EVALUATING COLLEGE PROSPECTS AND THEIR POTENTIAL TO SUCCEED IN THE NATIONAL ASSOCIATION: IDENTIFIYING SIGNIFICANT QUANTIFIABLE MEASURES

A THESIS

Presented to

The Faculty of the Department of Economics and Business

Colorado College

In Partial Fulfillment of the Requirements for the Degree

Bachelor of Arts

By

Samuel L. Markin

May 2018

EVALUATING COLLEGE PROSPECTS AND THEIR POTENTIAL TO SUCCEED IN THE NATIONAL BASKETBALL ASSOCIATION: IDENTIFIYING SIGNIFICANT QUANTIFIABLE MEASURES

Samuel L. Markin

May 2018

Economics

Abstract

As more money is committed to players, it is more pivotal than ever for NBA teams to find ways to accurately and comprehensively find young, cheaper talent in the draft better than their competitors. In this study we use all publicly available information about players that is available. The focus of this paper is to examine factors that contribute to early career success among professional players in the National Basketball Association (NBA) and to better understand these quantifiable measures available on draft day that can aide in predicting players' future performance. In this study we six separate OLS regressions with different sections of the data. One regression with all the player data, then we separate the others by one and two year players, three and four year players, guards, forwards, and big men. The independent variables used in this study are player position ( guard, , small forward, power forward, ), the player's college win shares per year, win shares in their first year, average college win shares per year, quality of conference, NBA combine agility, combine no-step vertical leap, and dummy variables for when they came out of college. The dependent variable in all of the regressions is the average wins per year through five years of each players NBA career.

Key Words: National Basketball Association, Player Development, Sports Economics

JEL Codes: (L83, Z2, Z20)

ii

ON MY HONOR, I HAVE NEITHER GIVEN NOR RECEIVED UNAUTHORIZED AID ON THIS THESIS

Signature

iii

TABLE OF CONTENTS

ABSTRACT…………………………………………………………………….….. ii.

1. INTRODUCTION………………………………………………….……… 1.

2. Previous Literature and Theoretical Considerations………………….……. 3.

2.1 Win Shares………………………………………………………….. 7.

2.2 Formula for Crediting Win Shares……………………………...…... 9.

3. DATA & METHODOLOGY………………………………………..…...... 10.

3.1 Regression Equations…………………………………………...…… 14.

4. RESULTS & ANALASYS………………………………………………..… 16.

4.1 General Findings……………………………………………….…… 18.

4.2 All Player Data…………………………………………………….... 18.

4.3 One & Two Year Players…………………………………………..... 20.

4.4 Three & Four Year Players………………………………………….. 20.

4.5 Guards Wings and Big Men…………………………………………. 21.

5. CONCLUSION……………………………………………………………… 22.

6. REFERENCES………………………………………………………………. 24.

iv Introduction

In an age where football seems to be on the decline as America’s number one sport per viewership and revenue, engagement in basketball, specifically the National

Basketball Association has only increased. The NBA has become the sport people look to for superstar personalities. Players like Lebron James and Stephen Curry are continually in the news, either because of their incredible play, or because they are speaking on important social issues.

Total league revenue has jumped from 2.66 billion dollars in 2001-02 season to

5.87 billion dollars in 2015-16, and finally up to 7.37 billion just a year later from last season alone (Statista). As revenues for the league have increased, player salaries have increased as well with the average salary around six million for the 2017-2018 season

(basketball-reference).

We are now seeing average players like Tim Hardaway make (72 million dollars) on a 4-year contract. Teams seem to be reaching for free agent talent, and handing out large sums of money in the hopes that an unproven player pans out. While the large market teams might not need to be good to generate the most revenue in the league, other smaller market teams do not have the same luxury. Small market teams do not usually attract top free agents, so the only way to acquire a superstar talent without breaking the bank for overrated players is through the draft. This brings us to the discussion about what quantifiable measures we should be using as we look to evaluate players entering the NBA draft.

Many of these smaller market teams rely on their performance to drive ticket sales and revenue as well, despite the revenue sharing statues currently in place via the latest

1 collective bargaining agreement (CBA). In addition to this, tanking, or purposely losing in order to get a better spot in the draft, continues to be put under the microscope. It is discouraged to lose on purpose in order to benefit your future and so the league has considered implementing a new rule to make it so this does not take place. One potential solution that has been mentioned is that the bottom 4-8 teams will have an equal opportunity to get the top pick in the lottery.

As the stakes become higher, and more money is committed to players, it is more pivotal than ever for NBA teams to find ways to accurately and comprehensively find young, cheaper talent in the draft better than their competitors. In the latest version of the

CBA, rookies have a scaled contract for the first three years based on their draft position and there is a team option on the fourth year. Expiring rookie contracts also qualify for the exception that allows teams to exceed the cap to resign their own free agents. Teams are also able to offer their own free agents more than the max if they meet certain criteria.

Every year there are tens of billions of dollars in exchanged in the NBA between teams, players, and other corporate entities with a stake in the league’s success, involved through sponsorships and advertisements. The collective bargaining agreement has been put into place to help the smaller market teams stay competitive. If a franchise can consistently find hidden talent in the draft, these rules make it so that they have a chance to keep their talent and grow organically with a salary cap friendly roster.

2 Previous Literature and Theoretical Considerations

Although we have seen an increase in technology and data collection in the present day, it seems as though it has not been utilized to its full potential. Many analysts and scouts may still argue that statistics can only tell you so much, and I do agree to an extent. However, there is a reason players perceived as the best tend to have the best statistics as well. We have seen statistical analysis put to use effectively, and it has gained traction since the release of Michael Lewis’ highly publicized book, Moneyball (2003), a true story where Lewis uses statistical analysis to put together a winning team on a budget in the MLB. While I am not here to say that statistical analysis and statistical analysis alone should be used to evaluate players and decide on personnel, I do believe that it does have a place in the world of player evaluation and analysis. As we proceed we will look at previous studies that identify how players are likely to reach the NBA level, as well as studies aimed at predicting draft orders, and studies that use college statistics to predict success and evolution at the NBA level.

A previous study by Musch and Grondin (2001) suggests that factors such as being born on the right side of the age cutoff as a youth may contribute to an athlete being identified as better as a result of being more physically and cognitively mature at an early age, and thus are given more opportunities and coaching to improve to reach their potential, the professional level. This has come to be known as the “season-of-birth effect” or “relative age effect”. The effect has been identified in hockey, soccer, , gymnastics, and other sports. This study only shows who may be more likely to make it to the professional level, but tells us little about how the athlete will preform once they arrive.

3 There have been studies like the one of Groothius, Hill, and Perri (2007) that compares measures of NBA to draft spot of which they were chosen. They find in this study that draft position is negatively related to NBA efficiency in a given year.

This shows us that scouts and franchises do get player evaluation right for the most part, however it does not explain why certain players taken earlier in the draft tend to be better, and why there are some hidden gems taken later in the draft that excel at the NBA level.

A study by Berri, Brook, and Fenn (2011) showed that the most important performance variable for predicting draft order was scoring. Also in this study, many other variables were used to predict NBA performance, which includes rebounds, steals, shooting efficiency, and team success. Additionally, they find that college performance measures and age predict both draft status and performance, but differ in that other variables such as height and team success predict draft status but not NBA performance.

The youth advantage, which suggests that age predicts relative success, has also been confirmed by separate analyses (Rodenberg & Kim, 2011). This leads to the explanation that franchise scouts and evaluators believe that certain measures, like scoring, are better indicators than others to use in order to make rational decisions come draft time.

Many different papers have discussed the effectiveness of teams in professional sports leagues to pick valuable prospects in their respective amateur drafts. Coates and

Oguntimein (2008), Camerer and Weber (1999), and Groothuis, Hill, and Perri (2007a,

2007b), all study this effectiveness as it relates to the NBA. Coates and Oguntimein

(2008) were the first to use college performance to predict draft performance and NBA success. This study showed that college performance predicted NBA careers and draft order. They also found that teams tend to stick with higher draft picks longer than lower

4 draft picks even when controlling for productivity. This finding is consistent with that of

Staw and Hoang (1995) as well as Camerer and Weber (1999). The study indicates that college level productivity measures can do a good job of explaining NBA career productivity. This is significant because some of the college statistics that they found to be best at explaining NBA career success are consistent with the statistics that correlate strongly with what the NBA values when making NBA compensation decisions. This helps us gain an understanding that NBA economic decision makers are rational, however we hope to gain further insight into the process of evaluating a players potential with hopes of revealing even clearer rational decisions.

What makes many of the studies different is the way they define NBA and college level success. We have seen from previous studies that some of the traditional relative measures of success tend to translate from college to the professional levels. However, most studies, aside from Coates and Oguntimein (2008), use many related variables to predict a measure of NBA performance. As is later noted, this potentially introduces co-linearity that may affect some important relations (Moxley and Towne,

2014).

Moxley and Towne (2014) attempt to define untapped potential using the information we now have available with the implementation of the NBA combine before each year’s drafts. Of the previous studies reviewed, none had athleticism measures or variables like arm span, which might be more useful than height alone. The metric that they use for NBA success is an advanced metric described at a win share. They legitimize using win shares as a reliable metric for NBA success by pointing out that the top 25 players of all time with the most win shares have all been selected to the hall of fame.

5 Moreover, it is an all-encompassing statistic that eliminates concerns of co-linearity. It is a metric that is calculated at the college level as well, so if deemed a reliable metric for player success, it can be easily analyzed and compared between the college and professional level. The predictor variables used in this study were: player position (guard, forward, center), age at the beginning of the next NBA season, the player's college win shares, college quality, NBA combine agility, combine no-step vertical leap, combine arm span, and combine weight. Position was coded 0 for guard, 1 for forwards, and 2 for centers. In the results of this study, the only variables that predicted NBA success were age, players' college win shares, and college quality. They also showed that NBA draft order appears to value anthropometric and athletic variables despite those variables not being necessarily related to NBA success.

In this study we use all public information about players that is available. Like some of the previous studies, the focus of this paper is to examine factors that are most likely to contribute to early career success among professional players in the National

Basketball Association (NBA) and to better understand the factors available on draft day that can aide in predicting players' future performance. Like Moxley and Towne, we use win shares to denote success both at the college and professional levels. Win shares is an advanced metric that calculates how much a single player contributes to the teams wins while taking into account both offense and defense. In short, it is an effort to credit a players total measurable contribution to his team’s win total during the season

(Basketball-reference.com). The goal of this study is to attempt to quantify the factors that affect success as a player transitions from the college to the NBA level. This is done with the use of combine statistics that are publicly available, and the advanced metric of

6 win shares which we use to denote “success”. While similar in this aspect, we will also attempt to create our own measures to quantify potential and work ethic. A key difference in our study is that we separate the players from the general data set by grouping one and two year players and three and four-year players in separate subsets of the data. We then analyze them using separate regressions. We also look at the players in pools from similar positions (guards, wings, big men) and run separate regressions on each of those subsets as well. Finally we look to bring them together to analyze the entire pool as a collective. Additionally, we use a more recent sample of players from 2005 to 2010.

Basketball is an ever-evolving game and in the last 10 years alone we have seen shifts in the way the game has been played. Aspects of the game seem to be more and less valued as time continues and new players with new, different skills, and specialized training enter the league. This study attempts to keep us up to date with the times, and keep us informed as to what is truly important in evaluating potential prospects.

Win shares

Win shares is a reliable metric of all-around performance that correlates well with other aggregate metrics of performance (Winston, 2009). The win shares metric uses a mixture of statistical efficiency measures, including how often players are involved in plays and a team's overall defensive efficiency with the player on the court (Winston,

2009). The statistic is not perfect because the defensive component is influenced by the quality of a player's teammates. However, the players ranked in the top 25 in NBA history in win shares have all been selected to the Basketball Hall of Fame and players leading the league in win shares tend to be those selected to the all star game and all-

NBA teams. As Moxley and Towne (2014) point out, another advantage of using win

7 shares is that the statistic is available for college players so it will be easy to compare the measures for success across both timeframes.

According to Baskeball-reference.com, the creators of the advanced metric, win shares as a statistic is meant to be conceptually simple. One Win Share should reflect the offensive and defensive contributions of a player that have led to one win during the season. The offensive side of win share calculation includes field goals, assists, and free throws. Credit is also given for offensive rebounds that enable point production. On defense a player receives credit for possessions where his team prevents a score and where the player himself contributes a stop. Generally, a stop is awarded through easily understood events like steals, blocks, and defensive rebounds. Taking all of this into account, a player that makes well-rounded contribution on both sides of the ball will do well in the system (Basketball-Reference.com).

Formula for Crediting Win Shares

The formulas for crediting offensive Win Shares are complicated and are outlined by

Basketballreferenece.com based on Dean Oliver's points produced and offensive possessions by the creator of the advanced metric. Examples are provided using Lebron

James statistics from the 2009-2009 seasons. Baskeballreference.com acknowledges that one should read Oliver’s book, basketball on paper to learn more about Oliver’s metrics.

BaskeballReference.com outlines the formula for calculating win shares as follows.

8 Crediting Offensive Win Shares to Players

A. 1977-78 to present NBA

1. Calculate points produced for each player. In 2008-09, James had an estimated 2345.9

points produced.

2. Calculate offensive possessions for each player. James had an estimated 1928.1

offensive possessions in 2008-09.

3. Calculate marginal offense for each player. Marginal offense is equal to (points

produced) - 0.92 * (league points per possession) * (offensive possessions). For James

this is 2345.9 - 0.92 * 1.083 * 1928.1 = 424.8. Note that this formula may produce a

negative result for some players.

4. Calculate marginal points per win. Marginal points per win reduces to 0.32 * (league

) * ((team pace) / (league pace)). For the 2008-09 Cavaliers this is 0.32 *

100.0 * (88.7 / 91.7) = 30.95.

5. Credit Offensive Win Shares to the players. Offensive Win Shares are credited using

the following formula: (marginal offense) / (marginal points per win). James gets credit

for 424.8 / 30.95 = 13.73 Offensive Win Shares.

Crediting Defensive Win Shares to Players

A. 1973-74 to present NBA

1. Calculate the Defensive Rating for each player. James's Defensive Rating in 2008-09

was 99.1.

2. Calculate marginal defense for each player. Marginal defense is equal to (player

minutes played / team minutes played) * (team defensive possessions) * (1.08 * (league

9 points per possession) - ((Defensive Rating) / 100)). For James this is (3054 / 19780) *

7341 * ((1.08 * 1.083) - (99.1 / 100)) = 202.5. Note that this formula may produce a

negative result for some players.

3. Calculate marginal points per win. Marginal points per win reduces to 0.32 * (league

points per game) * ((team pace) / (league pace)). For the 2008-09 Cavaliers this is 0.32 *

100.0 * (88.7 / 91.7) = 30.95.

4. Credit Defensive Win Shares to the players. Defensive Win Shares are credited using

the following formula: (marginal defense) / (marginal points per win). James gets credit

for 202.5 / 30.95 = 6.54 Defensive Win Shares (Basketball Reference.com).

Data/ Methodology

All data implemented in this study was collected from websites with publicly

available information. All combine participation info and results for a given year was

obtained from stats.nba.com. College statistics, NBA statistics, and draft information was

obtained from sportsrefrence.com and basketballrefrence.com respectively. Of the

previous studies, only Moxley and Towne included anthropometric variables like

standing vertical, and athleticism variables like arm span, which may be used as a proxy

for height, and could be more relevant to the game of basketball. The basic regression

models take the following equation form, differing in the number of Betas depending on

the number of independent variables in the regression:

!"#$%!&'()*+,-./-,0-+, = !(!! + !! + ⋯ !!) + !!

In this study we run six separate OLS regressions with different sections of the

data. One regression with all the player data, then we separate the others by one and two

year players, three and four year players, guards, forwards, and big men. It is ok that parts

10 of each data set do overlap, as we are simply looking for correlations. Different positions require different skill sets, so it may be more helpful to look at and compare players who are more similar to each other.

The independent variables used in this study are player position (, shooting guard, small forward, power forward, center), the player's college win shares per year, win shares in their first year, average college win shares per year, quality of conference, NBA combine agility, combine no-step vertical leap, and dummy variables for when they came out of college. The dependent variable in all of the regressions is the average wins per year through five years of each players NBA career. For this metric a player with 25 win shares through five years of their NBA career would receive a value of 5 (25win shares/5years =5 win shares per year). In other words this is the variable we chose to indicate success at the NBA level. As we separate the data out for the different regressions, different variables were added and taken away, as is displayed in equations

1.1-1.6.

Player position separated into five dummy variables, PG, SG, SF, PF, and C.

Players listed as multiple positions were given a value of 1 for both of those positions listed and 0 for the rest.

In equations 1.4-1.6 we separate the data by player position: Guards, Wings, and

Big Men. Players listed as a point guard (PG) or shooting guard (SG) were included in the guard regression (equation 1.4). Players listed as a small forward (SF) or power forward (PF) were included in the wings regression (equation 1.5). Players listed as power forwards (PF) and centers (C) were included in the big man regression. If a player was listed as a SG/SF they were included in both the guard and big man regressions.

11 Players who attended the NBA combine and did not play every year in the NBA were given a zero value for the years they didn’t play when calculating Average win shares per year in their first five years of the NBA. We attempt to hold constant the age effect as denoted by Musch and Grondin (2001) by including how many years of college the players in the sample played before they were drafted. Moreover, we only include players who played at least one of a professional NBA game, to see how players faired who were at least given an opportunity to compete, practice, and prove themselves at the highest level. All players who never played in the NBA were discarded from the sample.

College quality is a dummy variable given a value of zero or one depending on the conference that the player’s school was in. We decided to go with a power six- conference classification. Players who played at a school in one of the classic power five conferences (ACC, Big Ten, Big twelve, Pac twelve, and SEC) were given a value of 1.

The Big East was included as the sixth power conference because the classification of the power five conferences was originally dictated by college football. While the teams in each of these conferences are still considered powerhouses in college basketball, the caliber of the teams in the Big East conference seemed on par with the original five conferences, thus players from teams in the Big East were given a value of 1 as well.

Teams in these conferences may have been identified as better players already coming out of high school, and may have gained access to better coaching and development at these well-established basketball schools.

Combine measurements such as wingspan and weight were included in the regressions where the data had been separated by position. These variables would have

12 little significance in the larger, more general data sets used in equations 1.1-1.3 that included all different positions because big men will almost always have greater wingspans and weigh more.

The variable Average Annual Increase in Win Shares Per Year was added to the three and four year player regression as a proxy for practice and improvement over their time in college. This variable was calculated by finding the difference between the win share stat from their first and last year in college and dividing it by the number of years they played. We figured that players who improved more over their time in college would have a similar trajectory in the NBA, and thus be more successful than they appear to have been. (Equation 1.3).

The specified explained variables and equations used for each regression are as follows:

Variable(Name(( Variable(Description( *AvgNBAWinSharesP erYear( Average(Win(Shares(per(year(through(5(career(NBA(seasons( Player(was(listed(as(a(Point(Guard(or(a(combination(of(Point( PG( Guard(and(another(position( Player(was(listed(as(a(Point(Guard(or(a(combination(of(Point( SG( Guard(and(another(position( Player(was(listed(as(a(Small(Forward(or(a(combination(of(Small( SF( Forward(and(another(position( Player(was(listed(as(a(Power(Forward(or(a(combination(of( PF( Power(Forward(and(another(position( Player(was(listed(as(a(Center(or(a(combination(of(Center(and( C( another(position( Wingspan(as(measured(at(the(NBA(combine(prior(to(the(NBA( CombineWingspan( draft( CombineWeight( Weigt(as(measures(at(the(NBA(combine(prior(to(the(NBA(draft( (lane(agility(time(measured(at(NBA(combine(prior(to(NBA( LaneAgility( draft( Official(Standing(Vertical(Leap(measured(at(the(NBA(Combine( StandingVertical( prior(to(the(NBA(Draft( Dummy(Variable:(1,(if(played(at(any(of(the(power(6(College( Power6Confrence( Basketball(Conferences((

13 Dummy(Variable:(1,(if(played(only(one(year(of(College( One&Done( Basketball(prior(to(the(NBA(Draft( Dummy(Variable:(1,(if(played(three(or(four(years(of(College( 3or4YearPlayer( Basketball( FirstYearCollegeWinS Win(Shares(contributed(in(first(year(of(playing(college( hares( basketball( AverageCollegeWinS haresPerYr( Average(win(shares(contributed(per(year(through(college(career( AverageAnnualIncre aseinWS( Average(increase(in(win(shares(annually(over(college(career( "*"(Denotes(the(Dependent(variable(

Equation 1.1 – All Players

!!!!"#$%!&'()ℎ!"#$%#"&#!"

= !! + !!!" + !!!" + !!!" + !!!" + !!! + !!!"#$%&'(')*

+ !!!"#$%&$'()*"&+#, + !!!"#$%6!"#$%&#'& + !!!"#&!"#$

+ !!"3!"4!"#$!"#$%& + !!!!"#$%&'(#)*++','-"./ℎ!"#$

+ !!"!"#$%&#'())#&#*+,-ℎ!"#$%#"&#!" + !!

Equation 1.2 – One and Done Players

!!!!"#$%!&'()ℎ!"#$%#"&#!"

= !! + !!!" + !!!" + !!!" + !!!" + !!! + !!!"#$%&'(')*

+ !!!"#$%&$'()*"&+#, + !!!"#$%6!"#$%&#'&

+ !!!"#$%&#'())#&#*+,-ℎ!"#$%#"&#!" + !!

14 Equation 1.3 – Three and Four-Year Players

!!!!"#$%!&'()ℎ!"#$%#"&#!"

= !! + !!!" + !!!" + !!!" + !!!" + !!! + !!!"#$%&'(')*

+ !!!"#$%&$'()*"&+#, + !!!"#$%6!"#$%&#'&

+ !!!"#$%&#'())#&#*!"#ℎ!"#$%#"&#!"

+ !!"!"#$%&#!'(%)*'+$#%,#-.'/ℎ!"#$%#"&#!" + !!

Equation 1.4 – Guards

!!!!"#$%!&'()*+,-./-,0-+,

= !! + !!!"#$%&'(%&)*+,& + !!!"#$%&'(!"#ℎ! + !!!"#$%&'(')*

+ !!!"#$%&$'()*"&+#, + !!!"#$%6!"#$%&#'& + !!!"#&!"#$

+ !!3!"4!"#$%&#'"$ + !!!"#$%&#'())#&#*+,-ℎ!"#$%#"&#!" + !!

Equation 1.5 - Wings

!!!!"#$%!&'()*+,-./-,0-+,

= !! + !!!"#$%&'(%&)*+,& + !!!"#$%&'!!"#$ℎ! + !!!"#$%&'(')*

+ !!!"#$%&$'()*"&+#, + !!!"#$%6!"#$%&#'& + !!!"#&!"!"

+ !!3!"4!"#$%&#'"$ + !!!"#$%&#'())#&#*+,-ℎ!"#$%#"&#!" + !!

Equation 1.6 – Big Men

!!!!"#$%!&'()*+,-./-,0-+,

= !! + !!!"#$%&'(%&)*+,& + !!!"#$%&'!!"!"ℎ! + !!!"#$%&'(')*

+ !!!"#$%&$'()*"&+#, + !!!"#$%6!"#$%&#'& + !!!"#&!"#$

+ !!3!"4!"#$%&#'"$ + !!!"#$%&#'())#&#*+,-ℎ!"#$%#"&#!" + !!

15 Results and Analysis

Table 1.1 ( Table 1.1

1 & 2 year 3 & 4 Year All Players Players Players Guards Wings Big Men Avg NBA Avg NBA Avg NBA Avg NBA Avg NBA Avg NBA Win Win Win Win Win Win Shares Per Shares Per Shares Per Shares Per Shares Per Shares Per Year Year Year Year Year Year Variables

PG 0.279 0.385 0.526 (0.71) (0.39) (1.26)

SG -0.405 -0.509 -0.377 (-1.23) (-0.61) (-1.10)

SF -0.434 -0.388 -0.295 (-1.25) (-0.41) (-0.81)

PF -0.304 -0.356 -0.0811 (-0.79) (-0.44) (-0.19)

C -0.355 0.140 -0.915* (-0.70) (0.12) (-2.09)

Wingspan -1.434 -0.326 -0.319 (-1.60) (-0.32) (-0.24)

Weight (lbs) 0.00213 0.0108 0.0117 (0.18) (1.08) (0.74)

Lane Agility Time 0.414 1.020 0.160 0.400 0.627 0.412 (1.35) (1.46) (0.52) (0.85) (1.63) (0.71)

Standing Vertical Leap 0.150** 0.111 0.157*** 0.152* 0.200** 0.202

16 (3.14) (1.02) (3.39) (2.39) (2.99) (1.68)

Power 6 Confrence 0.708* 1.482* 0.298 1.061** 0.452 0.403 (2.25) (2.31) (1.00) (2.64) (1.08) (0.69)

One & Done -0.797 -1.088 -0.281 -0.533 (-1.26) (-1.05) (-0.34) (-0.47)

3 or 4 Year Player -1.425** -1.420 -1.302* -1.740 (-2.66) (-1.46) (-2.03) (-1.96)

First Year College Win Shares -0.0962 -0.136 -0.102 -0.210 (-0.57) (-0.53) (-0.46) (-0.71)

Average College Win Shares per Year 0.761*** 0.528** 0.795*** 1.026*** 0.669* 0.795* (3.63) (2.72) (7.38) (3.69) (2.37) (2.41)

Average Annual Increase In win shares 0.434* (2.04)

_cons -9.239* -15.36 -8.721* -1.463 -13.30* -10.94 (-2.32) (-1.75) (-2.10) (-0.22) (-2.15) (-0.85)

N 218 70 148 109 123 70

t statistics in parentheses, * p<0.05, **p<.01, ***p<.001 (

17 General Findings

Our results from the six regressions yield some interesting findings. Specific values can be found on table 1.1. We find certain athletic variables such as the standing vertical leap to be significant which differs from previous studies. Much of what we found was in line with our theory, and college win shares displayed itself as a reliable metric to use in the process of evaluating potential NBA prospects at different positions and through different points in their college careers. After the initially running each of models we ran the White Test to identify any heteroskedasticity. To correct for this we ran robust regressions with the hopes of normalizing the residuals in the results. We also ran correlation matrices to identify any multicollinearity between the independent variables. We did not have any problems with multicollinearity and believe this was due to the fact that we used one all encompassing statistic to denote success rather than all of the individual basic statistics.

All Player Data

For combine participants from 2005-2010, none of the positions appear to be more successful or significant at the NBA level than others. Standing vertical leap showed to be very statistically significant at the .01 level. This alludes to the fact the explosiveness and ability to play above the rim goes a long way to predicting success at the NBA level. Along with this, the distinction of whether the player played in one of the power six conferences displayed statistical significance at the .05 level. This is in line with our theory, as better players will likely be recruited to better basketball schools.

While there, they are likely to have better access to developmental tools, and will be used to be playing against better competition.

18 Being a three or four-year player correlated with being less successful at the NBA level with 99% statistical significance and a negative coefficient. This means that many players who stay in college may be staying there for a reason, either because they know they are not good enough to be drafted, or they need more time to prove themselves and develop. Some may stay because they simply want to complete their degree. It is interesting to note that while being a three or four-year player had a negative correlation to NBA success, being a one and done player was not significant as a positive indicator of success at the NBA level. From this we can derive that while a player may be hyped up as a legitimate NBA prospect, it is still important to look at his contributions to his team’s success and continue to evaluate further before making a final judgment.

Finally, average college win shares per year had incredibly high correlation with average NBA win shares per year at the .001 level (See Table 1.1). This shows that above all else; college production will correlate very highly with NBA production. In the context of this study, and in alignment with the findings of Moxley and Towne, it shows that Win Shares per year at the college level is correlated very correlated with Win

Shares through the first five years of a given NBA career.

We learned from evaluating the correlation measures, that while other variables have significance in showing performance at the NBA level they are not directly correlated to success at the college level. Therefore when evaluating a player it is of importance to look for all significant indicators of success in conjunction with each other.

In the subsequent regressions we look to segment the data to identify variables that may be more relevant given certain criteria.

19 1 & 2-Year Players

In evaluating one and two year college players on their own, and evaluating their success at the NBA level, the variables of statistical significance were the caliber conference they played in at the (p< .05 ) and their college production (p<. 01). None of the anthropometric and athletic variables proved to be of any statistical significance in this segment of the data.

One and done players are likely identified as being more gifted, and more athletic early on. Many of the one and done players know they will be drafted even before their first season begins, and in comparison with each other, athleticism may be normalized.

This regression shows that among one and two-year college players who declare for the draft, and eventually make it to the NBA, the most athletic ones are not necessarily the best fits for the NBA. By this stage in the process, a ton of the vetting has already been done. Scouts have been watching many of these players since before high school as they played for their AAU squads. Many of these players were recruited to elite basketball schools for these reasons. Accordingly, as is displayed in the regression results (table 1.1) an individual’s contribution to his team in his first and second seasons in a major conference is the main criteria that should be taken into account for scouting players that declare for the NBA draft after only a season or two playing college basketball.

3 & 4-year players

If we recall the initial regression which included all players who played in the

NBA and participated in the combine from 2005 to 2010, it showed that the three and four year players tended to be less productive in general at the NBA level. In fact there was a negative correlation between these players who were given a chance at the NBA

20 level and their NBA success. This regression is here to evaluate what attributes of these three and four-year players may defy the norm and lead to success at the NBA level.

Centers in this segment of the data are less likely to contribute at the NBA level

(p< .05) than any other position in the sample. As we evaluate the Big Man regression

(equation 1.6) later in our analysis, we may gain some insight as to why that is. Standing vertical leap was significant at the p< .001 level. While players who are not identified as top prospects right out of the gate may need more time to develop, the fact remains that most of these players will need the physical tools to compete at the NBA level.

In this regression, being from a particularly good conference had less of an effect on NBA success, while producing for your team still did (p< .001). Additionally, we looked at how the individuals improved over their college career. The average increase in win shares per year had significance at the p< .05 level. From this we can derive that you do want to see an increased contribution over a player’s college career, in conjunction with their overall contributions and athleticism variables.

Guards, Wings, and Big Men

In the positional analysis, standing vertical leap displayed significance for guards and wings at the p< .05 and p<.01 levels respectively, while no significance was shown for big men. This alludes to the fact that big men don’t need to jump very high to play above the rim. Also, it may signify certain physical thresholds that players with different body types cannot exceed. An interesting future study could use this information to identify these thresholds, and examine the production of players who are able to exceed these theoretical boundaries.

21 The quality of conference from which a player played, only showed significance for a guard’s success at the NBA level as p< .01, but showed no significance for wings or big men. One and done players were no better for any of the positions, while three or four year wings had a significant negative correlation with NBA success (p< .05). College production was very significant in relation to guards’ production in the NBA (p< .001) and was also significant for wings and big men (p< .05).

Conclusion

This paper looks at factors that are most likely to contribute to early career success among professional players in the NBA. It aims to better understand the details available that significantly predict players' future performance relative other players of similar standing. We have now seen how the different variables in each regression correlate to

Win Shares through the first five years of a given player’s career in the NBA. None have carried more significance than the advanced metric of win shares at the college level. It is important to take this information as theory. Evaluating players based on the displayed significant criteria alone may not be the best approach.

This study exhibits much of what we had already known about evaluating potential NBA prospects. Berri, Brook and Fenn (2011) showed that measures such as points, rebounds steals, shooting efficiency, and team success predicted NBA performance. Coates and Oguntimein (2010) displayed that college performance predicted NBA careers and draft order. We have seen college performance measures and age predict draft status and performance. Moxley and Towne (2014) used the win share metric at the college level to predict NBA success as denoted by win shares through 3 years with players from 2001-2006. They found that win shares were correlated with, and

22 can be viewed as a reliable metric for predicting win shares at the NBA level.

We expand upon this study by using a more recent sample of players (2005-

2010) and evaluating them through five years of their NBA careers, a time when many players begin to their “prime”. Furthermore, we look at and compare players from their respective positions, and when they chose to leave college.

However you choose to evaluate it, college production remains a key factor in determining whether a prospect will be successful at the NBA level. Instead of points, rebounds, and assists, we look at win shares as an all-encompassing metric that reduces co-linearity that tends to be present if you use multiple of the basic statistics. We also found the standing vertical leap to be significant for many of the subsets of data, a finding that differs directly from proceeding studies. Finally we also illustrate that different types of players require different criteria and metrics for evaluation.

Every player is different, and there are different psychological, and unquantifiable measures of commitment, practice, and leadership that go into an NBA prospects’ development. This is knowledge that that one would only be able to gain from speaking with the player or their previous coaches directly. This study is not meant to dismiss the value of a good scout, but is instead meant to be used as a tool for identifying the “it” factors of potential prospects. To put it simply, if a player produces in the form of win shares at the college level, they are from a good conference, and they are explosive, then according to our study they will be likely to succeed at the NBA level. A specific lacking in any of the significant variables as described in this study for a particular player may be cause to raise a flag for NBA front offices, and may warrant a more thorough and deeper evaluation of the potential prospect being evaluated.

23

References

Berri, D. J., Brook, S. L., & Feen, A. J. (2011). From college to the pros: predicting NBA amateur player draft. Journal of Productivity Analysis, 35, 25e35.

Camerer, C., & Weber, R. (1999). The econometrics and behavioral economics of escalation of commitment: a re-examination of Staw and Hoang’s NBA data. Journal of Economic Behavior and Organization. 39(1), 59e82.

Coates, D., & Oguntimein, B. (2010). The length and success of NBA careers: does college production predict professional outcomes? International Journal of Sports Finance, 5, 4e26.

Groothuis, P. A., Hill, J. R., & Perri, T. J. (2007). Early entry in the NBA draft the influence of unraveling, human capital, and option value. Journal of Sports Economics, 8 (3), 223 - 243.

Musch, J., & Grondin, S. (2001). Unequal competition as an impediment to personal development: a review of the relative age effect in sport. Developmental Review, 21(2), 147e167.

Rodenberg, R. M., & Kim, J. W. (2011). Precocity and labor market outcomes. Evi- dence from professional basketball. Economics Bulletin, 31, 2185e2190.

Staw, B. M., & Hoang, H. (1995). Sunk coasts in the NBA: why draft order affects playing time and survival in professional basketball. Administrative Science Quarterly, 40, 474e494.

Winston, W. L. (2009). Mathletics. Princeton, NJ: Princeton University Press.

Web References

NBA & ABA Player Directory. (2013, September). Sports Reference LLC. Retrieved from http://www.basketball-reference.com/players/.

NBA Advanced Stats. (2018, February). NBA Media Ventures LLC. Retrieved from https://stats.nba.com/draft/combine-anthro/.

Player Index. (2018, February). Sports Reference LLC. Retrieved from http://www. sports-reference.com/cbb/players/.

24 Total NBA Revenue 2001-2017. Statista. (2018, February). Retrieved from https://www.statista.com/statistics/193467/total-league-revenue-of-the-nba-since-2005/.

Books

Lewis, Michael. Moneyball: The Art of Winning an Unfair Game. New York: W.W. Norton, 2013

25