DISCRIMINATIVE POWER OF BASKETBALL GAME ...

Perceptual and Motor Skills, 2004,99, 1231-1238. O Perceptual and Motor Skills 2004

DISCRIMINATIVE POWER OF BASKETBALL GAME-RELATED STATISTICS BY LEVEL O F COMPETITION AND SEX ' JAIME SAMPAIO

SERGIO IBAREZ GODOY, SEBASTIAN FEU

Sports Sciences Department University of Trks-0s-MontesAlto Duoro

Faculty of Sport Sciences University of Extremadura

Summay.-The purpose of this study was to identify the basketball game-related statistics that best discriminate performances by sex of players and level of competition. Archival data were obtained from the International Basketball Federation boxscores for all games during men's senior (n = 62)' men's junior (n=64), women's senior (n = 62), and women's junior (n = 42) World Championships. The game-related statistics gathered included 2- and 3-point field-goals (both successful and unsuccessful), free-throws (both successful and unsuccessful), defensive and offensive rebounds, blocks, assists, fouls, steals and turnovers. For the analysis only the close games were selected ( N = 105, 1 to 12 points difference). Men's teams were discriminated from women's teams by their higher percentage of blocks and lower percentage of steals and unsuccessful 2-point field goals. Junior teams were discriminated from senior teams by their lower percentage of assists and higher percentage of turnovers. In the two-factor interaction, the teams were mainly discriminated by the game-related statistics identified for level of competition.

Traditionally, basketball coaching intervention has been based upon subjective observations of players and team performance. However, several studies have shown that such visual observations can be unreliable and inaccurate (e.g., MacDonald, 1984; Franks & Mdler, 1986). Therefore, quantitative analysis and evaluation of performance particularly through game-related statistics are being widely used among coaches in order to analyse game events with more valid and reliable data. Published research focused specifically on the association between game-related statistics and team performance is very limited. Akers, Wolff, and Buttross (1991) examined the game-related statistics from 229 college games (NCAA from USA) and concluded that winning teams had higher 2point field goal percentages, free-throw percentages, and rebounds, and lower numbers of turnovers and steals. O n the other hand, Karipidis, Fotinakis, Taxildaris, and Fatouros (2001) analyzed 53 games from European and World Championships and concluded that winning teams secured more defensive rebounds, obtained higher 2- and 3-point field-goal percentages, and had less unsuccessful 3-point field-goals. Using data from five NBA seasons (between 1993-1994 and 1997-1998), Melnick (2001) identified a positive

'Address correspondence to Jaime Sampaio, Departmento de Desporto, Universidade de Trhs0s-Montes e Alto Douro, Apartado 202, 5001-911 Vila Real, Portugal.

1232

J. SAMPAIO, ET AL.

correlation between team assists (a behavioural measure of teamwork) and win-loss record (the lower value of .42 was obtained in 1996-1997 and the higher value of .71 was obtained in 1994-1995). Also, the author compared "assisted team points" and "unassisted team points" in relationship to winloss record, and the results suggested that how a team scores points is more important than the number of points it scores. More recently, Sampaio and Janeira (2003) analysed 353 games from the Portuguese Professional League and concluded that in close games (games with final score differences below 3 points) winning teams had higher 2-point field-goal percentages and freethrow percentages and secured more defensive rebounds. In these studies, differences between winning and losing male senior teams were analysed. However, these results were also confirmed by analysing 64 games played in a junior World Championship (Ibhfiez, Sampaio, Shenz-Lbpez, Gimenez, & Janeira, 2003). A very important topic that has not been adequately investigated is the effect of sex of players and level of competition on team performance as measured by game-related statistics. In fact, the contrast between men and women's basketball and between junior and senior basketball teams has never been made with variables measured directly from the competition. Thus, the purpose of this study was to identify the basketball game-related statistics which best discriminate performances according to sex of players and level of competition.

Participants Archival data were obtained from International Basketball Federation official boxscores for all played games during men's senior (n = 62, Indianapolis in the USA, 2002), men's junior (n = 64, Lisbon in Portugal, 1999))women's senior (n = 62, Nanjing in China, 2002) and women's junior (n =42, Brno in the Czech Republic, 2001) World Championships. Procedure The game-related statistics were gathered by International Basketball Federation technicians and consisted of the following: 2- and 3-point fieldgoals (both successful and unsuccessful), free-throws (both successful and unsuccessful), defensive and offensive rebounds, blocks, assists, fouls, steals, and turnovers. All these game-related statistics were normalized according to ball possessions, defined as the period of play between when one team gains control of the ball and when the other team gains control of the ball, i.e., offensive rebounds are included in the same possession (Oliver, 2004). The opposing teams in a game will always have the same number of possessions. In this way, normalized game-related statistics were transformed to derived rate

DISCRIMINATIVE BASKETBALL STATISTICS

1233

variables and can be used to access team performance across the whole seas o n ( ~ ) Also, . to get a more convenient ratio after their normalization, all game-related statistics were multiplied by 100. For example, the performance of a team that converts 35 field-goals in an 80-possession game would be normalized to 43.8% (35/80), whereas the performance of a team that converts the same 35 field-goals in a 90-possession game would be normalized to 38.9% (35/90). Ball possessions were calculated by the following equation (Oliver, 2004) Ball possessions = (field-goals attempted) - (offensive rebounds) + (turnovers) - 0.4 x (free-throws attempted) .

Statistical Analysis Prior to the inferential statistical analysis the games were categorized through k-means clustering (NoruSis, 1993). This classification method produced four different clusters of greatest possible distinction according to game final score differences: (a) 1 to 12 points, (b) 13 to 27 points, (c) 28 to 46 points, and (d) above 47 points. For the subsequent statistical analyses only the close games were used (N= 105, 1 to 12 points) because they represent best the performance from confronting teams in each specific context (Ibhfiez, et al., 2003; Sampaio & Janeira, 2003). A discriminant analysis was performed to judge which of the game-related statistics are more useful in discriminating game performances by sex, by level of competition, and by sex x level. The interpretation of the obtained discriminant functions was based on examination of the structure coefficients greater than 10.301 (Tabachnick & Fidell, 1989). The accuracy of the equations was tested using the leave-one-out method of cross-validation (NoruSis, 1993). The statistical analyses were performed using SPSS software release 10.0.1, and significance was set at p I .05.

RESULTS The means and standard errors of game-related statistics for each simple effect and interaction are presented in Table 1 and Table 2. For sex of players, the obtained function in discriminant analysis was statistically significant ( p I .05) and had an overall percentage of successful classification of 86.0%. The structure coefficients from the function reflected an emphasis on steals and unsuccessful 2 point field-goals and a deemphasis on blocks (see Table 1). For level of competition, the obtained function in discriminant analysis was statistically significant ( p I .05) and had an overall percentage of successful classification of 84.0%. The structure coefficients from the function reflected an emphasis on assists and a de-emphasis on turnovers (see Table 1).

1234

J. SAMPAIO, ET AL.

TABLE 1 MEANS,STANDARD ERRORS, AND DISCRIMINANT ANALYSIS STRUCTURE COEFFICIENTS (SC) FROMGAME-RELATED STATISTICS BY SEXAND LEVEL OF COMPETITION Game-related Statistic

2-pt successful 2-pt unsuccessful 3-pt unsuccessful 3-pt successful Free-throw successful Free-throw unsuccessful Offensive rebounds Defensive rebounds Assists Fouls Steals Turnovers Blocks

SC

Sex Men

Women

M

SE

28.6 31.6 18.2 9.5 23.5 9.2 16.3 33.3 17.7 3 1.4 16.7 15.3 4.1

.56 .68 .58 .37 .83 .42 .54 .58 .45 .63 .53 .47 .23

M

SE

SC

Level Senior

M

SE

Junior

M

SE

29.9 .73 36.0 .88 15.2 .75 7.0 .47 20.7 1.06 9.2 .54 15.6 .69 3 1.9 .75 13.9 .58 27.4 .81 22.4 .68 15.1 .60 2.0 .29

Wilks Lambda Eigenvalue Canonical correlation *p I .05. tlSCl 2 30. TABLE 2 MEANS,STANDARD ERRORS, AND DISCRIMINANT ANALYSIS STRUCTURE COEFFICIENTS (SC) FROMGAME-RELATED STATISTICS I N THE INTERACTION OF SEXAND LEVEL OF COMPETITION Sex x Level

Game-related Statistic Men

2-pt successful 2-pt unsuccessful 3-pt unsuccessful 3 -pt successful Free-throw successful Free-throw unsuccessful Offensive rebounds Defensive rebounds Assists Fouls Steals Turnovers Blocks Wilks Lambda Eigenvalue Canonical correlation "p I .05. tlSCl 2 30.

Senior

Junior

M S E

M S E

Women Senior Junior

M S E

M S E

Structure Coefficients Function 1 2 3

27.6 .84 32.2 1.02 20.0 .87 11.4 .55 26.2 1.23 9.0 .62 14.7 .80 33.7 .87 23.7 .67 34.0 .94 17.6 .79 9.5 .70 4.8 .34 .12* ,42* .76" 2.6 .8 .3 .85 .67 .49

1235


For sex of players x level of competition, the obtained functions in discriminant analysis were all statistically significant ( p I .05) and classified correctly 83.5% of the cases. Discriminant Function 1 accounted for 69.4% of the variance, Discriminant Function 2 accounted for 22.2%, and Discriminant Function 3 the remaining 8.5%. The structure coefficients from Function 1 reflected an emphasis on turnovers and a de-emphasis on assists. In Function 2 the emphasis was on steals and unsuccessful 2 point field-goals and the de-emphasis was on blocks. Finally, in Function 3 the emphasis was on unsuccessful 3-point field-goals, unsuccessful free-throws, and offensive rebounds and the de-emphasis was on successful 2-point field-goals and assists (see Table 2). In Fig. 1 are presented the contrasts in group centroid distances to describe the game-related statistical profiles that differentiate between the four groups (sex x level of competition). Discriminant Function 1

Women Junior Women Senior

.

Men Senior

Men Junior

3,OO

-

.

CV C

.-0

Women Junior

C,

0 C

Women Senior

3 LL

z

m .-C E .-

0,oo

-

.

rn

Men Junior

b

Men Senior

U)

5 -3,OO -3,OO

I

1

0,OO

3,OO

Discriminant Function 3 FIG. 1. Territorial map of the groups relative to their sex and level of competition. The points indicate the group centroid.

1236

J. SAMPAIO, ET AL.

DISCUSSION The purpose of the present study was to identify the basketball gamerelated statistics that best discriminate performances by sex of players and level of competition. The known differences in several performance dimensions, e.g., cognitive, perceptual, motor, psychological, should configure each team performance in a game situation (Kioumourtzoglou, Derri, Tzetzis, & Theodorakis, 1998) and be reflected in game-related statistics; however, these effects are unknown. The normalization of game-related statistics to 100 ball possessions allowed to control for game rhythm and to compare them, even when gathered in different games. This methodological approach may have increased the validity of the subsequent analysis. However, is very difficult to establish comparisons across the available research because the normalization of game-related statistics is a recent methodological adjustment and was only used previously by Ibiifiez, et al. (2003) and Sampaio and Janeira (2003). In these two studies, the field goals, the rebounds, and the fouls were identified as the most discriminative game-related statistics between winning and losing teams. Thus, it could be expected that men's senior teams could have better performances in these game-related statistics. However, one interesting finding in our results was the lack of discriminant power for all of these gamerelated statistics. Men's teams performance was best discriminated by their higher percentage of blocks and lower percentage of steals and unsuccessful 2-point fieldgoals. The effect of sex of players is probably determined by physical performance differences that seem to configure game tactics and strategies. Physical performance differences between men and women are known and attributable to a variety of factors that are not mutually exclusive. The most common are the anthropometric characteristics (height, weight, body proportion, and composition) and differences in daily physical activity (Durkin, 1987; Derri, Kioumourtzoglou, & Tzetzis, 1998; Kioumourtzoglou, et al., 1998). Because men are taller and have a higher center of gravity, they may be less focused on stealing the ball, which usually is an action closer to the floor and more focused on blocking a field goal, an action more common for taller players. Conversely, the lower percentage of unsuccessful 2-point field goals for men's teams is the results also of a lower percentage of attempted field goals; otherwise there would be statistically significant differences in successful 2-point field goals. The previously discussed results from steals and blocks may help to explain this fact, i.e., women's teams attempt more 2-point field goals compared to men's because of their higher percentages of steals and lower percentages of blocks. Senior teams were discriminated from junior teams by their higher percentage of assists and lower percentage of turnovers. Basketball players are


1237

systematically confronted with very complex and rapidly changing environments. The player is required rapidly to gain information from the ball, teammates and opponents and decide on an appropriate response with time and space pressure (Williams, 2000). In these situations, the player's ability to anticipate events from previously perceived components of an action sequence is part of a skilled performance. In several game situations, skilled players perform unhurried actions as a result of their increased ability to "read the game" (Helsen & Starkes, 1999; Williams, 2000). For example, a skilled centre can secure an offensive rebound near the basket and, based on contextual information, decide to pass the ball to the guard to restart the offensive phase. Probably, the response in this game situation of a less skilled centre could be attempting a field-goal that would be unsuccessful or be blocked by a defender. In addition, senior teams were discriminated by their mastery of ball-handling skills (assists and turnovers) and teamwork (assists). In fact, successful offences have the goal of selecting carefully field-goal opportunities and are dependent upon the quality of player's decision-making and field-goal execution as well as upon team coordination. Probably, senior players are faster and more accurate in recognizing and recalling patterns of play, anticipate better opponents' actions, have more effective and directed visual-search strategies, and are more accurate in their expectations regarding the outcome of specific contexts (Allard, Graham, & Paarsalu, 1980; Abernethy & Russell, 1984; Helsen & Starkes, 1999; Williams, 2000). It seems clear that these characteristics have stronger implications for dribbling and passing the ball during the offensive phase than for field-goal attempts. In fact, field-goals can be considered the result of the quality of all precedent actions. In the study of the interaction of sex x level of competition, the gamerelated statistics that best discriminated among the four groups (Function 1, 69.4% of the variance) were the same as previously identified for the level of competition. Additionally, but with less importance (Function 2, 22.2 % ) , were the unsuccessful 2-point field-goals, assists, and steals. These results reinforce the importance of assists and turnovers and strongly suggest that a very important training time should be directed to tasks that integrate perceptual, decision, and motor components. The ability of the discriminant functions in correctly classifying the teams in their respective groups was high, denoting the quality of the discriminant functions and the power of the structure coefficients in explaining variability amongst groups. This way, it seems very appropriate to evaluate teams' game performance according to normative data for sex of players and level of competition. The territorial map (see Fig. 1) provided an initial model for discriminating these high-level basketball teams. The identification of the sport skills underlying high-level performances

1238

J. SAMPAIO, ET AL.

in these contexts can be information of great value to sport coaches and more specifically to the basketball coaches, allowing them to establish and monitor playing patterns. Also, it can help coaches understand better the differences between players-different sex of players and level of competition -and better regulate training sessions and competition. REFERENCES ABERNETHY, B., & RUSSELL, D. G. (1984) Advance cue utilization by skilled cricket batsmen. Australian Journal of Science and Medicine in Sport, 16, 2-10. AKERS, M. D., WOLFF,S., &BUTTROSS, T. (1991) An empirical examination of the factors affecting the success of NCAA Division I college basketball teams. The Journal of Business and Economic Studies, 1(2),57-7 1. ALLARD, F., GRAHAM, S., & PAARSALU, M. L. (1980) Perception in sport: basketball. Journal of Sport Psychology, 2, 14-21. E., &TZETZIS,G. (1998) Assessment of abilities in basketball: a DERRI,V., KIOUMOURTZOGLOU, preliminary study. Perceptual and Motor Skills, 87, 91-95. K. (1987) Social cognition and social context in the construction of sex differences. DURKIN, In M. A. Baker (Ed.), Sex d2jrferences in human pevformance. Chichester, U K : Wiley. Pp. 141-170. FRANKS, I., & MILLER,G. (1986) Eyewitness testimony in sport. Journal of Sport Behavior, 9, 39-45. HELSEN, W. F., &STARKES, J. L. (1999) A multidimensional approach to skilled perception and performance in sport. Applied Cognitive Psychology, 13, 1-27. Isbrj~z,S., SAMPAIO, J., SAENZ-LOPEZ, P., G I M ~ N EJ., Z ,& JANEIRA, M. (2003) Game statistics discriminating the final outcome of junior world basketball championship matches (Portugal 1999). Journal of Human Movement Studies, 45, 1-19. KARIPIDIS, A., FOTINAKIS, P., TAXILDARIS, K., &FATOUROS, J. (2001) Factors characterizing a successful performance in basketball. Journal of Human Movement Studies, 41, 385-397. KIOUMOURTZ~GLOU, E., DERRI,V, TZETZIS, G., &THEODORAKIS, Y. (1998) Co nitive, perce tual, and motor abilities in skilled basketball performance. Perceptual a n f Motor ski& 86, 77 1-786. MACDONALD, N. (1984) Avoiding the pitfalls in player selection. Coaching Science Update, 5, 41-45. MELNICK, M. J. (2001) Relationship between team assists and win-loss record in the National Basketball Association. Perceptual and Motor Skills, 92, 595-602. NORUSIS,M. (1993) SPSS for Windows release 6.0. Chicago, IL: SPSS, Inc. OLIVER,D. (2004) Basketball on study: rules and tools for performance analysis. Dulles, NY: Brassey's Inc. SAMPAIO, J., & JANEIRA, M. (2003) Statistical analyses of basketball team erformance: understandin teams' wins and losses according to a different index of b& possessions. internationa?loumal of Per/ormance Analyiis in Sport, 3, 40-49. TABACHNICK, B., ELFIDELL, L. (1989) Using multivariate statistics. New York: Harper & Row. WILLIAMS, A. M. (2000) Perceptual skill in soccer: implications for talent identification and development. Journal of Sports Sciences, 18, 737-750.

Accepted November 9, 2004