Factors Associated with Success Among NBA Teams

 

Abstract

Data from the 1997-1998 National Basketball Association (NBA) regular season were analyzed to determine factors that best predicted success, as measured by winning percentage. A total of 20 variables were examined. A multiple regression analysis revealed that field goal conversion percentage was the best predictor of success, explaining 61.4% of the variance in winning percentage. The average three-point conversion percentage of the opposing teams explained a further 18.9% of the variance. These two variables combined explained 80.3% of the variance in winning percentage. The finding pertaining to field goal conversion percentage suggest that the attainments of the offense are more important than are the defensive attainments in predicting the success levels of NBA teams. These and other implications are discussed.

Introduction

The game of basketball was invented in December 1891 by Dr. James A. Naismith while an instructor in the physical training department of the International Young Men’s Christian Association (YMCA) Training School in Springfield, Massachussets (Fox, 1974). Naismith’s goal was to answer the challenge of Dr. Luther H. Gulick, his department head, who wanted an indoor game to be invented that (1) would attract young men during the winter, when baseball and football were out of season, and (2) would replace gymnastics and calisthenics, which provoked little interest (Fox, 1974). Naismith, known as “the father of basketball,” incorporated features of soccer, U.S. football, rugby football, field hockey, and other outdoor sports in developing the game of basketball.

By 1946, professional basketball had acquired a large and faithful following among U.S. sports fans, who wanted to watch their former collegians in action. During this period, there was the American Basketball League (ABL) on the East Coast and the National Basketball League (NBL) in the Midwest. In June, 1946, the Basketball Association of America was formed, which effectively replaced the ABL and competed directly with the NBL (Fox, 1974). The BAA and the NBL merged in 1950 as the National Basketball Association (NBA), comprising 17 teams. The NBA was reduced to 10 teams in 1951, as 7 NBL teams with marginal franchises dropped out (Fox, 1974). However, in the 1970s, the NBA expanded to 22 teams. Presently, the NBA contains 29 teams, with 15 teams in the Eastern Conference (with 7 teams representing the Atlantic division and 8 teams representing the Central division) and 14 teams in the Western Conference (with 7 teams representing the Midwest division and 7 teams representing the Pacific division). Basketball is now one of the most popular sports in the United States. Indeed, in the 1997-1998 season (the last time a full 82-game season was played), a total of 8,877,309 people attended an NBA game (The Sports Network, 1998), with an average attendance of 17,135 people per game (USATODAY, 1999).

Currently, at the end of the regular season, that is, when each team has played 82 matches, the top eight teams in each conference qualify for the playoffs. These eight teams then participate in a knockout tournament with the eventual winners of this stage within each conference advancing to the NBA finals. Because the teams which advance to the playoffs are those that have the highest winning percentages in their respective divisions during the regular season, knowledge of factors which predict success during this period would be of educational value for NBA coaches and analysts. Indeed, the former group could use this information to target coaching interventions.

Basketball is abound with empirical facts. Surprisingly, however, only descriptive statistics (e.g., averages, totals, percentages) tend to be utilized. Conversely, few inferential statistical analyses are undertaken on NBA data. Yet, such analyses provide consumers with information regarding the relationships among variables. As such, inferential statistics can yield very detailed and important information to consumers of professional basketball. Moreover, inferential statistics can be used to determine factors that predict the performance levels of teams.

To date, only a few studies have investigated correlates of basketball-related performance. Of those that have, the majority have involved an examination of psychological antecedents of basketball performance. For example, Whitehead, Butz, Vaughn, and Kozar (1996) found that increased stress (assumed to be present in games as opposed to practices) among members of an NCAA Division I men’s varsity team was associated with longer pre-shot preparations and a greater incidence of overthrown shots.

Newby and Simpson (1994) reported (1) a statistically significant negative relationship between minutes played by a sample of men and women college basketball players and mood, (2) a statistically significant negative relationship between the number of assists and depression, (3) a statistically significant negative relationship between the number of turnovers committed and mood, and (4) a statistically significant positive relationship between the number of turnovers committed and degree of tension. The researchers concluded that success in basketball is negatively related to psychopathology.

Both Pargman, Bender, and Deshaires (1975) and Browne (1995) found no relationship between free-throw and field goal shooting and field independency/field dependency. Additionally, Shick (1971) found no relationship between hand-eye dominance and depth perception and free-throw shooting ability in college women. Hall and Erffmeyer (1983) examined the effect of imagery combined with modeling on free-throw shooting performance among female college basketball students. These researchers noted that players who shot free throws under the conditions of videotaped modeling combined with relaxation and imagery were significantly more accurate than were those who shot in the relaxation and imagery condition only.

All the above studies investigated correlates of specific basketball skills (e.g., free-throw shooting), and, with a few exceptions (e.g., Butz et al., 1996), these skills typically were examined under simulated conditions. Such studies, although interesting, have limited utility for basketball coaches, in particular, because they does not provide any information as to why or how a team wins a basketball game. Indeed, the only inquiry found determining factors associated with success among basketball players was that of Steenland and Deddens (1997). These researchers studied the effects of travel and rest on performance, utilizing the results for 8,495 regular season NBA games over eight seasons (1987-1988 through 1994-1995). Findings revealed a statistically significant positive relationship between the amount of the time that elapsed between games and performance level. Specifically, more than 1 day between games was associated with a mean increase of 1.1 points for the home team and 1.6 points for the visitors. Peak performance occurred with 3 days between games. The researchers theorized that the negative effects of little time between games may be due more to insufficient time for physical recovery than to the effects of circadian rhythm (i.e., jet lag). However, although not statistically significant, they also found that visiting teams performed four points better, on average, when they traveled from the west coast to the east coast than when they traveled form east to west.

Surprisingly, no other study has investigated predictors of success among NBA teams. Even more surprising is the fact that no research appears to have examined what factors directly associated with skill level (e.g., field goal conversion percentage) best predict a team’s winning percentage. This was the purpose of the present inquiry. A secondary goal was to determine whether offensive or defensive factors would have more predictive power. It was expected that knowledge of these factors could help coaches to decide where to focus their attention, as well as assist analysts and fans in predicting a team’s performance.

Method
The data comprised all 21 unique team-level variables (when both team averages and totals were presented, only the averages were utilized, since they rendered totals redundant) that were presented on the official NBA website (i.e., http://www.nba.com) for the 1997-1998 regular professional basketball season. (The 1997-1998 NBA season was chosen because it represented the last time a full 82-game season was played.) These variables comprised winning percentage, which was treated as the dependent measure and 20 other variables which were utilized as independent variables. All variables are presented in Table 1. Scores pertaining to each variable for each team were analyzed using the Statistical Package for the Social Sciences (SPSS; SPSS Inc., 1999).

Table 1
Pearson Product-Moment Correlations of Winning Percentage and Selected Variables for the 1997-1998 Regular NBA Season
Variable   Winning
Percentage 
three-point conversion percentage .38  
field goal conversion percentage .78* 
free-throw conversion percentage .03  
average number of offensive rebounds per game -.31 
average number of defensive rebounds per game .47  
number of total rebounds .19  
average number of assists per game .61*  
average number of steals per game .08 
average number of blocks per game   -.13 
number of points scored per game .57* 
field goal conversion percentage of the opposing teams -.68* 
average three-point conversion percentage of the opposing teams -.50  
average free-throw conversion percentage of the opposing teams .18  
average number of offensive rebounds per game of the opposing teams -.49  
average number of defensive rebounds per game of the opposing teams   -.71* 
average number of total rebounds of the opposing teams -.69*  
average number of assists per game of the opposing teams -.70*  
average number of steals per game of the opposing teams -.45  
average number of blocks per game of the opposing teams -.58*   
average number of points scored per game of the opposing teams -.70*  
* statistically significant after the Bonferroni adjustment

Results and Discussion
Table 1 presents the correlations between winning percentage and each of the selected variables. It can be seen that, after adjusting for Type I error (i.e., the Bonferroni adjustment), winning percentages increased with field goal conversion percentage, number of assists per game, and number of points scored per game, and decreased with field goal conversion percentage of the opposing teams, average number of defensive rebounds per game of the opposing teams, average number of total rebounds per game of the opposing teams, average number of assists per game of the opposing teams, average number of blocks per game of the opposing teams, and average number of points per game of the opposing teams.

An all possible subsets (APS) multiple regression (Thompson, 1995) was used to identify which combination of independent variables best predicted NBA teams’ success. Again, success was measured by NBA teams’ regular season winning percentages. For this study, the criterion used to determine adequacy of the model was the maximum proportion of variance explained (i.e., R2), which provides an important measure of effect size (Cohen, 1988). Specifically, all variables were included except for those that represented (1) the total number of points scored or the total number of rebounds (use of the number of defensive rebounds and offensive rebounds rendered use of the total number of rebounds redundant). Consequently, a total of 16 independent variables were analyzed.

The multiple regression analysis revealed that the following two variables made a statistically significant contribution (F [2, 26] = 53.12, p < .0001) to the model: field goal conversion percentage and average three-point conversion percentage of the opposing teams. The regression equation was as follows:

winning percentage =
-159.53 + {(7.90) X field goal conversion percentage} – {(4.24) X average three-point conversion percentage of the opposing teams}

The regression equation indicates that every 1 percentage increase in field goal conversion rate is associated with a 7.90% increase in winning percentage. The confidence interval corresponding to this variable suggests that we are 95% certain that every 1 percentage increase in field goal conversion rate is associated with an average increase in winning percentage of between 6.00% and 9.80%. Additionally, every 1 percentage increase in the three-point conversion rate of the opposing teams is associated with a 4.24% decrease in winning percentage (95% confidence interval is 2.49% to 5.99%).

With respect to predictive power of the model, field goal conversion percentage explained 61.4% of the variance in winning percentages, whereas average three-point conversion percentage of the opposing teams explained 18.9%. These two variables combined to explain 80.3% of the total variance in winning percentage (adjusted R2 = 78.8%). In the study of human behavior, this percentage is extremely large, suggesting that an NBA team’s success can be predicted with an excellent degree of accuracy.

Conclusions
The purpose of this study was to determine which variables best predict whether an NBA team’s success rate. The finding that field goal conversion percentage explains more than three times the variance in success than does the average three-point conversion percentage of the opposing teams suggests that the attainments of the offense are more important than are the defensive attainments in predicting whether an NBA team will be successful. Thus, the present finding is in contrast to Onwuegbuzie (1999a), who identified four multiple regression models which adequately predicted the winning percentages of National Football League (NFL) teams for the 1997-1998 regular football season–the most notable being a two-variable model comprising turnover differential (which explained 43.4% of the variance in success) and total number of rushing yards gained by the offense (which explained a further 9.3% of the variance). Based on these models, Onwuegbuzie concluded that, outside the 20-yard zone, the attainments of the defense are more important than are the offensive attainments in predicting whether an NFL team is successful.

The present result pertaining to NBA teams also is in contrast to Onwuegbuzie’s (1999b) replication study of NFL teams for the 1998-1999 football season in which a model was identified containing the following five variables: (1) turnover differential (which explained 54.4% of the variance); (2) total number of rushing yards conceded by the defense (which explained 21.3% of the variance); (3) total number of passing first downs attained by the offense (which explained 9.4% of the variance), (4) percentage of third-down plays that produce a first down (which explained 4.1% of the variance), and (5) total number of penalties conceded by the opponents’ defense resulting in a first down (which explained 4.1% of the variance). Onwuegbuzie concluded that defensive gains are better predictors of success than are offensive gains because the first two variables, which explained more than 75% of the variance, were characteristics of the defense.

The finding that field goal percentage rate explained a very large proportion of the variance in success (i.e., 61.4%) highlights the importance of offensive efficiency not only of the starting players but also of the “bench” players, since the latter group also contribute to the field goal percentage rate. Nevertheless, the fact that three-point conversion percentage also made a contribution to the regression model, albeit a smaller one, suggests the importance of teams forcing the opposition to hurry their three-point shots and to take these shots from non-optimal parts of the basketball court.

Although a significant proportion of the variance in winning percentage was explained by the selected variables, this study also should be replicated using data from other seasons. Furthermore, regression models should be fitted using college basketball data. Information from such analyses should help coaches and analysts alike to obtain objective data which can be used to monitor the performance of NBA teams.

References

Browne, G.S. (1995). Cognitive style and free throw shooting ability of female college athletes. Unpublished master’s thesis, Valdosta State University, Valdosta, Georgia.

Cohen, J. (1988) Statistical power analysis for the behavioral sciences. New York: Wiley.

Fox, L. (1974). Illustrated history of basketball. New York, NY: Grosset & Dunlap.

Hall, E.G., & Erffmeyer, E.S. (1983). The effect of visuo-motor behavior rehearsal with video taped modeling of free-throw shooting accuracy of intercollegiate female basketball players. Journal of Sport Psychology, 5, 343-346.

Newby, R.W., & Simpson, S. (1994). Basketball performance as a function of scores on profile of mood states. Perceptual and Motor Skills, 78, 1142.

Onwuegbuzie, A.J. (1999a). Defense or Offense? Which is the better predictor of success for professional football teams? Perceptual and Motor Skills, 89, 151-159.

Onwuegbuzie, A.J. (1999b, November). Is defense or offense more important for professional football teams? A replication study using data from the 1998-1999 regular football season. Paper presented at the annual meeting of the Midsouth Educational Research Association, Point Clear, AL.

Pargman, D., Bender, P., & Deshaires, P. (1975). Correlation between visual disembedding and basketball shooting by male and female varsity athletes. Perceptual and Motor Skills, 41, 956.

Shick, J. (1971). Relationships between depth perception and hand-eye dominance and free-throw shooting in college women. Perceptual and Motor Skills, 33, 539-542.

SPSS Inc. (1999) SPSS 9.0 for Windows. [Computer software]. Chicago, IL: SPSS Inc.

Steenland, K., & Deddens, J.A. (1997). Effect of travel and rest on performance of professional basketball players. Sleep, 20(5), 366-369.

The Sports Network. (1998). Statistics: 1997-1998 NBA attendance. The Sports Network, 21(21).

Thompson, B. (1995). Stepwise regression and stepwise discriminant analysis need not apply here: A guidelines editorial. Educational and Psychological Measurement, 55, 525-534.

USATODAY. (December 28, 1999). Inside the numbers. Retrieved January 28, 2000 from the World Wide Web: http://www.usatoday.com/sports/basketba/skn/numbers.htm.

Whitehead, R., Butz, J.W., Vaughn, R.E., & Kozar, B. (1996). Stress and performance: An application of Gray’s three-factor arousal theory to basketball free-throw shooting. Journal of Sport Behavior, 19(4), 354-364.

Footnote
1 Due to space constraints, the intercorrelations among all the variables is not presented. However, this can be obtained by contacting the author.


Address correspondence to Anthony Onwuegbuzie, Department of Educational Leadership, College of Education, Valdosta State University, Valdosta, Georgia, 31698 or e-mail (TONWUEGB@VALDOSTA.EDU).

Print Friendly