Daniel Ricci 1 1. Introduction In professional sports, the ultimate goal is winning a championship. To achieve that goal, teams must win as many games as possible during the regular season to qualify for the playoffs. Therefore, only in an alternative universe is there an incentive for teams to lose games during the regular season. Yet that is exactly what happens in the National Basketball Association (NBA) year in and year out as losing teams compete for the worst record, in what is otherwise known as “tanking.” Tanking is defined as team management’s intention of doing as little as it can to win. It is a concerted effort for a team to deliberately be just as bad as it can be for several months during the season. The question begs: Why do teams engage in this tanking philosophy? The answer is that tanking is a byproduct of a flawed NBA draft lottery system incorporated by the NBA nearly a quarter century ago. In this system, a team can be rewarded for putting an atrocious product on the floor every night. Therefore, this paper tests whether or not tanking is a successful tactic NBA teams pursue, which rewards teams in the long-term for consistently losing games in the present. Constantly losing conveys an opportunity for general managers to select high in the upcoming NBA Draft once the season concludes. The theory is that the higher the pick in the NBA Draft, the better the player and the more likely that player will become a star in the league. That discussion has accelerated over the last year: many NBA teams last offseason went into “rebuild mode” and released or traded away many of their veteran players that could have possibly contributed to more victories. The incentive to rebuild last offseason and potentially be a bad team for this season likely correlates with the strength of the upcoming 2014 NBA Draft. The 2014 class is projected as the strongest draft class in the last decade with collegiate freshmen
In professional sports, the ultimate goal is winning a championship. To achieve that goal, teams must win as many games as possible during the regular season to qualify for the playoffs. Therefore, only in an alternative universe is there an incentive for teams to lose games during the regular season. Yet that is exactly what happens in the National Basketball Association (NBA) year in and year out as losing teams compete for the worst record, in what is otherwise known as “tanking.” Tanking is defined as team management’s intention of doing as little as it can to win. It is a concerted effort for a team to deliberately be just as bad as it can be for several months during the season. The question begs: Why do teams engage in this tanking philosophy? The answer is that tanking is a byproduct of a flawed NBA draft lottery system incorporated by the NBA nearly a quarter century ago. In this system, a team can be rewarded for putting an atrocious product on the floor every night. Therefore, this paper tests whether or not tanking is a successful tactic NBA teams pursue, which rewards teams in the long-term for consistently losing games in the present.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Daniel Ricci
1
1. Introduction
In professional sports, the ultimate goal is winning a championship. To achieve that goal,
teams must win as many games as possible during the regular season to qualify for the playoffs.
Therefore, only in an alternative universe is there an incentive for teams to lose games during the
regular season. Yet that is exactly what happens in the National Basketball Association (NBA)
year in and year out as losing teams compete for the worst record, in what is otherwise known as
“tanking.” Tanking is defined as team management’s intention of doing as little as it can to win.
It is a concerted effort for a team to deliberately be just as bad as it can be for several months
during the season. The question begs: Why do teams engage in this tanking philosophy? The
answer is that tanking is a byproduct of a flawed NBA draft lottery system incorporated by the
NBA nearly a quarter century ago. In this system, a team can be rewarded for putting an
atrocious product on the floor every night. Therefore, this paper tests whether or not tanking is a
successful tactic NBA teams pursue, which rewards teams in the long-term for consistently
losing games in the present.
Constantly losing conveys an opportunity for general managers to select high in the
upcoming NBA Draft once the season concludes. The theory is that the higher the pick in the
NBA Draft, the better the player and the more likely that player will become a star in the league.
That discussion has accelerated over the last year: many NBA teams last offseason went into
“rebuild mode” and released or traded away many of their veteran players that could have
possibly contributed to more victories. The incentive to rebuild last offseason and potentially be
a bad team for this season likely correlates with the strength of the upcoming 2014 NBA Draft.
The 2014 class is projected as the strongest draft class in the last decade with collegiate freshmen
Daniel Ricci
2
phenoms such as Andrew Wiggins, Joel Embiid, and Jabari Parker expected as the top three
selections.
While team management sometimes believes their best interest is to tank and receive a
top draft pick, the NBA wishes for the opposite. League executives fear teams that eagerly
maximize losses in the present will hurt the product of the league, since fans see the act of
tanking as disgraceful. Therefore, the NBA and each team’s front office face a principal-agent
problem when teams decide tanking for a higher draft pick is beneficial. Long-term profit
motivates both sides, but the league and the team front offices each seek different approaches.
This paper explores that principal-agent relationship and possibly detects if tanking is a
successful tactic from the upper management of NBA teams.
This research paper is divided into seven sections. Section 2 provides an overview and
history of the NBA Draft lottery and also summarizes why teams tank for top draft picks.
Section 3 surveys literature regarding tanking and competitive balance. Section 4 derives an
economic theory related to the principal-agent problem and discusses if tanking in the present
benefits profit in the long-run. Section 5 introduces the methods of the empirical model, while
Section 6 analyzes the results and concludes whether the tanking strategy is efficient. Finally,
Section 7 concludes with a summary of the results and potential future research that builds off
this exercise.
2. History of the NBA Draft
Throughout most of its history, the NBA used a reverse-order draft system to determine
who receives the number one overall pick. In this system, the team that finishes with the worst
record is awarded the number one pick, the second-worst record with the second pick, etc.
Daniel Ricci
3
However, in order to eliminate teams losing purposefully and eradicate their incentive to obtain
the worst record, the NBA implemented an equal draft lottery system in 1985 (Soebbing, 2011).
Under this system, no teams had an incentive to tank because all non-playoff teams had an equal
chance at the number one pick. Yet in the equal draft lottery, teams close to qualifying for the
postseason can win the lottery and draft the best players. Therefore, they can further distance
themselves from the worst teams and competitive balance would decrease.
After only five years of the equal draft lottery system, NBA commissioner David Stern
again changed the system after harsh criticism and comments made around the league. Stern
employed a weighted lottery system, which provided the worst team a greater chance of winning
the NBA lottery. After the Orlando Magic won the 1993 NBA lottery, despite having the best
record of non-playoff teams and only a 1.5% chance of obtaining the number one pick, the NBA
again modified the system. NBA officials overwhelmingly voted to increase the chances of the
number one pick for the team that finished with the worst record. The team that finishes with the
worst record now has a 25% chance of obtaining the number one overall selection. Moreover,
that team can drop to no worse than the fourth selection in the lottery, which is the system that is
in place today.
However, by increasing the odds that the worst team wins the lottery, the NBA is
indirectly reintroducing the incentive to tank. Therefore, the NBA is faced with the decision of
balancing two competing goals. The NBA is set on achieving its first goal, which is to increase
competitive balance by giving the bad teams the chance to land a superstar at the top of the draft.
However, this makes it increasingly difficult for the NBA to achieve the second goal: reducing
the possibility that teams tank.
Daniel Ricci
4
Indeed, since the NBA introduced the weighted draft lottery system in 1990, it appears
they succeeded in increasing competitive balance and giving bad teams the opportunity of
drafting a superstar. The table below shows the percentage of players drafted from 1990-2010
selected to multiple All-NBA teams. The All-NBA team is composed of the top players in the
league, voted upon by a panel of sportswriters at the end of each season.
Draft Selection Percentage of players selected to multiple All-NBA teams
Top 5 65.85%
Picks 6-10 14.63%
Teens 9.76%
Rest of Draft 9.76%
Table 2.1
Thus, an astounding 66% of players named to at least two All-NBA teams were drafted in
the top five of the NBA Draft. This percentage does not include high school stars such as Kobe
Bryant, Amare Stoudemire, and Tracy McGrady – all likely top five selections if required to
attend college for one year due to today’s NBA minimum age rule. The full list of players
accompanied by their respective draft pick and number of All-NBA selections is displayed in the
Appendix. Therefore, while competitive balance increases, the results also establish that teams
have a strong incentive to tank due to the past success of selections in the top five of the draft.
3. Literature Review
3.1. Is tanking really a common phenomenon in the NBA?
Various authors over the last decade explore literature regarding tanking in the NBA.
Select authors believe that teams do not intend to tank during a given season. Gonzalez et al.
(2013) approach tanking from a game theory point of view. They argue that teams often get
stuck in the middle, which is the worst position for an NBA team. Gonzalez et al. (2013) claim
Daniel Ricci
5
that many teams pursue their dominant strategy, which is winning a championship. However,
since many other teams pursue their dominant strategy of winning an NBA title, it creates a
vicious cycle where teams spend decades straddling between the line of barely making the
playoffs or drafting late in the lottery. As long as these teams continue pursuing their dominant
strategy, they guarantee themselves that they won’t be relevant now or in the future. Instead, they
can maximize their payoffs by tanking and selecting high in the lottery as opposed to winning a
championship. Therefore, Gonzalez et al. (2013) conclude that tanking is still not a common
behavior in the NBA; teams not only pursue their dominant strategy, but they also hypothesize
that teams still place high value on the ethics of fair play.
In addition, Deeks (2014) believes that “tanking” is a mythical term that media members
substitute for “rebuilding.” He believes that some NBA teams’ optimal strategy is weakening
their current product in exchange for future flexibility and assets. Deeks (2014) identifies teams
this past season such as the Utah Jazz and Phoenix Suns, in which both teams unloaded a lot of
talent during the offseason to acquire future draft picks and younger players. Thus, this led the
media to believe that both teams tanked during the offseason. However, both teams performed
above expectations this year, and thus Deeks (2014) concludes that: 1) no team is intentionally
trying to tank and 2) youth does not guarantee poor play. Instead, it is a rebuilding process and
supposed tankers are simply better at building a young team through asset accumulation and
management. Instead, Deeks (2014) argues, “The worst team in the NBA…is the one who tries
to be good.” The weakness to his argument is that Deeks (2014) is looking at evidence from only
the 2014 season.
Daniel Ricci
6
3.2. Do teams intentionally tank?
While some authors believe that tanking is not a common practice, many others scholars
find evidence that NBA teams employ a tanking strategy. Taylor and Trogdon (2002) became the
first group of authors to test whether NBA teams intentionally tank. The authors investigate how
likely teams intentionally tank under the three NBA draft formats: reverse-order, equal chance
lottery, and the weighted lottery system. The results demonstrate that teams eliminated from
postseason play are 2.2 times more likely to lose under the reverse order system. When the NBA
changed the draft format to equal chance lottery, the tanking incentive diminished entirely. Yet
when the NBA implemented the weighted lottery system, tanking behavior returned as teams
eliminated from playoff contention became 2.5 times more likely to lose when playing another
team already eliminated from postseason play. This work provides clear evidence that NBA
teams did lose games to improve draft position only in seasons where losing guaranteed a better
pick. This paper refers back to Taylor and Trogdon’s (2002) piece and assumes that all losing
teams tank to improve draft position and therefore acquire a better talent.
Price et al. (2010) employ an empirical model similar to Taylor and Trogdon (2002) in
order to determine tanking behavior of NBA teams. Price et al. (2010) find that no tanking
occurred in the traditional reverse-order draft system (contrary to Taylor and Trogdon) because
in the reverse-order system, the draft order is deterministic and the marginal benefit of shirking –
an increase in one position in the draft – is constant. However, their results are mostly consistent
with Taylor and Trogdon’s (2002) primary findings, where they conclude that teams do have a
strong incentive to tank when there is a lottery system in place. In fact, Price et al. (2010) found
that tanking is a lot more consistent after the introduction of the rookie pay scale in 1995, where
teams no longer needed to pay their rookie draft picks hefty contracts. Therefore, that saved
Daniel Ricci
7
teams substantial amounts of money on their payroll, which resulted in higher profits at the end
of the year. Overall, by promoting competitive balance and also deterring tanking, the NBA
actually created a highly competitive secondary tournament once teams were eliminated from
postseason contention. In this secondary tournament, the losing teams compete for the number
one overall draft selection rather than a championship, which is really the ultimate goal.
Therefore, Price et al. (2010) also contribute to this paper by proving teams clearly have an
incentive to tank and acquire a high-end superstar. In addition, payroll is one of the control
variables further employed in the models in Section 6.
Walters and Williams (2012) compare games at the end of the season in which a non-
playoff team can potentially change its lottery odds by winning or losing games to teams where a
win or loss affects their lottery position. Their results show that playoff-eliminated teams are
14% more likely to lose a game when doing so directly increases their chance of acquiring a
better draft lottery position. They also find that losing teams more likely drop even more games
towards the end of the season via a tanking strategy; therefore, this paper assumes that all losing
teams in the NBA tank at some point in the season. Strictly speaking, even if a team does not
seek out tanking at the beginning of the season, they eventually tank at the end of the season so
that they can receive a better draft pick.
Research extends beyond the NBA sphere to conclude if the NBA is the sole offender of
tanking in professional sports. Borland et al. (2009) explore whether tanking exists in the
Australian Football League (AFL) that utilizes a reverse-order draft system to determine the
number one draft pick. Furthermore, the AFL also incorporates a special assistance draft, which
awards more selections to teams that do not receive a threshold of wins. Therefore, teams should
have an even stronger incentive to tank if eliminated from postseason contention. However, even
Daniel Ricci
8
under both a reverse-order draft system and special assistance draft, Borland et al. (2009) found
that teams showed no signs of tanking in the Australian Football League. The authors investigate
the main reason why there is no tanking behavior in Australian football: there is a difference in
roster size and inability of teams to scout and project talent in the amateur draft compared to the
NBA. Sanderson and Siegfried (2003) originally develop the argument, which the next segment
examines further.
Overall, these studies conclude that tanking does indeed occur in the National Basketball
Association. Therefore, this paper builds off that notion and infers that not only does a losing
team exploit tanking as a common behavior, but also assumes all losing teams tank at a certain
point in the season.
3.3. Why do teams tank and how does it affect competitive balance?
The upcoming literature investigates why teams engage in tanking behavior and how it
affects competitive balance. As Section 2 demonstrates, teams have a very strong incentive to
tank and select at the top of the NBA Draft to acquire a superstar. Sanderson and Siegfried
(2003) hypothesize that NBA teams tank because, “one player could constitute 20% of a starting
line-up (one out of five on the court) and where there is more agreement about a player’s
potential than in football or baseball, which have larger rosters and predictions of performance
are less reliable” (Sanderson and Siegfried, 2003). Therefore, the authors conclude that a top
draft pick should increase competitive balance in the NBA since worse teams have a high
probability of landing a superstar. Therefore, this paper contributes to Sanderson and Siegfried’s
(2003) piece by analyzing whether teams are indeed more likely to draft a better player by
selecting high in the draft.
Daniel Ricci
9
This contrasts with Fort and Quirk (1995), where they contend the amateur draft has no
impact on the league’s competitive balance. Instead, they establish that tanking teams strive for
profit maximization instead of win maximization. In a classic form of monopsony, drafted
players are restricted to only play for the team that drafts them under a low-salaried rookie
contract. Therefore, management purchases a player’s talents at a cheap rate and potentially
resells that talent at a higher price if that player is approaching the end of his contract. Team
executives therefore remain satisfied with profit maximization, whether they obtain a superstar
or not. The following section (Section 4) builds off this knowledge when examining the theory
behind tanking and its relationship with profit maximization.
Soebbing (2011) conducts a study and finds supportive evidence to Fort and Quirk’s
(1995) hypothesis. Soebbing (2011) conducted a difference-in-difference model using data from
the entire regular season history of the NBA (1946-2010) to measure competitive balance in the
NBA. He finds that the amateur draft is an anti-competitive practice used by management to
keep player costs down and improve profits of the owners; it is not in team management’s
interest to fulfill the NBA’s desire to use the draft as a way of increasing league-wide
competitive balance. Teams pursue their best interest, which Soebbing assumes leads to a
potential principal-agent problem between the league (the principal) and the executives (the
agents). Therefore, this paper further analyzes whether this principal-agent relationship exists
and if team management approaches the system in a different way than originally designed by
the league.
Daniel Ricci
10
4. Theory
In this section, this paper develops the economic theory behind tanking. Optimization
theory assumes that agents act rationally, where individuals make decisions that result in utility
maximization. At first glance, it appears that teams act irrational by tanking; rational behavior by
most NBA teams suggests teams should always maximize their wins, which should generate
greater profit. However, a breakdown of the tanking theory demonstrates that losing in the short-
term in the NBA may be beneficial to long-term wins, and therefore long-term profit.
Agents always want to maximize their revenue and minimize their costs to optimize
profit. For simplicity, this paper refers to the agents as the owner, general manager, and head
coach of each team respectively. Each of the three positions’ incentive is pursuing profit in the
long-run. Clearly it is in the owner’s interest to pursue profit in the long-run since that is
understandably his purpose of owning a professional team: making money. For the general
manager and the head coach, it is not as straightforward. The general manager and head coach
theoretically first maximize long-term wins. Once the team maximizes long-term wins, long-run
profit follows since the owner will likely reward the general manager and head coach for their
efforts with a multi-million dollar contract extension. Therefore, in a round-about-way, all three
positions – the owner, general manager, and head coach – have a strong incentive of pursuing
long-run profit.
In economic theory, profit maximization for a firm is typically measured by the following
equation shown in Equation 4.1:
Maximize rKwLpQ with respect to Q, L and K subject to ),( KLfQ
[4.1]
Daniel Ricci
11
Firms want to maximize profit by increasing total revenue (pQ), but labor (wL) and
capital (rK) costs constrain them. This paper builds off the theoretical framework of this formula
in order to explain why NBA teams engage in tanking behavior.
To begin, this paper assumes a two-period model. Teams want to maximize total revenue
in periods one and two, while minimizing costs in those same two periods. For the total cost
function in this model, it is assumed that capital is zero. Equation 4.2 displays the total cost
functions for both periods:
11 * LwTC
22 *LwTC
[4.2]
Therefore, the number of players on the roster and the salaries paid to each player only
determine the total costs. This paper assumes costs are fixed and total costs in Period One equal
total costs in Period Two. Future research could potentially examine if costs should be
incorporated into the model, but for simplicity the paper assumes fixed costs.
With costs fixed, the team’s incentive is maximize total revenue. Total revenue also has
one function in each period. If a team wins more games in Period One )( 1wins , that generates
more total revenue in Period One. The same is true for Period Two, and therefore total revenue is
a function of wins as shown in Equations 4.3:
)( 11 winsgTR
)( 22 winsgTR
[4.3]
The draft pick (draft) that a team receives is also a function of wins in Period One: the
fewer the wins, the higher the draft pick. Equation 4.4 demonstrates this function:
Daniel Ricci
12
)( 1winshdraft
[4.4]
The last function investigates wins in Period Two. A successful team that is winning
games in Period One has a good chance of winning games and success in Period Two, assuming
that most of the same players remain on the team. However, a bad team in Period One also has a
chance of success in Period Two, if they can acquire a superstar with a high pick in the NBA
Draft. Therefore, wins in Period Two )( 2wins is a function of how a team performs in Period
One and also its draft position (h(wins1)). It is defined in Equation 4.5 as:
))(,( 112 winshwinsfwins
[4.5]
The agents in the model want total revenue maximization in periods one and two combined.
Therefore, they are willing to lose more and take less in Period One by tanking if that leads to greater
revenue in Period Two. It is necessary to take the derivative form of the functions and determine the
effect that winning games in Period One has on total revenue in both periods combined. The effect of
winning games in Period One on total revenue in Period One is simple as shown in Equation 4.6:
01
1
1
1
wins
g
dwins
TR
[4.6]
Equation 4.6 is positive, which determines that winning more games in Period One
increases total revenue for that period. However, breaking down the effect of total revenue in
Period Two is more complicated as displayed in Equation 4.7:
Daniel Ricci
13
)(*)(*)()(*)(
1
2
1
2
1
2
wins
h
h
f
f
g
wins
f
f
g
dwins
TR
>0 >0 >0 >0 <0
(1) (2)
[4.7]
Equation 4.7 exhibits the effect of a change in wins in Period One on total revenue in
Period Two. The first half of the equation indicates that an increase in wins in Period One
produces more wins in Period Two, which in turn leads to higher Period Two total revenue.
Therefore, the sign on the derivative for the first half of the equation should be positive. The
second half of the equation suggests that more wins in Period One leads to a worse draft pick,
which leads to lower wins in Period Two, and therefore, lower Period Two total revenue.
Therefore, the sign on the derivative for the second half of the equation is negative. In total,
Equation 4.6 and Equation 4.7 can combine to form Equation 4.8, which specifies the effect of a
change in wins in Period One on overall total revenue ( 21 TRTRTR ):
1
2
1
2
1
1
1
*)(*)()(*)()(wins
h
h
f
f
g
wins
f
f
g
wins
g
dwins
TR
(1) (2)
[4.8]
Tanking only increases total revenue when the first term of the equation is greater than the
second. This is further displayed in Equation 4.9, which suggests that winning fewer games in Period One
contributes more total revenue in both periods combined. Therefore, it is in the agents’ best interest to
tank in Period One only when the absolute value of the first term of the equation is greater than the
absolute value of the second term of the equation.
Daniel Ricci
14
1
2
1
2
1
1 *)(*)()(*)()(wins
h
h
f
f
g
wins
f
f
g
wins
g
(1) (2)
[4.9]
Moreover, if it is in the agent’s best interest to tank, this contrasts with the principal’s interest of
league-wide competitive balance. Conversely, if the first term in the equation is greater than the second
term in the equation in absolute value terms, the team’s incentive should be to win games in Period One
and thus, no principal-agent problem should occur.
Overall, the NBA wants competitive balance and has a certain set of rules that it wants its agents
(the league owners and management) to follow. However, the theory above demonstrates that some teams
may have an incentive of approaching long-term wins, and therefore long-term profit, in a different way
than the NBA initially designed. Strictly speaking, the system currently in place still allows teams to
operate within the rules, but not by the original design of the league. The following section analyzes
whether this hypothesis is true and if tanking does indeed contribute to long-term wins.
5. Data Summary
This paper collects data from the website Basketball Reference, a comprehensive
statistical website of the NBA. The primary purpose is determining how tanking affects long-
term wins. For this paper the long-term is defined as an average total (wins, for example)
between the player’s third season through his last season as a member of the team that drafted
him. The following page further examines the long-term. Therefore, teams that finished with a
bottom-six record in the NBA for a given season are automatically designated with a dummy
variable of tanking, whereas teams that did not finish in the bottom-six are not labeled as tanking
Daniel Ricci
15
for the given season. This assumption draws back to the literature review in Section 3 when
authors such as Taylor and Trogdon (2002) and Walters and Williams (2012) find substantial
evidence that losing teams tank, especially after elimination from playoff contention. The data
begins in the year 1994 since that is the first year that the NBA introduced the current lottery
system. Data is collected up until 2009 to determine the players drafted that season and their
effect on long-term wins.
If players drafted in the top six all played for the team that originally drafted them the
following season, then six more players are collected for the treatment group – teams that did not
finish in the top six. Players from teams that did not finish in the top six are randomly selected in
alphabetical order. If however, a top-six draft pick was traded to a team that finished outside the
top six in a previous season, then that player is not included as an observation; that team would
have no incentive to tank since their draft selection no longer belonged to them. For example, in
1997, the Grizzlies traded an unprotected1 2003 first round pick to the Pistons, which resulted in
the second overall pick in the draft that year. Since the Grizzlies did not own that selection, they
theoretically would have had no incentive to tank for that season; therefore, the Pistons selection
of Darko Milicic at the No. 2 spot is not included in the data set. Instead, the pertaining situation
only calls for five players in both the control and treatment group. In all, from the year 1994-
2009, there are 150 observations (75 tanking teams and 75 non-tanking teams) signifying that
150 players played for the team that drafted them the following season.
The variable longwin is the primary dependent variable in the model and examines how
many wins the team had in the player’s third season with the organization. If the player is on the
1 Unprotected signifies that a team can receive any draft pick via a trade, regardless if it is a lottery pick. Trades sometimes involve protected
lottery picks: If Team A trades a protected lottery pick the previous season to Team B, but Team A finishes with a poor record and ends up in the lottery after the season, then Team A holds on to that lottery draft pick.
Daniel Ricci
16
team that originally drafted him for more than three seasons, then the number of wins is averaged
for all seasons until his last season as a member of the team. For example, long-term wins for
LeBron James is the average number of wins for the Cleveland Cavaliers from 2005 (James’
third year on team) through 2010 (James’ last year on the squad). In a more general example, a
player on a team that wins 30 games in season three, 33 games in season four, and 36 games in
season five – but then signed by a different team after that fifth season – played on a team with a
long-term win total of 33 games. Obviously long-term wins are a great measure of determining a
team’s success, but only for the regular season. Therefore, the variable playoffs measures how
many playoff appearances that the player participated in the long-term on the team that drafted
him. The final dependent variable is winshare, which signifies a player’s long-term contribution
to the team. The variable winshare measures the percentage of a team’s wins that that player
accounts for by himself; it examines the overall effectiveness of the player and how he alone
contributes to long-term wins. Win shares is calculated through an advanced statistics method,
developed by Basketball Reference. The method is similar to the framework developed by the
famous baseball statistician Bill James. To measure win shares, Basketball Reference provides a
detailed formula that accounts for many variables such as points, offensive possessions, and a
player’s marginal offensive efficiency. The measure therefore calculates how many wins that
player alone contributed to his respective team that season. By taking the long-term win share
divided by long-term wins for the team, this estimates the variable winshare for this paper:
The first crucial independent variable in the model is AllStars. The variable AllStars
signifies the number of All-Star players that the examined player played with in his career by
season in the long-term. For example, Blake Griffin played with two others All-Stars in the long-
term – Chris Paul in 2012 and also in 2013. To control for other factors, this papers introduces
Daniel Ricci
17
two other independent variables. PER determines the long-term average player efficiency rating
for a player’s respective team, while also excluding the current examined player. Player
Efficiency Rating is a system designed by former ESPN statistician and current Memphis
Grizzlies Vice President of Basketball Operation John Hollinger, which examines a player’s
productivity on a per-minute basis. Therefore, this variable should control for team quality.
Payroll controls for a general manager and how he builds his team. General Managers that
highly pay their players likely have a better team and therefore, a larger long-term win total. The
final variable, tanking, is a dummy variable; as previously mentioned, it automatically assumes
that a team that finishes with a bottom-six record is described as tanking for that season. Table
4.1 in the Appendix displays a table that calculates the mean values for each variable.
The data appears on first glance that tanking is indeed a favorable phenomenon. Long-
term wins for a tanking team is greater by more than two wins, while tanking teams also are .3
times more likely of going further in the playoffs. The number of players that generate no long-
term wins is greater for non-tanking teams. However, this is likely because teams that draft
players higher may have a “larger leash” – this may be credited to loss aversion. Since teams
invest a high draft selection on the player, they are therefore less willing to cut their losses
because they believe the player still has a high potential. Thus, teams that draft high are less
willing to give up on a top draft pick compared to a team that drafted a player in the mid-to-late
first round who does not have as high of a potential. The biggest discrepancy in the table is likely
the win share, where the player from a tanking team generates on average 11.87% of the team’s
wins in the long-term, while a player from a non-tanking team only accounts for 6.4% of his
respective team’s long-term wins. Tanking teams have fewer All-Stars in the long-term on
average because it is more likely that the player that they drafted in the top six is already an All-
Daniel Ricci
18
Star, making it difficult for a second player on the same team to also become an All-Star. Lastly,
tanking teams on average likely have a lower payroll ranking because the general manager
originally assembled a low-salaried team with worse players, which eventually contributed to a
top draft pick. Regression analysis in the following section further detects if it is worth it to tank
and receive a higher draft pick for a potential higher win total.
6. Results
The Appendix displays full tables with results. In this section, this paper examines the
effect of long-term wins on tanking, since the theory predicts that tanking should contribute more
long-term wins. Model I is a simple model and observes the effect of only tanking on long-term
wins; the linear equation is described in Equation 6.1:
ii1i kingta+=longwin
Equation 6.1
Model II also employs long-term wins (longwin) as the dependent variable, but includes
the following independent variables: tanking (the dummy variable), PER, AllStars, and Payroll.
The linear equation for Model II is revealed in Equation 6.2:
iii3i2i1i +payroll+PER+AllStars+kingta+=longwin 4
Equation 6.2
As Table A demonstrates, predicted theory is not in accordance with the results. The
variable tanking is not significant in both models, and therefore tanking does not guarantee and
increase long-term wins for an NBA team. However, when examining Model II, both PER and
All-Stars are significant: if one more All-Star is added to the team, then that contributes to 1.26
Daniel Ricci
19
more wins in the long-term. If the overall team (outside of the examined player) increases its
average Player Efficiency Rating by a full point, then that contributes to an astounding 7.85 more
wins in the long-term. Therefore, the results can be interpreted in the following manner: long-
term wins are not fully dependent on tanking and which player a team receives in the NBA Draft.
Instead, the wins are determined based on how many other All-Stars are on a team and the
overall team quality, while excluding the player originally drafted.
The results from above are quite surprising, considering that tanking for a player has no
effect on the long-term wins of an NBA franchise. However, does that demonstrate that there is
absolutely no benefit to tanking? The same process from above projects the effect of tanking –
however, this time on the drafted player and his individual contribution to wins (winshare).
Model III examines whether only tanking contributes to a better win share and the equation is
shown in Equation 6.3:
ii1i kingta+=winshare
Equation 6.3
Likewise, Model IV also investigates the effect of tanking on a player’s contribution, but
also include AllStars, PER, and payroll, just like Model II. The linear equation is displayed