enumerating them does motivate us to design a study to identify measurable factors which can enhance the accuracy and predictability of talent research. In this study, we attempt to predict which athletes should be selected for their league’s All-Star team by utilizing a Dynamical Systems perspective. Talent selection is a process which needs to take into account that successful performance is a result of the interaction of physical (strength, speed), perceptual (hitting pitches at different speeds and trajectories), and psychological factors (self-efficacy).
Participants were 68 boys between 11 and 14 years of age who played in a recognized organization (i.e. Little League or Dixie Youth League). One hundred participants were tested during the study but 32 participants had incomplete data sets and were excluded from the statistical analysis. Approval from the University’s Institution Review Board was obtained prior to administering the testing protocol. In addition, permission was obtained from league officials (i.e. league president or appropriate board members). A letter was sent to parents of children in the league asking them to volunteer for this study and participants’ parents signed a consent form in order for children to par-
ticipate. The organizations were located in Central Virginia. Each participant was required to be a member of an official team within that organization in order to be eligible for the league’s All-Star team.
A total population sample was better suited to the research than a selection because to determine if the measurable factors that differentiate successful versus unsuccessful performance would predict All-Star selection, then a full range of performance levels was required. We tested five independent variables: (1) lower body power potential (organismic constraint), (2) upper body power potential (organismic constraint), (3) 30-yard dash (organismic constraint), (4) hitting 60 mph fastballs and right- and left-handed curveballs (task constraint), and (5) level of self-efficacy for hitting a baseball (organismic constraint). We needed approximately 100 participants in order to reveal major differences between successful and unsuccessful performance.[fn]Stevens, J. (2002). Applied multivariate statistics for the social sciences (4th ed.). Mahwah, NJ: Erlbaum.[/fn]
Procedure, Instrumentation, and Independent Measures
Data collection occurred through participant observations. No subjectivity was involved in testing any of the independent variables. Objective tests aimed at differentiating between successful and unsuccessful batting performance were administered, as follows:
Vertical Jump (physical test). Lower body power potential was measured using a Vertec?stationary vertical jump tester. A piece of athletic tape was placed on the floor indicating where the participant’s feet should be placed and each participant was asked to extend his jumping arm to provide a standing reach measurement. Each participant was allowed three opportunities to jump as high as possible and touch the Vertec. This device includes colored plastic swivel vanes arranged in ½-inch increments attached to a telescopic metal pole that can be adjusted to the athlete’s standing reach.[fn]Klavora, P. (2000). Vertical jump tests: A critical review. Strength and Conditioning Journal, 22, 70–74.[/fn] The participant was allowed to bend at the knees and swing the arms upward to initiate his jump. If the participant used a drop step technique (dropping the foot back then forward to initiate the jump), that jump would be disqualified and the tester would re-instruct the participant on the correct form.
Medicine ball toss (physical test). The side medicine ball toss has been shown by Ikeda, Kijima, Kawabata, Fuchimoto, and Ito to be a more skill-specific movement than the exercise used in the field tests conducted by Friend and LeUnes (1990) and Grove (2001).[fn]Ikeda, Y., Kijima, K., Kawabata, K., Fuchimoto, T., & Ito, A.(2007). Relationship between side medicine-ball throw performance and physical ability for male and female athletes. European Journal of Applied Physiology, 99, 47–55.[/fn] It used a 2kg medicine ball with the participant standing in his batting position facing perpendicular to the throwing area. The participant held the medicine ball in both hands and rotated his body back and then forward (similar to a swinging motion). During the forward rotation, the participant released the medicine ball to throw it for maximum distance.
30-yard dash (physical test). The 30-yard dash time was conducted in an indoor facility so surface and weather could be controlled. Timing began when the participant swung the bat and ended when they touched first base. Time was recorded using a Cronus Pro Survivor stopwatch and then recorded for data analysis (Leavitt, personal communication, April 12, 2006). Pilot testing revealed a reliability of .892 which was within the acceptable statistical parameters. In addition the test was conducted by the primary researcher for each participant tested.
Hitting tasks (perceptual). Each participant took fifteen swings on three types of pitches from a pitching machine 46 feet away. Each saw five 60 mph fastballs, and five each of 50 mph right- and left-handed curveballs A Jugs II pitching machine was used in order to ensure each participant received the same treatment. Balls used were Jugs Sting Free dimpled white batting cage balls to ensure uniformity of pitches. Each area of the batting cage had a point value and the total number of points was calculated for each round.
Self-efficacy. The fifth and final test was a questionnaire that assessed the participant’s perceived self-efficacy. Players’ self-efficacy beliefs can contribute to their successful or unsuccessful performance in baseball. “Perceived self-efficacy is a judgment of capability to execute given types of performances.”[fn]Bandura, A. (2005). Adolescent development from an agentic perspective. Self-Efficacy Beliefs of Adolescents, 1–43. Information Age Publishing: Charlotte, NC.[/fn] Outcome expectations are judgments about how the individual performed in the skills being tested. The individual is likely to derive his outcome expectations from three different forms that include positive and negative physical, social, and self-evaluative performance outcomes.[fn]Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.[/fn] This perceived efficacy is critical because it affects behavior, goals and aspirations, outcome expectations, perceptions of preventing successful outcomes, and opportunities for success.[fn]Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191–215.[/fn], [fn]Bandura, A. (Ed.). (1995). Self-efficacy in changing societies. New York: Cambridge University Press.[/fn] Since self-efficacy is measuring a perception of how one expects to perform a given task, the modified self-efficacy questionnaire was administered after the physical and perceptual measures were collected. Participants were instructed to answer the questions based on their previous year participating in baseball rather than from the tests they just participated in on a scale of 0–100% confident. (Sample: 100 percent would correspond to “Overall I am confident I can perform well when it comes to batting in baseball.”)[fn]Bandura, A. (2005). Adolescent development from an agentic perspective. Self-Efficacy Beliefs of Adolescents, 1–43. Information Age Publishing: Charlotte, NC.[/fn]
All variables were inspected for means and standard deviations. Following that, correlation coefficients were calculated between independent variables prior to the main analysis. Scores on vertical jump (VJ), medicine ball side throw (MB), home-to-first times (H1), 60 mph fastballs (60), left-handed curveballs (LC), right-handed curveballs (RC), and scores from a self-efficacy questionnaire (SE) were recorded. Season-long batting average and on-base plus slugging percentage (OPS) were recorded and, along with VJ, MB, H1, 60, LC, RC, and SE, were evaluated using a stepwise discriminant analysis which includes independent variables (predictor variables) and a dependent variable which is a grouping variable.[fn]Pedhazur, E.J. (1997). Multiple regression in behavioral research (3rd ed.). New York: Wadsworth.[/fn] This particular grouping variable was determined by whether the individuals made their respective All-Star team.
Performance Characteristics of Participants
Scores were obtained for the seven independent variables in one testing session by the primary researcher. Batting averages (BA) and OPS were also collected at the same time unless the coach needed to calculate or locate participants’ statistics for those two categories. Each independent variable was recorded on a data sheet and later entered for data analysis. Thirty-two participants had incomplete data (missing OPS and BA) and were therefore excluded from the final data analysis. Each participant was asked to disclose whether they made their league’s all-star team. Group statistics are also shown in Table 2 to provide an example of mean differences between groups. Table 2 shows there are differences between groups which suggest that there are differences between all-stars and non-all-stars in this study.
Table 3 illustrates a test of equity of group means which provide evidence as to which variables provide statistically significant differences between all-stars and non-all-stars for all independent variables with BA and OPS producing the best fit at the .05 level. Based on the significance level, it is unlikely the researcher mistakenly found differences. A correlation matrix (Table 4) shows variables selected in the study were independent from one another. In other words, this study did not measure the same underlying concept twice. Table 4 illustrates a correlation matrix. Two
variables (BA and OPS) had higher levels of multicollinearity; however, they did not exceed the .85 level (meaning these two variables are highly correlated but they are not measuring the same construct).
Statistical analysis indicated that four independent variables contributed significantly to the grouping differences (BA, OPS, LCB, and SE).
Accuracy of the Predictions of All-Stars versus Non-All-Stars
Results of the study show 80.9 percent of participants were correctly classified into the predicted group based
on the independent variables selected. Significant independent variables (BA, OPS, LCB, SE) correctly classified 80.9 percent of the sample population indicating independent variables perform reasonably well. Table 5 illustrates the accuracy of the prediction. There was a slight difference in predicting group membership when adjusting for the differences in group size. Table 6 illustrates that predicting All-Star membership improved by more than two percent. However, correctly predicting non-All-Stars decreased slightly. Discriminant function analysis allows the researcher to reject the null hypothesis that physical, psychological, and perceptual characteristics cannot determine placement on an All-Star team. It is important to mention that 11–12 year-olds are selected to different All-Star teams than 13–14 year-olds. Due to this distinction age was not a consideration in the analysis because the age groups are selected to different All-Star teams.
From the data listed above, the following hypotheses cannot be rejected. Discrete physical performance indicators are a poor determinant of success in batting in baseball. On the other hand, there is evidence that perceptual ability in striking a baseball is a useful grouping variable. In other words, selecting a test that measures one’s ability to strike a ball at varying speeds and trajectories can differentiate an All-Star from a non-All-Star. Finally, based on the population sampled, self-efficacy can differentiate between individuals that make their league’s All-Star team and those who do not based on previous performance accomplishments. The discriminant function analysis seems better able to discriminate when a person should belong on the All-Star team as opposed to those who should not.
The present study was designed to examine differences in end-of-season success in Little League baseball based on physical, perceptual, and psychological variables as well as two end-of-season statistics (BA and OPS). The theoretical basis of this study was Dynamical Systems Theory which helps explain talent development within a multi-dimensional approach. Dynamical Systems Theory (Newell, 1986) suggests skill acquisition arises from an interaction between constraints within the organism, in the environment, and in the task (Summers, 2005) but no previous talent research study appears to exist based on multi-dimensional measures. Results of the present study found only left-handed curveballs (likely because a majority of the participants were right-handed), self-efficacy, and the two end-of-season statistics were significant at the .05 level. This shows that talent selection in youth baseball is a multi-dimensional process. Each league should collect BA and OPS throughout the season to aid in making objective All-Star selections. This may appear intuitive to most people but thirty-two participants were excluded from data analysis because BA and OPS were not collected.
Another practical suggestion for coaches would be to utilize the information provided in this study to determine strengths and weaknesses of players. If a coach knew what aspects of batting a child struggled with, the coach could provide remediation for that child to improve performance. By improving performance levels, self-efficacy would likely improve as well. Coaches could also improve self-efficacy beliefs of players by modifying tasks during practices to facilitate more mastery experiences. This change would be considered a task constraint in the Dynamical Systems Theory. By modifying the task the coach can facilitate a successful response.
In order to be accurate when selecting an All-Star roster, certain information should be collected in order for a league to make an informed and objective decision. Practicality should be the driving factor when applying the results of this research to selecting individuals for inclusion on the All-Star team. It appears critical for teams to collect information about season-long performance when selecting their All-Star roster. The collection of this information is not labor-intensive and would require someone to go through a scorebook at the end of the game and record information throughout the course of a season (usually includes about 15–20 games).
Coaches also should realize that despite the fact that physical variables lacked statistical significance, it is still important to be fit and strong. Nevertheless, physical variables should not be a determining factor when predicting the best baseball players. With that said, coaches should still provide sound physical training to their players as a function of making them better baseball players. The bottom line is that being able to hit a baseball is more important than having a great vertical jump (Everett, 1952). That may sound like an intuitive statement but youth coaches often make All-Star selections based on information that is not relevant to baseball performance throughout the season (32 participants in this study were excluding because season-long statistics were not collected).
There is still much work to be done when evaluating talent. The predictive accuracy of a model built with the Dynamical Systems Theory shows it is useful to examine talent through a multi-dimensional approach. However, practitioners need to further scrutinize this research to determine if the results are generalizable for the offensive variables. Researchers should also expand the testing to include fielding and pitching variables so as not to exclude that important aspect of the game when selecting individuals for inclusion on an All-Star team in youth baseball.
DR. STEPHEN SMITH is an assistant professor of Health and Physical Education and Head Golf Coach at Lynchburg College in Virginia.