The Efficacy of Statistics in Basketball:
A Review of the Literature
Introduction
The increase in technology over recent years has led to
its increased use in sports. One of the most common ways to use technology is
to attempt to more efficiently analyze the game of basketball. Statistical
analysis can be used individually or as a team in order to best understand the
efficiency being offered in each facet of the game. Basketball has become
increasingly reliant on statistics in games and in strategy and preparation.
Advanced statistics are becoming more popular. It is important to understand
the scope in which statistical analysis can be effective. The economist Aaron
Levenstein said, “Statistics are like bikinis. What they reveal is interesting,
but what they conceal is vital” (as cited in Popik, 2011). Statistics can simultaneously
allow individuals to see what they are looking for and hide that which they do
not want to see. Previous research dedicated to the use of statistics in
basketball ultimately all points toward three purposes:
1.
Predict outcomes.
2.
Establish factors of success.
3.
Understand form and function.
Using statistics to predict outcomes is the least researched but could
have major contributions to the game of basketball. This is the purpose that
this proposed research study would look into further.
Predict Outcomes
Little research has been completed with the purpose of
understanding the predictive ability of statistics in basketball. Essentially,
Ken Pomeroy (2012) is the sole contributor to this purpose. Pomeroy uses concepts
developed by Dean Oliver in his book, Basketball
on Paper (2004), to rank all NCAA Division I teams throughout each season
and ultimately to attempt to predict the national championship. Using
statistical categories such as effective field goal percentage, turnover
percentage, offensive rebounding percentage, free throw rate, and others,
Pomeroy describes the effectiveness of both teams and individuals in an effort
to predict their success over the course of a season. Each number is calculated
per 100 possessions in order to eliminate the differences that may occur as a
result of pace of play. The numbers are also adjusted to account for competition.
Pomeroy estimates the points a team would score per 100 possessions against an
average Division I team using previous box score data and refers to that number
as Adjusted Offensive Efficiency. Similarly, he estimates the points a team
would allow per 100 possessions, known as Adjusted Defensive Efficiency. The
other major category Pomeroy uses is the Adjusted Tempo, which describes how
many possessions a team averages per 40 minutes against other teams wanting to
play at a Division I tempo. A Division I tempo averages approximately
sixty-seven possessions per game (Pomeroy, 2012). Ultimately, Pomeroy ranks
teams according to their Pythagorean Winning Percentage, which is calculated
using the following formula:
Pyth = (AdjO^10.25)/(AdjO^10.25+AdjD^10.25)
where AdjO is Adjusted Offensive Efficiency and AdjD is Adjusted Defensive
Efficiency. (Pomeroy, Ratings Glossary, 2012)
Pomeroy’s attempt to predict the success of
teams throughout seasons is unmatched and it is his research that this study
will attempt to emulate at a different level.
Establish Factors of Success
Much research completed with the
goal of establishing factors of success is done on a post hoc basis.
Researchers look back on previous statistics in an attempt to explain why what happened
actually happened. Dean Oliver published an entire book, Basketball on Paper (2004), which explained how to keep tempo free
statistics and the purpose for which they could be used. Oliver shows how statistics
can highlight effectiveness of individual players, strategy, chemistry, and the
team’s effectiveness as a unit. Oliver provides several formulas for analyzing
the game of basketball that evaluate players and teams statistically and
objectively, instead of the subjective analysis so often used throughout the
game. Oliver also wrote an article that goes into further details about the
four main factors of successful basketball analysis that were initially
outlined in Oliver’s Basketball on Paper
(2004). The four factors are identified as shooting, turnovers, rebounding, and
free throws. Ultimately, successful teams are more efficient than their
opponents by having better numbers in each of these four factors. By breaking
these factors down using a self-made computer program called Roboscout, Oliver
is able to explain how statistics impact building a team and developing game
plans for specific opponents.
Csatalijay, James,
Hughes, and Dancs (2012) conducted a study on the differences between winning
and losing close quarters in a basketball game. The objective of this study was to compare the
performance differences between winning and losing teams over the course of
four quarters in relation to the closeness of the score. This study analyzed
the results of each of the four quarters played over the course of twenty-six
games by a Hungarian 1st division basketball team in order to
understand the statistical categories that result in the biggest difference
between winning and losing in each quarter of a game. Close quarters were those
with a 0-5 point differential. Balanced quarters had a 6-11 point differential.
Unbalanced quarters had a 12-22 point differential. Quarters that were tied
were not analyzed. Quantitative analysis through SPSS showed that there were 20
different statistical categories that were statistically significant in
determining success within quarters. When the data was limited to only close
quarters, five significant factors became apparent: number of successful free throws, number of
defensive rebounds, number of total rebounds, percentage of offensive rebounds,
and percentage of defensive rebounds. This study provided ideas into which
statistical categories to specifically consider in addition to others that
other research may support.
In his presentation on statistics in
basketball, Coach Mike Neighbors provided a spreadsheet of data illustrating
the statistical categories in which national champions most often led the
nation. Neighbors collected data from ncaa.org over an ongoing period from
2002-present in order to determine the statistical category that was most
likely to predict a national championship. His research concluded that a team’s
field goal percentage offense was most important in winning a national
championship, although it was not the exclusive predictor. Other factors that
proved to play a role included field goal percentage defense, free throw
percentage, rebound rate, and turnover rate. This is an invaluable resource in
considering additional statistical predictors of success in the game of
basketball and in considering different perspectives to approach the usefulness
of statistics as successful predictors.
Understand Form and Function
The use of statistics to
understand form and function allows for coaches and players to practice and
prepare for what matters, and make corrections in process and form.
Fontanella’s book, The Physics of
Basketball (2006),
uses the concepts of physics to explain the sport of basketball and how to play
the game most effectively according to the laws prescribed by physics. Using
forces, gravity, acceleration, collisions, angles, and other ideas, the author
gives evidence as to the best way to shoot and other necessary concepts related
to the game. The book gets into specifics such as the most effective places at
which one should hit the backboard and where to aim on the rim so that the ball
is given the highest probability of bouncing through the hoop. Overall, this
book provides physical evidence that can be incorporated into statistical
research to either support or negate the mathematical evidence.
Sampaio (2004) conducted a study
with the purpose of determining how sex of players and the level of competition
affect basketball statistics. This study used data collected by the
International Basketball Federation from the World Championships at the men’s
senior, men’s junior, women’s senior, and women’s junior championships. Typical
box score categories were included in the analysis. The authors found that
men’s teams had a higher percentage of blocks, lower percentage of steals, and
lower percentage of unsuccessful two-point field goals as opposed to women’s
teams. For the level of competition, junior teams illustrated a lower
percentage of assists and a higher percentage of turnovers when compared to
senior teams. The main discriminatory factor in basketball statistics proved to
be the level of competition, according to this study.
Maymin, Maymin, and Shen (2012)
learned more about factors affecting successful free throws in their shooting. This
physics-based study aimed to determine the influence of different trajectory
parameters on an individual’s ability to successfully shoot free throws. By
identifying key areas of focus in successful free throws, the authors could
provide guidelines by which free throw shooting could be improved easier and
faster than previously thought. This study used tracking data from STATS LLC
over the course of 158 National Basketball Association games in the 2010-2011
season. STATS LLC uses installed technology on the basketball goal in order to
collect data. The five factors leading to efficient free throw shooting are the
height of the release, the launch velocity, the vertical launch angle, the
left-right deviation, and the backspin. This study demonstrates that
professional players struggle individually with these factors and that they do
not all struggle in the same area. When individuals can understand exactly
which area is giving them problems, the problem becomes easier to correct.
Arkes
(2010) aimed to better understand the “hot hand” concept in basketball,
specifically the evidence for the effect of making the first free throw on
making the second free throw. Arkes used data collected by 82games.com over the
course of the 2005-2006 National Basketball Association season. Variables
accounted for included the day of the season, the player, whether the player
made the free throw, the place of that shot in the set order of free throws,
the quarter in which the free throw took place, and how many free throws the
player had made in previous attempts. The data showed that a successful prior
free throw leads to a 2-3% increase in the probability of making the second
free throw, which supports the idea of a “hot hand”. This seemed to be stronger
for frequent free throw shooters as opposed to those who went to the line more
sporadically. Despite the fact that this study only used data specifically for free
throws, it does suggest the reality of the hot hand phenomenon.
Use of Technology
One of the primary reasons that statistical analysis is
becoming more popular in the field of athletics is the advent of applicable
technology. Plenty of software and Internet sources allow for statistical
analysis of sport. Major college coaches are becoming more open to the use of
advanced statistical analysis. Proponents of such analysis include Brad Stevens
of Butler, Roy Williams at the University of North Carolina, and Mike
Krzyzewski of Duke (Phelps, 2012). Phelps provides several examples of
companies dedicated to furthering the use of statistical analysis in basketball
such as Krossover, TurboStats, and DakStats (2012). Each of these is geared
toward use by coaches for measuring the efficiency of their players and teams
as a whole. Krossover separates itself from TurboStats and DakStats by using
video uploaded by coaches in order to complete the statistical analysis, as
opposed to putting the burden on the coach to plug in numbers after each game.
The use of Krossover will be integral to the success of this research study, as
will be later discussed.
KenPom is a different type of technology used to do more
than just calculate efficiency. Using tempo free stats and advanced statistics
prescribed by Oliver (2004), Pomeroy ranks each NCAA Division I team on the
objective basis of their statistics and attempts to predict the national
champion of the NCAA Division I Men’s Basketball tournament each year according
to their advanced statistics accumulated over the course of the season (2012).
Pomeroy’s technology and research is the only previous research that has
contributed to the predictive ability of statistics in basketball, which was
discussed in more detail in the “Predict Outcomes” section.
Methodology
In order to best
understand the efficacy of statistics in basketball and possibly learn more
about their predictive ability, this research study will examine previous box
scores and use statistical analysis in order to determine the variability
between results of each statistical category. The purpose of this research
study will be to determine the predictive ability of statistics in determining
winners of conference basketball games in the Wisconsin Intercollegiate Athletic
Conference (WIAC) women’s basketball games within the last five years.
The first step to take
in achieving the purpose is to prove the validity, objectivity, and reliability
of published box scores from the games. This will be done using the services provided
by Krossover Intelligence, Inc.. Krossover’s software provides statistical
analysis of game film by uploading the film on Krossover’s website then
receiving results within thirty-six hours post upload. Researchers will
randomly select three WIAC games played during each of the past five seasons.
Each of the three games will be uploaded to Krossover along with the rosters of
each team competing in the game. Krossover has agreed to provide twenty-five
free uploads toward the completion of this study so these services will be
complimentary. Upon receiving the statistical analysis provided by Krossover,
researchers will compare the numbers provided by Krossover to the published box
scores of each school’s Sports Information Director postgame. If there are no
statistically significant differences in the two sets of numbers, then the
published box scores will be proven valid, objective, and reliable. However, if
the numbers are significantly different, then the published box scores will not
be usable for research because the numbers are not valid, objective, or
reliable.
After the box scores are
proven to be reliable and consistent in the data provided, then the box scores
of all previous WIAC women’s basketball games can be collected and analyzed.
Researchers will calculate advanced tempo-free statistics for each game in
addition to the basic statistics already provided. Data will be inserted into
Excel and SPSS to calculate more advanced statistical categories using
algorithms initially published by Dean Oliver in Basketball on Paper (2004). Following data collection, data will be
analyzed to determine the variance in results. Describing the variance will
allow researchers to see the statistical categories that are most significant
in determining the winners of basketball games, if there are any.
Conclusion
Overall, the areas of
establishing factors of success and understanding form and function through
statistics seem to be the best researched in basketball statistics. The entire
field is quite under researched but it is a topic that is growing in popularity
with the continued growth of technology. This proposed study would contribute
to the topic by providing more research into the predictive ability of
basketball statistics as a whole and, more specifically, in the elite
conference of NCAA Division III women’s basketball.
Annotated Bibliography
Arkes, J. (2010). Revisiting
the hot hand theory with free throw data in a multivariate framework. Journal of Quantitative Analysis in Sports,
6(1), 1-10. doi: 10.2202/1559-0410.1198
This study aimed to better understand the “hot hand”
concept in basketball, specifically the evidence for the effect of making the
first free throw on making the second free throw. Arkes used data collected by
82games.com over the course of the 2005-2006 National Basketball Association
season. Variables accounted for included the day of the season, the player,
whether the player made the free throw, the place of that shot in the set order
of free throws, the quarter in which the free throw took place, and how many
free throws the player had made in previous attempts. The data showed that a
successful prior free throw leads to a 2-3% increase in the probability of
making the second free throw, which supports the idea of a “hot hand”. This
seemed to be stronger for frequent free throw shooters as opposed to those who
went to the line more sporadically. Despite the fact that this study only used
data specifically for free throws, it does suggest the reality of the hot hand
phenomenon.
Csataljay G., James, N.,
Hughes, M., & Dancs, H. (2012). Performance differences between winning and
losing basketball teams during close, balanced and unbalanced quarters. Journal of Human Sport & Exercise, 7(2),
7356-364.
The objective of this study was to compare the
performance differences between winning and losing teams over the course of
four quarters in relation to the closeness of the score. This study analyzed
the results of each of the four quarters played over the course of twenty-six
games by a Hungarian 1st division basketball team in order to
understand the statistical categories that result in the biggest difference
between winning and losing in each quarter of a game. Close quarters were those
with a 0-5 point differential. Balanced quarters had a 6-11 point differential.
Unbalanced quarters had a 12-22 point differential. Quarters that were tied
were not analyzed. Quantitative analysis through SPSS showed that there were 20
different statistical categories that were statistically significant in
determining success within quarters. When the data was limited to only close
quarters, five significant factors became apparent: number of successful free throws, number of
defensive rebounds, number of total rebounds, percentage of offensive rebounds,
and percentage of defensive rebounds. This study provided ideas into which
statistical categories to specifically consider in addition to others that
other research may support.
Fontanella, J. (2006). The Physics of Basketball. Baltimore,
MD: The Johns Hopkins University Press.
Fontanella’s book uses the concepts of physics to explain
the sport of basketball and how to play the game most effectively according to
the laws prescribed by physics. Using forces, gravity, acceleration,
collisions, angles, and other ideas, the author gives evidence as to the best
way to shoot and other necessary concepts related to the game. The book gets
into specifics such as the most effective places at which one should hit the
backboard and where to aim on the rim so that the ball is given the highest
probability of bouncing through the hoop. Overall, this book provides physical
evidence that can be incorporated into statistical research to either support
or negate the mathematical evidence.
Krossover Intelligence Inc.
(2012). Krossover: Get your game brain on. Retrieved from
krossover.com.
Maymin, A., Maymin, P., &
Shen, E. (2012). Individual factors of successful free throw shooting. Journal of Quantiative Analysis in Sports,
1-15. Retrieved from http://ssrn.com/abstract=1947166
This physics-based study aimed to determine
the influence of different trajectory parameters on an individual’s ability to
successfully shoot free throws. By identifying key areas of focus in successful
free throws, the authors can provide guidelines by which free throw shooting
can be improved easier and faster than previously thought. This study used
tracking data from STATS LLC over the course of 158 National Basketball
Association games in the 2010-2011 season. STATS LLC uses installed technology
on the basketball goal in order to collect data. The five factors leading to
efficient free throw shooting are the height of the release, the launch
velocity, the vertical launch angle, the left-right deviation, and the
backspin. This study demonstrates that professional players struggle individually
with these factors and that they do not all struggle in the same area. When
individuals can understand exactly which area is giving them problems, the
problem becomes easier to correct.
Neighbors,
M. (2012). Statistical Predictor of Wins:
Statistical Comparison. Address at the Women’s Basketball Coaches
Association National Convention in Denver, CO illustrated on spreadsheet.
In his presentation on statistics in
basketball, Coach Mike Neighbors provided a spreadsheet of data illustrating
the categories that were most predictive of success over the course of the
season. Neighbors collected data from ncaa.org over an ongoing period from
2002-present in order to determine the statistical category that was most
likely to predict a national championship. His research concluded that a team’s
field goal percentage offense was most important in winning a national
championship, although it was not the exclusive predictor. Other factors that
proved to play a role included field goal percentage defense, free throw
percentage, rebound rate, and turnover rate. This is an invaluable resource in
considering additional statistical predictors of success in the game of
basketball and in considering different perspectives to approach the usefulness
of statistics as successful predictors.
Oliver, D. (2004). Basketball on Paper. Dulles, VA:
Brassey’s Inc.
This is an entire book on the different types of
performance analysis that can be used throughout the sport of basketball. Using
a possession-based concept, Oliver shows how statistics can highlight
effectiveness of individual players, strategy, chemistry, and the team’s
effectiveness as a unit. Oliver provides several formulas for analyzing the
game of basketball that evaluate players and teams statistically and objectively,
instead of the subjective analysis so often used throughout the game.
Oliver, Dean (2004).
Roboscout and the four factors of basketball success. Journal of Basketball Studies. Retrieved from www.rawbw.com/~deano
This article goes into further details about the four
main factors of successful basketball analysis that were initially outlined in
Oliver’s Basketball on Paper (2004).
The four factors are identified as shooting, turnovers, rebounding, and free
throws. Ultimately, successful teams are more efficient than their opponents by
having better numbers in each of these four factors. By breaking these factors
down using a self-made computer program called Roboscout, Oliver is able to
explain how statistics impact building a team and developing game plans for
specific opponents.
Phelps, M. (2012). Strength
in numbers. Coaching Management, 20(6),
18-26.
This article discusses the way that advanced statistics
are growing in the game of basketball. Phelps identifies those coaches who use
advanced metrics best, including Brad Stevens at Butler. Additionally, he
defines and elaborates on some advanced statistics that are becoming more
popular within basketball, especially those stressed by Ken Pomeroy. Statistics
discussed include effective field goal percentage, turnover percentage,
offensive rebounding percentage, free throw rate, adjusted efficiency and
tempo. The article highlights the advantages of using advanced statistics.
These advantages include redefined positions, objective scouting, ability to identify
team tendencies, and the ability to isolate individuals. Finally, Phelps
reviews some ways to implement advanced metrics in a basketball program,
whether coaches choose to do it themselves or use analysis aids such as
Turbostats, Hudl, or Krossovr.
Pomeroy, Ken (2012). Advanced analysis in college basketball.
Retrieved from http://www.kenpom.com
The KenPom website is a website devoted to the advanced
statistical analysis of basketball. Ken Pomeroy, the author, uses concepts
developed by Dean Oliver in his book, Basketball
on Paper (2004), to rank all NCAA Division I teams throughout each season
and ultimately to attempt to predict the national championship. Using
statistical categories such as effective field goal percentage, turnover
percentage, offensive rebounding percentage, free throw rate, and others,
Pomeroy describes the effectiveness of both teams and individuals in an effort
to predict their success over the course of a season.
Popik, B. (2011). Statistics
are like bikinis—what they reveal is interesting, but what they conceal is
vital. The Big Apple. Retrieved from
www.barrypopik.com
Sampaio, J., Godoy, S.I.,
& Feu, S. (2004). Discriminative power of basketball game-related
statistics by level of competition and sex. Perceptual
and Motor Skills, 99, 1231-1238.
The purpose of this study is to determine the basketball
statistics that are affected by sex of players and the level of competition.
This study used data collected by the International Basketball Federation from
the World Championships at the men’s senior, men’s junior, women’s senior, and
women’s junior championships. Typical box score categories were included in the
analysis. The authors found that men’s teams had a higher percentage of blocks,
lower percentage of steals, and lower percentage of unsuccessful two-point
field goals as opposed to women’s teams. For the level of competition, junior
teams illustrated a lower percentage of assists and a higher percentage of
turnovers when compared to senior teams. The main discriminatory factor in
basketball statistics proved to be the level of competition, according to this
study.
