ArticlePDF Available

Predicting the WNBA Draft: What Matters Most from College Performance?


Abstract and Figures

The reverse-order draft has been the subject of a number of studies in the economics literature. These studies generally examine the quality of decisions teams make in this process. The results in the studies of the NFL, NBA, and MLB all highlight problems with the player evaluation process. This study contributes to this literature and the broader literature on gender economics via an examination of the WNBA. Similar to other studies, we also find issues with decision-making in the WNBA draft.
Content may be subject to copyright.
International Journal of Sport Finance, 2015, 10, 299-309, © 2015 West Virginia University
The Women’s National Basketball Association (WNBA) utilizes a version of the
reverse-order draft lottery similar to the NBA. While previous research has examined
factors impacting the draft picks and subsequent professional performance in the
NBA,1we do not believe this subject has been explored for the WNBA. In fact, little
has been published on this league. This particular study will build upon the work of
Berri and Krautmann (2013), which presented a model of player performance for the
WNBA. This work will be paired with the approach to the NBA draft introduced by
Berri, Brook, and Fenn (2011). In the end, this research will explore both what factors
determine draft position in the WNBA and how such factors relate to subsequent per-
Conceived in 1996, the WNBA has experienced the typical growing pains of many
relatively young sports leagues.2As we saw in the early history of the NBA, teams in
the WNBA have come and gone while profits and attendance have expanded and con-
tracted. With respect to the number of teams, the league began with eight teams in
Predicting the WNBA Draft:
What Matters Most from
College Performance?
Jill Harris1and David J. Berri2
1Pomona College
2Southern Utah University
Jill S. Harris, PhD, is a visiting assistant professor of economics and teaches
Economics of Sport and Economics of Crime. Her research interests include the
nature and behavior of the National Collegiate Athletic Association (NCAA),
women in sport, acquatic sports, and non-compliance behavior.
David J. Berri, PhD, is a professor of economics in the Department of Economics
and Finance. His current research focuses on the economics of sports, specifically
the topics of consumer demand, competitive balance, worker productivity, and
women in sport.
The reverse-order draft has been the subject of a number of studies in the economics
literature. These studies generally examine the quality of decisions teams make in this
process. The results in the studies of the NFL, NBA, and MLB all highlight problems
with the player evaluation process. This study contributes to this literature and the
broader literature on gender economics via an examination of the WNBA. Similar to
other studies, we also find issues with decision-making in the WNBA draft.
Keywords: WNBA, draft, gender, basketball
Harris, Berri
300 Volume 10 • Number 4 • 2015 • IJSF
1997, expanded to 16 teams by 2000, before contracting to its current 12 teams by
2010. Currently the league remains with just the following teams: Atlanta Dream,
Chicago Sky, Connecticut Sun, Indiana Fever, Los Angeles Sparks, Minnesota Lynx,
New York Liberty, Phoenix Mercury, San Antonio Silver Stars, Seattle Storm, Tulsa
Shock, and Washington Mystics.
Relative to the NBA, these teams appear—when we look at with-in season varia-
tion—to be somewhat competitive.3When we look at league championships, though,
we see less competition. Of the 18 franchises that have existed in the WNBA, only eight
have ever won a title. And 15 of the league’s 17 titles have been won by just six organ-
One proposed solution to such imbalance embraced by sports leagues throughout
North America is the reverse-order draft. This institution rewards the worst teams in
a league with a choice of the best available talent not currently employed in the league.4
For this to improve competitive balance, the teams selecting first must be selecting
more productive players than those taken later. There is a problem, though, with the
selection process. At the time of the draft a team has not seen how a given player will
perform against the talent seen in the professional league. Consequently, it is possible
the worst teams do not actually get better via the draft.
This is essentially the story told by much of the previous research on drafts in pro-
fessional sports. Both Massey and Thaler (2006) and Berri and Simmons (2011)
uncovered problems with how talent is selected in the NFL draft. With respect to the
NBA, Kahn and Sherer (1988) reported no statistical relationship between draft posi-
tion and a player’s statistical performance in college. More recently, Berri et al. (2011)
report that draft position is related to scoring totals,5but other factors—like shooting
efficiency and rebounds—that are more closely aligned with winning were not found
to matter much in a player’s draft position. These results are similar to those of Coates
and Oguntimein (2010). Using data from 1987 to 1989, the authors found points
scored were important for draft position, but not indicative of professional point scor-
ing. Rebounds, blocks, and assists were correlated more with pro performance.
The importance of scoring in player evaluations in the NBA is not a new finding.
Berri and Schmidt (2010) offer evidence of the primacy of scoring in terms of salary
allocation.6But a study of salaries in the WNBA is problematic. The collective bargain-
ing agreement in the WNBA results in salaries that are much more regimented and far
lower than the NBA.7Consequently this study into player evaluation in the WNBA
turns to the draft, an arena that allows us to examine which factors decision-makers
consider in evaluating basketball talent.
Research on the WNBA in this area is interesting for at least four reasons. More
broadly, an examination of decision-making in sports—where labor productivity data
is abundant—allows one to assess whether or not decision-makers act in a fashion
consistent with the neoclassical model. Related to this point, decision-makers in the
NBA have been found to focus on scoring (seemingly) above all else when evaluating
player talent. Prior research establishes that efficiency in scoring is not significant in
player evaluation. Will the WNBA management suffer the same efficiency blindness?
A third reason for interest is the analysis of player productivity from the college draft
into the pro league. Specifically, can we determine whether the factors highlighted on
draft day predict WNBA performance? Berri et al. (2011) reviewed how Hollinger
Predicting the WNBA Draft: What Matters Most from College Performance?
Volume 10 • Number 4 • 2015 • IJSF 301
(2003) and Oliver (2004) assert that wins are determined by a team’s ability, relative to
its opponent, to elicit points from its possessions. Gaining possession is a function of
rebounding, steals, and turnovers. We are curious whether these factors have more
impact in the WNBA than they did in the Berri et al. (2011) research.
Finally, this research may contribute to the broad literature on gender economics
and how the work of females and males is evaluated. The WNBA and NBA are closely
linked; the NBA initially owned the WNBA and many decision-makers in the former
have extensive experience with the latter. Papers published in organizational behavior,
higher education, and other behavioral journals point to differences in performance
evaluation based on gender. For example, in a widely cited paper, Basow (1995) found
significant interactions between teacher gender and student gender in student course
evaluations. Male professors’ evaluations were not impacted by student gender while
female professors’ evaluations were impacted by student gender. Female instructors
received their lowest ratings from male students and highest ratings from female stu-
dents. This type of effect is reported in a variety of studies across multiple fields.8Our
findings suggest female player performance does not appear to be evaluated different-
ly than male player performance. If this result holds then it would certainly be of inter-
est to audiences outside of sports economics.
Modeling the Draft
This study of the draft uses amateurs chosen by WNBA teams from NCAA programs.
International players are excluded since comparable statistics for their pre-profession-
al experience are not available.9We collected draft data from 2010 to 2013 and college
performance data from the drafted players’ last year of college play. Summary statistics
for our data are detailed in Table 1. The sample is small—only 128 players. However,
the model generates results similar to those found in Berri et al. (2011), suggesting
managers in the WNBA suffer similar impairments to the NBA when it comes to
assessing player performance and productivity.
The average player in our sample is 6-foot tall, the tallest is 6-foot-6, and the small-
est is 5-foot-4. Unlike the male athletes drafted into the NBA, all the female players in
the sample completed four years of college, making the age range tighter. The average
age is a little over 22 while the oldest players were 24 when drafted. Another interest-
ing feature of the data is that more draft picks come from the SEC than any other con-
ference (the ACC and PAC 10 have the honors in the NBA study conducted by Berri et
al.). Almost twice the number of draft picks in the WNBA played in the NCAA Final
Four versus the NBA. Since the dependent variable is draft position, a variable that
positively impacts draft position will have an estimated coefficient with a negative
As with Berri et al. (2011), the dependent variable is draft pick selection number.
This value ranges from 1 to 36 in each draft year (12 teams with a three-round draft).
Each player’s performance is adjusted for position played as in other prior work by
Berri and Berri et al.11
What factors impact draft position? We begin with player performance in college. As
noted by Berri et al. (and in contrast to the work of Kahn and Scherer [1988]) players
with better college performance statistics should be drafted higher. After all—as noted
earlier—the primary stated purpose of the draft is to give poor performing teams
Harris, Berri
302 Volume 10 • Number 4 • 2015 • IJSF
access to better players. Next, height and age are included in the model. Although it is
cliché: you cannot teach height. Particularly in the female game we might expect the
short supply of tall women to be even more critical to the draft decision. Because the
advantage of height relative to position played could also impact draft position we
model height relative to position.12
Given the attention paid to age and experience in the NBA it makes sense to include
age as a control characteristic in our sample. However, the women’s game is not as
lucrative at the professional level. Female athletes have fewer incentives to leave college
Table 1. Descriptive Statistics for Dependent and Independent Variables (2010-2013)*
Variables Label Mean SD Min. Max.
Draft position PICK 18.37 10.33 1 36
Points scored PTS 18.80 4.70 8.3 31.03
Rebounds REB 6.18 3.43 0.35 13.17
Assists AST 4.35 1.92 1.36 13.60
Steals STL 2.80 0.93 1.18 5.42
Blocked Shots BLK 0.86 1.14 -1.57 3.29
Personal Fouls PF 2.45 0.76 0.81 4.94
Turnover % TOPER 0000
3 pt field goal % 3FGPER 0.29 0.14 0 0.75
2 pt field goal % 2FGPER 0.47 0.07 0.31 0.71
Free Throw % FT 0.74 0.09 0.47 0.89
Age AGE 22.39 0.55 22 24
Height, in RELHT 72.10 2.20 64.80 78.20
Final Four player DFIN4 0.16 0.37 0 1.0
Played in ACC DACC 0.19 0.39 0 1.0
Played in PAC 12 DPAC12 0.10 0.31 0 1.0
Played in Big East DBIGEAST 0.05 0.22 0 1.0
Played in SEC DSEC 0.24 0.43 0 1.0
Played in Big Ten DBIG10 0.09 0.29 0 1.0
Played in Big 12 DBIG12 0.09 0.29 0 1.0
Played in Conf USA DCONFU 0.03 0.17 0 1.0
Played in Mt West DWEST 0.09 0.29 0 1.0
Played in Colonial DCOLON 0.01 0.10 0 1.0
Played in American DAMER 0.08 0.27 0 1.0
Played in Sun DSUN 0.01 0.10 0 1.0
Played Center DC 0.14 0.34 0 1.0
Played Forward DF 0.40 0.49 0 1.0
Played Guard DG 0.45 0.50 0 1.0
* There are 128 observations. Notes: PTS, REB, AST, STL, BLK, and PF are per 40 min
and adjusted for position played. TOPER is also adjusted for position played.
Turnover Percentage = [(Turnovers)/(Turnovers + Field Goal Attempts + o.44*Free
Throw Attempts)] Sources: college performance data from; height data is
Predicting the WNBA Draft: What Matters Most from College Performance?
Volume 10 • Number 4 • 2015 • IJSF 303
early to enter the WNBA. Therefore, we do not expect age to be as important in the
draft selection story here.
In addition to the regressors mentioned above we include dummy variables for ath-
letic conference to capture the difference in quality of college team played for and the
degree of competition faced. We also include a dummy for Final Four experience. As
with the NBA, experience in post-season play may be perceived as a plus by decision-
makers and should improve draft position. Finally, dummy variables for the draft class
years are included in the model for potential variations in the draft pool year to year
over the sample.
To summarize, we expect draft pick to be influenced by college performance (cap-
tured by PROD—a vector of player specific position adjusted performance statistics
including points, rebounds, steals, blocked shots, assists, turnovers, and measures of
shooting efficiency), RELHT or relative height, AGE, college conference given by dum-
mies for each, experience in post-season play (DFIN4), and relative quality of pool of
draftable athletes in a given year (captured by a time dummy for each year, D10, etc.).
These influences are all noted in Equation 1, which we will estimate in an effort to
explain where a player will be chosen in the WNBA draft.
PICK n= β0 + αNPROD + β1RELHT + β2DFIN4 + β3DCHAMP + β4AGE
+β5DACC + β6DPAC12 + β7 DBIGEAST + β8DSEC + β9DBIG10 + β10DBIG12 +
β11DAMER + β12 DCOLON + β13DWEST + β14DCONFU + β15DSUN + β16 DC +
β17 DF + β18DG + αJDYEAR + e it (1)
Empirical Findings
Equation 1 is estimated with four years of draft data. Estimation of the model was con-
ducted with both a Poisson Distribution model and a Negative Binomial model.13 The
estimations are reported in Table 2. Since the Poisson and Negative Binomial models
return coefficients from a Maximum Likelihood estimation process the coefficients are
not equivalent to estimated slopes. The coefficients are used to estimate marginal
effects at the sample means. These marginal effects are reported in Table 2.
Before discussing the results on the performance measures, the non-performance
factors invite comment. As was the case in the NBA, shorter female players are at a dis-
advantage. Other things the same, the taller you are—relative to the average at your
position—the better you are going to do in the draft. In contrast to the NBA, age is not
important in the WNBA draft sample. This is not surprising given the tight distribu-
tion of ages compared to the men when entering the draft. Appearing in the Final
Four, however, is important for drafting. A player with this experience will see her draft
position improve by almost six slots. In addition, coming out of the ACC, SEC, or the
Big East gives a bigger boost than from the PAC 12 or Big Ten.
Turning attention to the performance factors, which of those predicted skills that
were statistically significant had the most economic significance? As Table 2 indicates,
points scored, assists, and shooting efficiency from the two-point range are all positive
influencers of draft position while personal fouls work against the athlete. Steals,
blocks, and rebounds are not telling much of the story of draft selection in our sam-
ple. Given these predicted results, how meaningful are they?
Table 3 reports how an estimated one standard deviation increase in each statistical-
ly significant performance variable impacts draft position. Again, we suspect—given
Harris, Berri
304 Volume 10 • Number 4 • 2015 • IJSF
the aforementioned research into the NBA—that scoring might matter the most. And
the results in Table 3 indicate that scoring matters most! A one standard deviation
increase in points improves draft position by over eight slots; a similar increase in
assists improves drafting by almost five slots and improved field goal percentages
improve position by three slots.14 Relative to the NBA, assists are about twice as valu-
able in improving draft position in the WNBA while points scored and efficiency are
about a third more valuable to the potential professional player.
Table 2. Estimation of Equation (1) (2010-2013)
Variable Poisson Z-stat Negative Bin Z-stat
PTS -0.95*** -9.58 -1.08*** -5.51
REB -0.05 -0.28 -0.24 -0.63
AST -2.18*** -7.07 -2.58*** -4.32
STL -0.41 -0.74 -0.21 -0.19
BLK -0.52 -1.08 -0.68 -0.71
PF 1.41** 2.67 2.11** 2.04
TOPER 0000
3FGPER 4.87** 2.05 3.47 0.78
2FGPER -50.38*** -5.90 -42.71** -2.69
FT -7.09 -1.29 -1.85 -0.17
RELHT -0.90*** -5.80 -1.18** -2.73
DFIN4 -5.83*** -4.64 -5.65*** -3.25
AGE -0.41 -0.62 -0.28 -0.22
DACC -6.92*** -6.34 -7.32*** -3.48
DPAC12 -3.34** -2.61 -3.51 -1.39
DBIGEAST -4.99*** -3.92 -5.79** -2.51
DSEC -6.75*** -6.29 -6.60*** -3.13
DBIG12 -3.95*** -3.22 -4.59** -1.95
-3.57 -1.18 -2.47
DCONFU 3.26 1.10 5.01 0.73
DWEST 3.46* 1.68 2.85 0.68
DCOLON 2.23 0.76 1.89 0.33
DAMER -2.71 -1.59 -4.08 -1.35
DSUN 1.59 0.64 1.57 1.12
DC 0.78 0.40 1.26 0.32
DF 0.16 0 -0.03 -0.01
DG -0.16 -0.13 -0.20 0.62
Observations: 128. *Denotes significance at 10%, ** 5%, *** 1%
Predicting the WNBA Draft: What Matters Most from College Performance?
Volume 10 • Number 4 • 2015 • IJSF 305
Performance from College to the Pros
Are the factors that impact a college player’s draft selection important to subsequent
professional performance in the WNBA? Answering this question requires a measure
of performance.
Berri (2008) demonstrates how the statistics gathered by the NBA (and thus the
WNBA) can be used to estimate a player’s Wins Produced.15 This model of performance
proves very reliable season to season and explains about 94% of the variation in team
wins. Using this metric we now investigate how each drafted player’s Wins Produced per
40 minutes (WP40) might be related to the performance statistics used in the draft model.
Specifically we estimate the following regression:16
WP40n= λ0+ γNPROD + λ1RELHT + λ2DFIN4 + λ3DCHAMP + λ4AGE +
λ5DACC + λ6DPAC12 + λ7DBIGEAST + λ8DSEC + λ9DBIG10 + λ10DBIG12 +
λ11DAMER + λ12DCOLON + λ13DWEST + λ14DCONFU + λ15DSUN + λ16DC +
λ17 DF + λ18DG + e it (2)
where PROD is a collection of player statistics including points, rebounds, steals,
blocked shots, assists, and measures of shooting efficiency.
The estimated results for Equation 2—for the first year of a player’s WNBA
career17—are reported in Table 4.
The results indicate that only PTS, PF, 2FGPER, and DC are significant predictors of
first-year performance. To provide some sense of the relative influence on WP40 of
each of these significant factors an estimated elasticity coefficient is reported, with
shooting efficiency found to have the largest impact. This result runs counter to what
we found when examining where a player is chosen in the draft. That study noted the
primacy of points scored. Although points scored does predict future performance,
our results indicate that more attention should be paid to shooting efficiency.
Clearly, given the results in Table 4, where a player is drafted does not reveal very
much about her subsequent performance in the WNBA. But “how much” is relative.
For an alternative view, consider Table 5, where we consider how much of future per-
formance is explained by where a player is selected.
Table 5 considers two different measures of player performance. The first is WP40.
The second is NBA Efficiency.18 As Berri and Schmidt (2010) note, the former is cor-
related with team wins but not as correlated with player evaluations in basketball. NBA
Efficiency has the opposite properties.
Table 3. The Impact of a One Standard Deviation Increase in Statistically Significant
Performance Variables (2010-2013)
Variable # slots a player gains from + 1 s.d
PTS -8.76
AST -4.95
2FGPER -2.99
PF 1.60
Observations: 128
Harris, Berri
306 Volume 10 • Number 4 • 2015 • IJSF
As Table 5 notes, only about 6% of a player’s career performance as captured by
WP40 is explained by where a player is drafted. Explanatory power increases when
NBA Efficiency is used to measure performance. Of course—as noted in Berri and
Schmidt (2010)—NBA Efficiency is a poor measure of player productivity.
We should note that similar results held in the work done by Berri et al. (2011) with
respect to the NBA draft. This confirms the strange result that—armed with more
information (in-person observations of the player in action and other qualitative
data)—decision-makers do not predict professional performance very well. Our
model of college performance measures seems to predict future performance better
than that used by the teams. Furthermore, management seems to be evaluating female
and male players in a similar fashion. Any bias in the hiring process is clearly linked to
points scored by the player; NBA and WNBA executives alike tend to overemphasize
this statistic. Even though this research does not directly test for gender neutrality it
hints indirectly at it. The absolute dollar value of poor decisions is obviously lower in
the WNBA than in the NBA. Still, important consequences exist for decision-makers.
Losing coaches tend to be fired. Given the short supply of tall productive players it is
reasonable to imagine most teams want to make judicious draft decisions. In this
regard, if these choices reflect the sum of collective wisdom of the team management,
the draft does not seem to really accomplish its intended purpose.
Table 4. How Much Career Performance (WP40) is Explained by Factors Influencing Draft
Variable 1st Year t-stat Elasticity
PTS 0.002** 1.97 1.02
REB -0.001 -0.31
AST 0.002 0.86
STL -0.007 -1.34
BLK -0.004 -0.80
PF 0.008* 1.63 0.53
3FGPER -0.000 -0.02
2FGPER 0.224*** 2.94 2.86
FT -0.020 -0.38
RELHT -0.001 -0.71
DFIN4 0.008 0.78
AGE -0.002 -0.36
DC -0.32* -1.85
DF -0.010 -0.87
DG 0.052 1.03
obs 80
R-squared 0.23
* denotes significance at 10%, ** 5%, *** 1%
Predicting the WNBA Draft: What Matters Most from College Performance?
Volume 10 • Number 4 • 2015 • IJSF 307
Concluding Observations
This research on the draft in the WNBA confirms prior results in the NBA with respect
to performance predictions. Specifically it shows that teams appear to consistently rely
on faulty indicators when making draft picks. For example, factors like Final Four
experience, conference affiliation, and relative height influence draft selection but fail
to accurately predict the players’ performance in the WNBA. In contrast to studies of
NBA performance, rebounds, steals, and turnovers are not an important part of the
story for draft selection in the WNBA. This result could be a function of our smaller
sample size, but may also be indicative of a different type of play on the court. With
more data this is certainly another potential line of inquiry to pursue.
Since evidence of gender bias has been found in a variety of studies, these results
could be important in this field as well. If the performance of female basketball play-
ers is evaluated the same way male performance is then additional studies into other
sports might provide a new laboratory for those interested in measuring productivity
and marginal revenue product.
Given the above, female college basketball players can benefit from two pieces of
advice aside from standing up as tall as possible when their height is officially measured:
(1) score as many points as you can during your college career and (2) do whatever you
can to make sure that career occurs in the SEC, ACC, or Big East. For the time being, it
looks like that’s what matter most to WNBA teams when they make their draft decisions.
Basketball Reference. (n.d.). Retrieved from
Bertrand, M., & Hallock, K. (2001). The gender gap in top corporate jobs. Industrial and Labor
Relations Review, 55, 3-21. Retrieved from
Basow, S. A. (1995). Student evaluations of college professors: When gender matters. Journal of
Educational Psychology, 87, 656-665.
Berri, D. J. (2005). Economics and the National Basketball Association: Surveying the literature
at the tip-off. In J. Fizel (Ed.), The handbook of sports economics research (pp. 21-48). London,
UK: M.E. Sharpe, Inc.
Berri, D. J. (2010). Measuring performance in the National Basketball Association. Working paper.
Berri, D. J., Brook, S., & Fenn, A. (2011). From college to the pros: Predicting the NBA amateur
player draft. Journal of Productive Analysis, 35, 25-35.
Berri, D. J., Brook, S. L., & Schmidt, M. B. (2007). Does one simply need to score to score?
International Journal of Sport Finance, 2, 190-205.
Berri, D.J., Brook, S., Fenn, A., Frick, B., & Vicente-Mayoral, R. (2005). The short supply of tall
people: Explaining competitive imbalance in the National Basketball Association. Journal of
Economic Issues, 39, 1029-1041.
Table 5. How Much Career Performance Can Draft Position Explain (as captured by WP40
and NBAEfficiency)
Year Observations WP40 NBAEfficiency
1st Year out 80 -0.051*** -7.300***
R-squared 0.06 0.14
Harris, Berri
308 Volume 10 • Number 4 • 2015 • IJSF
Berri, D. J., & Krautmann, A. C. (2013). Understanding the WNBA on and off the court. In M.
Leeds & E. S. Leeds (Eds.), Handbook on the economics of women in sports (pp. 132-155).
Cheltenham, UK: Edward Elgar Publishing.
Berri, D. J., & Schmidt, M. B. (2010). Stumbling on wins: Two economists explore the pitfalls on the
road to victory in professional sports. Princeton, NJ: Financial Times Press.
Berri, D. J., Schmidt, M. B., & Brook, S. L. (2006). The wages of wins: Taking measure of the many
myths in modern sport. Palo Alto, CA: Stanford University Press.
Berri, D. J., & Simmons, R. (2011). Catching a draft: On the process of selecting quarterbacks in
the National Football League amateur draft. Journal of Productive Analysis, 35, 37-49.
Coates, D., & Oguntimein, B. (2010). The length and success of NBA careers: Does college pro-
duction predict professional outcomes? International Journal of Sport Finance, 5, 4-26.
Gourieroux, C., Monfort, A., & Trognon, A. (1984). Pseudo-maximum likelihood methods:
Applications to poisson models. Econmetrica, 52, 701-720.
Hollinger, J. (2003). Pro basketball prospectus: 2003-2004. Washington, DC: Brassey’s Inc.
Humphreys, B. (2000). Equal pay on the hardwood: The earnings gap between NCAA Division
I basketball coaches. Journal of Sports Economics, 1, 299-307.
Kahn, L. M., & Sherer, P. D. (1988). Racial differences in professional basketball players’ compen-
sation. Journal of Labor Economics, 6, 40-61.
Marlowe, C. M., Schneider, S. L., & Nelson, C. E. (1996). Gender and attractiveness biases in hiring
decisions: Are more experienced managers less biased? Journal of Applied Psychology, 81, 11–21.
Massey, C., & Thaler, R. H. (2006). The loser’s curse: Overconfidence vs. market efficiency in the
National Football League draft. NBER Working Paper, (W11270).
NCAA statistics. (n.d.). Retrieved from
Oliver, D. (2004). Basketball on paper. Washington, DC: Brassey’s Inc.
Quinn, K. G. (2008). Player drafts in the major North American sports leagues. In B. R.
Humphreys & D. Howard (Eds.), The business of sport, Vol. 3. Westport, CT: Praeger
Staw, M. M., & Hoang, H. (1995). Sunk costs in the NBA: Why draft order affects playing time
and survival in professional basketball. Administrative Science Quarterly, 40, 474-494.
Taylor, B. A., & Trogdon, J. B. (2002). Losing to win: Tournament incentives and the draft lottery
in the National Basketball association. Journal of Labor Economics, 20, 23-41.
Treme, J., & Allen, S. (2009). Widely received: Payoffs to player attributes in the NFL. Economics
Bulletin, 29, 1631-1643.
Weir, K., & Wu, S. (2014). Criminal records and the labor market for professional athletes: The
case of the National Football League. Journal of Sports Economics, 15, 617-635.
WNBA official website. (n.d.). Retrieved from
1A sample of this literature would include Coates D, Oguntimein B (2010) and Berri, DJ, Brook
S, Fenn A, (2011).
2The league began play in 1997. Recent notable events include the Shock relocating from Detroit
to Tulsa, a league-wide multiyear marketing partnership with Boost Mobile landing a logo on
jerseys of 10 out of the 12 teams,; and in April of 2011, the appointment of Laurel J. Richie as
President of the league.
3 The Noll-Scully measure of competitive balance ranges between 1.1 to 2.6 for 1997 to 2013 for
the WNBA; with an average value of 1.9 In comparison, the NBA – from 1997-98 to 2013-14
had a range of 2.3 to 3.4 with an average value of 2.8. Berri and Krautmann (2013) discuss
potential reasons for this difference.
Predicting the WNBA Draft: What Matters Most from College Performance?
Volume 10 • Number 4 • 2015 • IJSF 309
4Much of this talent is found in the college ranks for the NBA and NFL. For MLB and NHL,
high school talent is also considered. The WNBA also primarily drafts from the college ranks.
And like the NBA, the top picks in the draft are assigned via a lottery. Rookies may only sign a
three year, non-guaranteed contract with an option in favor of the team for a fourth year.
5The Berri, Brook, and Fenn (2011) paper looked at drafts from 1995 to 2009. A similar argu-
ment with respect to college scoring was also made by Coates and Oguntimein (2011) in a study
of drafts from 1987 to 1989. For research on other character traits like criminal violations and
draft order see Weir and Wu (2014). Treme and Allen (2009) find faster more accomplished
football players are drafted earlier but 40 yard dash times are not correlated with professional
performance in early career years.
6These authors also cite research that scoring dominates the allocation of minutes and voting
for post-season awards (by coaches and the media).
7For the 2013 season minimum salaries ranged from $37,950 to $55,000 and maximum salaries
range from $105,000 to $107,500. Many players supplement their income by playing in interna-
tional leagues in the off –season.
8See for example Bertrand and Hallock (2001), Humphreys (2000) or Marlowe Schneider and
Nelson (1996)
9This approach follows the lead of Berri, Brook, and Fenn (2011).
10 The -1.57 minimum for blocks occurs due to the position adjustment process. If the “’worst”
blocker in our sample performs well below the average for her position it is possible for the posi-
tion adjustment to return a negative value. See footnote 12 for a full description of how posi-
tion adjustments are made.
11 Position bias is overcome by calculating a position adjusted value for each metric. Each play-
er’s per-minute performance with respect to points, rebounds, steals, blocked shots, assists, and
turnovers is determined. Then, the average per-minute accumulation at each position in our
data set is subtracted. The average value of the statistic across all positions is added back in. After
these steps, the result is multiplied by 40 minutes (the length of a college game), to return the
player’s per 40 minutes production of each statistic.
12 Relative height is determined by calculating the average height—in inches—of the drafted
players in the sample at each position. The position average is then subtracted from each play-
er’s height. The average height in the entire sample is then added back in.
13 As discussed in Berri, et al (2011) we adopted a two-step negative binomial quasi-gneralized pseu-
do-maximum likelihood estimate to correct for overdispersion and to generate a robust variance-
covariance matrix. More information on this estimator can be found in Gourieroux, et al. (1984).
14 Note that for an increase in personal fouls of one standard deviation draft position will
decrease by a little over 1 slot.
15 This model was updated for Berri and Schmidt (2010). Details can be found at http://wage-
16 This is equation (1) with WP40 as the dependent variable (instead of Pick).
17 Relative to the aforementioned work of Berri, Brook and Fenn on the NBA draft, our sample
for the WNBA study is quite small.
18 NBA Efficiency is calculated as follows: PTS + TREB + STL + AST + BLK – All missed shots
– Turnovers. As Berri and Schmidt (2010) note, this model does not explain wins very well. That
is because it de-emphasizes shooting efficiency. But it does a good job of explaining player eval-
uation in the NBA (primarily because the NBA decision-makers tend to de-emphasize shooting
... There are general ideas about attributes which translate into future performance value. Yet, effects tend to be small and oftentimes not generalizable (e.g., Berri, Brook & Fenn, 2011;Harris & Berri, 2015;Moxley & Towne, 2015). Furthermore, this view only considers looking at hard performance data. ...
... Baseball (e.g., Caporale & Collier, 2013;Sims & Addona, 2016), the National Football League (e.g., Hendricks, DeBrock & Koenker, 2003;Massey & Thaler, 2013) and the Women's National Basketball Association (e.g., Harris & Berri, 2015;Hendrick, 2016). ...
... There are general ideas about attributes that translate into future performance value, but effects tend to be small and often are not greatly generalizable (e.g., Berri, Brook & Fenn, 2011;Harris & Berri, 2015;Moxley & Towne, 2015). And this view only considers looking at hard performance data. ...
Full-text available
The thesis explores the intricacies of the NBA Draft. This mechanism is a policy a professional basketball league installed to foster competitive balance within their self-created market. The goals of this regulation are clear and noble. It provides great intentions and attractive potential benefits for most stakeholders involved. The entire league and its franchises would immensely profit from a perfectly functioning draft regulation. Nonetheless, historically it has failed to produce many of the positive outcomes it was intended to provide. This thesis explores this dissonance through the lens of behavioral economics. The application of the NBA Draft policy hinges on human decision-making. However, making choices in a complex environment like professional sports is incredibly difficult. It can be brutally unforgiving due uncontrollable factors. And yet, managerial decision-making quality within the mechanism can almost certainly be improved. The primary objective of this dissertation is to investigate the entire NBA Draft mechanism from a behavioral economic perspective. Using this approach, the overarching goal is to identify segments within the underlying managerial decision-making processes that offer room for decision-making quality improvement. These improvements in judgements and choices which ultimately could lead to a superior policy performance on a league-wide level, could either be achieved due to avoiding error-producing biases or enhancing the information subsequent draft decisions are based on. To reach this main objective, four academic papers were written to tackle important sub-issues. All articles provide sources for decision-making quality improvements within the NBA Draft setup. These are not only supposed to increase the performance of the individual managers and franchises, but also to enhance the results of the overarching league-wide policy with benefits for many stakeholders.
... Berri and Schmidt (2010) demonstrate that statistics can be used to estimate a player's "wins produced" in the NBA through the use of linear regression models. Building upon this work, Harris and Berri (2015) confirm that the same is true for players in the Women's National Basketball Association (WNBA). They also conclude that a subset of a player's college statistics are significant predictors of first-year performance in the WNBA. ...
... Many researchers have sought to model professional productivity using collegiate statistics (see Sec. 2). While Berri et al. (2011) and Harris and Berri (2015) claim a statistically significant relationship between collegiate statistics and professional draft pick, they do not report an adjusted R-squared value regarding the models' goodness of fit. They both show, however, that the same statistically significant factors affecting draft pick explains up to 23% and 36% of the variation in career performance, respectively. ...
Full-text available
We address a team-building problem for a basketball team, where the team decides on new players to draft, current players to trade with those of other teams, and/or which free agents to acquire, to maximize its total value. Here, the team value is considered as the “ability to win a match.” While the individual values of current players in the league are directly obtained from historical data, we propose a predictive model to predict the individual values of collegiate players. We also consider an additional value that comes from the synergy potential among players on the team. We consider “archetypes” of players that represent logical groupings of players based on their characteristics. To estimate the synergy potential between two players, we develop a new metric that encompasses the synergy potential between archetype pairs that each player belongs to. Since the archetypes of collegiate players are unknown, before the synergy estimation, a classification model is performed in order to predict archetypes of collegiate players. These inform a mixed-integer nonlinear programming model for the draft, trade and free agent acquisition problem that maximizes the total team value and balances the synergy among players. We reformulate the objective function in order to overcome computational challenges due to its nonlinearity. Finally, we conduct a case study on a National Basketball Association (NBA) team using 2019–2020 data. Results are discussed and compared with the actual decisions to validate our approach.
... consider factors that influence the NBA amateur draft where teams select players. Similarly, Harris and Berri (2015) examine factors that influence the WNBA draft. While the NBA recruits players directly from high school, the WNBA rarely does so. ...
... The empirical approach is similar to DuMond et al. (2008) and Harris and Berri (2015). Findings from the count model align with some of the results from the DuMond et al. (2008) study but also contrast in intriguing ways. ...
Full-text available
Each year high school football players sign letters of intent with college football programs. The NCAA governs this matching market with strict rules that are designed to protect amateurism. DuMond et al. (J Sports Econ 9(1):67–87, 2008) develop a model of athlete choice. I consider the matching puzzle from the program’s perspective: What factors increase the likelihood that a school will successfully recruit an athlete? Like DuMond et al., I find that the state of play matters. However, my results suggest that football programs are willing to recruit outside their borders. In addition, the results align with prior findings about cheating in the NCAA. This extends the literature on college sport recruiting and may provide insight into other matching puzzles in academic, medical, and business job markets.
... Similar results have been found for drafts in other professional sports leagues. Harris and Berri (2015) (WNBA), Massey and Thaler (2006) (NFL), and Spurr (2000) (MLB), all find significant differences between the characteristics that determine a player's draft position and those that determine his or her success in the league. My paper complements this literature on the relationship between college performance and draft position and future professional performance, by looking specifically at the role teammates play in causing this wedge between what predicts draft success and playing success. ...
Full-text available
Evaluating amateur basketball players based on their college performances is challenging due to the impact of teammates on observed output. In this paper, I look at the effect that college teammates have on a player’s draft position and develop predictions for NBA player performance that account for these effects. I find that players that score more in college due to their teammates are drafted higher, indicating that college choice can significantly impact a prospect’s draft position. I then use the player predictions in a matching model to evaluate how effectively NBA teams account for complementarities between players when evaluating college talent. The results suggest that teams could improve the offensive value added of their draft picks by 20.33% by better taking into account teammate effects.
... Additionally, certain boxscore-statistics do not translate as smoothly from the amateur to the professional level as one would expect (e.g. Coates and Oguntimein, 2010;Harris and Berri, 2015;Salador, 2011). ...
Purpose The NBA Draft policy pursues the goal to provide the weakest teams with the most talented young players to close the gap to the superior competition. But it hinges on appropriate talent evaluation skills of the respective organizations. Research suggests the policy might be valid but to date unable to produce its intended results due to the “human judgement-factor”. This paper investigates specific managerial selection-behavior-influencing information to examine why decision-makers seem to fail to constantly seize the opportunities the draft presents them with. Design/methodology/approach Athleticism data produced within the NBA Draft Combine setting is strongly considered in the player evaluations and consequently informs the draft decisions of NBA managers. Curiously, research has failed to find much predictive power within the players pre-draft combine results for their post-draft performance. This paper investigates this clear disconnect, by examining the pre- and post-draft data from 2000 to 2019 using principal component and regression analysis. Findings Evidence for an athletic-induced decision-quality-lowering bias within the NBA Draft process was found. The analysis proves that players with better NBA Draft Combine results tend to get drafted earlier. Controlling for position, age and pre-draft performance there seems to be no proper justification based on post-draft performance for this managerial behavior. This produces systematic errors within the structure of the NBA Draft process and leads to problematic outcomes for the entire league-policy. Originality/value The paper delivers first evidence for an athleticism-induced decision-making bias regarding the NBA Draft process. Informing future selection-behavior of managers this research could improve NBA Draft decision-making quality.
... Researchers mostly analyze the structure as well as the decision-making within the process or look for inefficiencies from an economic standpoint. Papers on the draft can be found for the NHL (e.g., Deaner et al., 2013;Tingling et al., 2011), the MLB (e.g., Caporale & Collier, 2013;Sims & Addona, 2016), the NFL (e.g., Hendricks et al., 2003;Massey & Thaler, 2013), and the WNBA (e.g., Harris & Berri, 2015;Hendrick, 2016). ...
Full-text available
The NBA installed the draft mechanism to fairly distribute young amateur players among its franchises. As this policy hinges on appropriate talent evaluation skills of the respective organizations, it can be considered a proxy for decision‐making under uncertainty. Such judgments are prone to fallacies and systemic mistakes. The article found the RSCI rank as a problematic metric, which is the source for systematic draft errors. It can be shown that in many cases, managers do not deviate enough from the pre‐draft rankings of players, leading to systematically over‐ and undervaluing certain groups of talents. This can be described as a decision‐quality‐lowering anchoring bias.
Players entering the Women's National Basketball Association (WNBA) are allocated to teams based on a draft where lower performing teams in the regular season earn the highest odds of selecting the best players, creating an incentive for teams to lose games. Using data from the 2000–2019 seasons, results indicate that WNBA teams are more likely to lose after they have been eliminated from postseason playoffs regardless of the gender of the coach or general manager. Evidence of tanking is concentrated in seasons after the WNBA made a draft lottery change in 2015, which was intended to reduce tanking.
Draft decisions by National Basketball Association (NBA) teams are notoriously poor. Analytics can help but are often dismissed for being too overfit, complex, risky, and incomplete. To address these concerns, we train separate leave-one-out random forests machine learning models for each collegiate NBA prospect from 2006 through 2019 with a conservative utility function on a novel comprehensive dataset including the raw text of scouting reports, combine measurements, on-court stats, mock draft placements, and more. Despite being unable to draft high school or international players, the resulting model outperforms the actual decisions of all but one NBA team, with an average gain of $100 million. Target shuffling shows that the model does not overfit and feature shuffling shows that handedness and ESPN mock draft rating, but not other mock drafts, are most important. NBA teams may be missing value by not following a disciplined, model-driven, prescriptive analytics approach to decision making.
Mit der Einführung der Nachwuchs Basketball Bundesliga (NBBL, U19) und Jugend Basketball Bundesliga (JBBL, U16) wurden professionelle Strukturen zur Talentförderung im deutschen Basketball geschaffen. Ein Merkmal dieser Professionalisierung sind Scoutingdaten, die bei jedem Punktspiel erfasst werden. Im Rahmen einer prospektiven Kohortenstudie wurde der Zusammenhang zwischen Scoutingdaten von N=155 Basketballspielern der NBBL-Saison 2010/2011 und dem späteren Erfolg der Athleten im Erwachsenenalter untersucht. Der Erfolg orientierte sich dabei an der höchsten Spielklasse der Spieler in der Saison 2015/16, 2016/17 oder 2017/18. Im Sinne der Prognoserelevanz der Scoutingdaten unterscheiden sich spätere Profis und Amateure hinsichtlich einer Reihe von Merkmalen, besonders bei den erzielten Punkten und der Gesamtleistung der Spieler, welche über das Maß der Effektivität abgebildet wird. Die differenzierte Betrachtung der Profis zeigt, dass sich internationale Profis insbesondere hinsichtlich der Effektivität von nationalen und regionalen Profis abheben, aber keine Trennung zwischen nationalen und regionalen Profis möglich ist. Die Ergebnisse belegen, dass Scoutingdaten mit späterem Erfolg im Erwachsenenalter zusammenhängen und wertvolle Unterstützung bei der Identifikation von Toptalenten bieten können.
Measuring the productivity of labor in sport is often a difficult task. How do we separate the production of a quarterback from his offensive linemen, for example? In water polo, the task is made simple. Water polo is a complex invasion sport like basketball. Thus, we model the offensive and defensive efficiency of teams (as in Berri, Manag Decision Econ, 20(8): 411–427, 1999) in order to determine the impact of individual player effort on wins. From these results, we estimate the team wins produced and predict wins from eight seasons of collegiate water polo.
Full-text available
This is the first paper to formally analyze character discrimination in the specific labor market for professional athletes. We observe all 1,273 players drafted into the National Football League between 2005-2009 to determine what effects character concerns, namely arrest records and team suspensions, have on draft status and performance in the NFL. Prospects that have a history of formal criminal charges or are suspended for team or university violations fall between 16-22 spots in the draft. The impacts of character concerns on performance depend on the nature of the issue. Players that have a history of suspension (non-criminal related) start and play in fewer games per season, but having an encounter with law enforcement does not negatively predict performance. This suggests that having a problem with coaches or fellow players is an indicator of future problems, but having a run in with the law does not adversely affect on the field performance. We also find some evidence that conditional on where players are picked in the draft, teams undervalue those with criminal records and overvalue those with other types of suspensions.
Full-text available
In recent years a wealth of literature has been offered examining the economics of professional team sports. Much of this work follows in the neoclassical tradition, employing the standard assumptions and focusing primarily on the impact of individual decision making. The purpose of this work is to show that the blinders imposed by the narrow focus of the neoclassical tradition on the issue of competitive imbalance may alter the conclusions that a broader view suggests. Economists from the time of Adam Smith have trumpeted the virtues of competition. From the perspective of individual firms, though, profits are typically increased when competition is eliminated. However, in professional sports, the elimination of competition effectively removes the primary source of revenue. In the words of Walter Neale, “Pure monopoly is a disaster. [Former heavy-weight champion] Joe Louis would have had no one to fight and therefore no income ” (1964, 2). The analysis of Neale extends beyond the obvious case of the boxing champion to any professional sport. As noted by Mohamed El-Hodiri and James Quirk (1971, 1306)
Full-text available
The relationship of managerial experience to gender and attractiveness biases was examined in decisions involving suitability for hire and probable organizational progression within a typical financial institution. Each of 112 managers evaluated 4 equivalent résumé-data sheets, to which different candidate photographs were attached. The photographs were varied using a 2 X 2 (Gender X Attractiveness) design wherein each photograph depicted a woman or a man who was either highly attractive or slightly below average in attractiveness. For both ratings and rankings of candidates, clear evidence of attractiveness and gender biases were present. The extent of the bias was generally smaller for the most experienced managers, although less attractive female applicants were routinely at a disadvantage regardless of managerial experience. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Full-text available
Student evaluations completed over a 4-year period at a private liberal arts college were analyzed for the effects of teacher gender, student gender, and divisional affiliation. A significant multivariate interaction between teacher gender and student gender was found for each of the 4 semesters examined. Overall, the ratings of male professors appeared to be unaffected by student gender. In contract, female professors tended to receive their highest ratings from female students and their lowest ratings from male students. This interaction generally remained when possible confounding factors (such as teacher rank) were partialed out. The mean ratings received by female and male professors also varied as a function of the divisional affiliation of the course. Implications of these findings are discussed. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Full-text available
This article examines the earnings gap between male and female National Collegiate Athletic Association (NCAA) head basketball coaches. Within women's basketball, female head coaches earn base salaries 9% higher than males. Estimates from a human capital earnings model suggest that differences in job performance and institution-specific gender segregation explain this earnings gap. The base salaries of women's head basketball coaches of both sexes are also compared to those of men's head basketball coaches. Women's head coaches of either gender earn about 50% of the base salary of the average men's head coach. The earnings differential between women's and men's basketball coaches is attributable to occupation and not the coach's gender, thus prestige from men's sports, customer discrimination, and the preferences of athletic directors may explain this gap.
This article explores the several empirical attempts to quantify the productivity of National Basketball Association (NBA) players, and argues that models should be judged both on the consistency of the measures through time and on the correlation with winning percentages. The scoring models have a good job of explaining player salary. It also appears to be quite consistent over time. The plus-minus approach can explain wins. Wins Produced and Win Shares seem to trump the scoring models or the plus-minus models. Although research in basketball is not without its challenges, the performance metrics the researcher can employ in basketball might be the best player data available in the study of professional team sports.
This study represents one of the first quantitative field tests of the sunk-cost effect. We tested whether the amount teams spent for players in the National Basketball Association (NBA) influenced how much playing time players got and how long they stayed with NBA franchises. Sunk costs were operationalized by the order in which players were selected in the college draft. Draft order was then used to predict playing time, being traded, and survival in the NBA. Although one might logically expect that teams play and keep their most productive players, we found significant sunk-cost effects on each of these important personnel decisions. Results showed that teams granted more playing time to their most highly drafted players and retained them longer, even after controlling for players' on-court performance, injuries, trade status, and position played. These results are discussed in terms of their implications for both sunk-cost research and the broader literature on managerial decision making.