Download as pdf or txt
Download as pdf or txt
You are on page 1of 34

Optimal Dynamic Clustering Through Relegation

and Promotion: How to Design a Competitive


Sports League
Martin L. Puterman / Qingchen Wang
University of British Columbia

Abstract

This paper investigates how the structure of a relegation-promotion system


impacts the competitiveness of a sports league. It proposes a rigorous
mathematical model of a multi-division hierarchical sports league made up of
teams with intrinsic skill levels (ISLs) that change from year to year. Since team
skill changes over time, modification in division (or cluster) composition is
necessary to optimize competitiveness. This is accomplished through promoting
teams with the best records at the end of a season to a higher division and
relegating teams with poor records to lower divisions. Such mechanisms are
fundamental to the English football league system and the PGA Tour/Nationwide
Tour. For reasons discussed in the paper, we use data from the National
Basketball Association (in which there is no relegation system) to develop
statistical models for year-to-year variability in ISLs and for match outcomes
based on the ISLs of the two teams. We then develop a multiple season simulation
model to investigate the effect of the number of teams relegated and promoted,
the schedule, and the variability of year-to-year ISLs on competitiveness of the
divisions. For the NBA data, we find that in a three-division league with ten teams
in each division, relegating and promoting three teams at the end of the season
results in the most competitive divisions as measured by the long run average
within-division ISL standard deviation and the percentage of teams assigned to
the correct division. The effect of schedule is minimal.
Introduction

The expressions “relegation” and “promotion” are foreign to most North


American sports enthusiasts. In North American sports leagues, a closed set of
teams competes against each other throughout a regular season, after which a
playoff determines the league champion. However, in Europe and elsewhere in
the world, the standard practice for sports leagues, especially those in football
(soccer), is to have a large open set of teams competing in a tiered-division system
with only the best teams competing in the highest division. At the end of each
season, the top teams in a division are promoted to a higher division and those
who finish at the bottom of their division are relegated to a lower division. In
some leagues there are playoffs, but the relegation and promotion system is
fundamental.
Although relegation and promotion is not explicitly used in North
American sports, a closely related system exists in professional golf. Players
compete in a two-tiered league made up of the PGA Tour and the Nationwide
Tour with player relegation and promotion taking place at the end of each season.
At the end of the 2010 season, the top 125 players on the PGA Tour money list
remained on the tour for the next year (the remainder are relegated out) and the
top 25 on the Nationwide Tour were promoted to the PGA Tour. In addition,
there is a multi-staged “Qualifying School” tournament referred to as Q-School,
with complex rules of inclusion in which the top 25 players and ties are promoted
to the PGA Tour. Connolly and Rendleman (2010) provide an insightful analysis
of the potential errors associated with the existing Q School structure using
simulation and as well suggested some potential restructuring.
The following quote suggests that this concept appears to have some
traction in the business world as well.

“As many as 60 Goldman Sachs executives could be stripped their


partnerships this year to make way for new blood, … (The Globe
and Mail, October 6, 2010, p. L7)”

There are a few significant differences in structure between the European


leagues that use relegation and promotion and North American leagues that do
not. These are most likely the result of divergent evolution between North
American and European sports as they solved problems in different ways. North
American sports leagues do not use a tiered-division system and the only way to
enter a new team is through acquiring a franchise from the league ownership and
then creating a team in that city subject to stringent league rules. This involves a
significant entry fee, and more importantly, would likely not be approved if the
new team was to take away geographical market share from any existing team.
The rules governing geographical markets make it extremely difficult for any
team to enter the North American sports’ major leagues and keep league sizes to
around 30 teams. European leagues, such as the English football league system,
do not honor geographical market rules, and entry into the top division, which is
referred to as the Premier League, is achieved through merit. There are over seven
thousand teams in the English football league system, and every team has the
opportunity to play in the Premier League if they are able to successfully climb up
the ladder system.
Another significant difference in structure is the North American leagues’
attempt to balance competition via salary caps, revenue sharing, and reverse order
player drafting. The salary cap is used by the NFL and requires teams to spend no
more and no less than specified amounts of money on player salaries each season.
This prevents wealthy teams from gaining a competitive advantage, and ensures
that no team is putting finances ahead of performance. Revenue sharing is another
system that helps financially strapped teams to be competitive. With revenue
sharing, all teams within the league receive a share of the total revenue generated
by the league, providing poorer teams more money than they could earn on their
own. Finally, reverse order drafting awards teams with weak records higher draft
positions thus enabling them to improve in the subsequent season.
Last but not least, the existence of playoffs to determine championships in
North American leagues creates a different form of measuring merit than that in
European leagues. The first and foremost goal of teams in North American
leagues is to win the championship, followed by making the playoffs. So while in
European leagues the 16th and 17th best teams may have minimal separation in the
eyes of fans, the same regular season results in a North American league such as
the NBA could be the difference between success and failure since the 16th place
team will make the playoffs while the 17th place team will not. This principle
forces teams within reach of the playoffs to make personnel decisions that will
help them win in the short term, while teams that have no chance to make the
playoffs will make transactions that reduce immediate quality but may have long
term benefits.
Relegation and promotion works like this: at the end of each season, the
bottom teams from each division are relegated (moved) from that division into
the next lower division and the top teams from that division are promoted
(moved) into the next higher division. Thus, if a team performs poorly, it will be
demoted into a lower division, and if a team performs well it will be promoted to
a higher division. In leagues such as the English football league system, divisions
have different numbers of teams that they relegate and promote, but to keep the
division size constant, the number of teams moved out from one division must be
the same as are brought in. To be in a superior league is of great value to the team
and fans, but more importantly, there are significant financial implications of
being relegated or promoted (“Newcastle reveals the financial cost of relegation,”
Ian Ferris, www.tribalfootball.com, May 11, 2009).
Putting aside the distinct differences between North American and
European sports leagues, the relegation and promotion system brings to European
sports some obvious benefits that North American sports lack. In North American
sports, games played between two weak teams during the latter part of the season
when neither are in contention for the playoffs become generally irrelevant to fans
resulting in poor attendance. For example a late 2009-2010 season National
Basketball Association (NBA) game between the Minnesota Timberwolves and
the Los Angeles Clippers, or a Major League Baseball (MLB) game between the
Cleveland Indians and Kansas City Royals draws minimal crowds, and give the
public little reason to pay attention other than as basis for gambling. More
importantly, when there is nothing on the line, there is less incentive to win, thus
diluting the product. In the English Premier League in which the bottom three
teams are relegated to a lower-tiered division, fans pay attention to late season
games involving weak teams because losing means being relegated. In order to
avoid such a fate, games are more competitive, creating a better product for fans.
The underlying goal of relegation and promotion is to maintain
competitiveness in the face of changing team skills. Assuming that there is
overlap of quality between teams playing in different divisions (i.e., the weaker
teams of division one and the stronger teams of division two), employing
relegation can result in more competitiveness within divisions as teams will
continue to move up or down the league hierarchy. By putting teams of similar
skill levels in the same division, the competition level will increase. Therefore, we
would see a team like the Kansas City Royals win a few more games than they do
now if MLB were to employ a tiered-division system based on team quality. This
could have a positive effect towards the progression of its management, coaches,
and players while boosting fan morale.
While we believe relegation and promotion can be beneficial to North
American sports, our main objective in writing this paper is to provide a rigorous
model of a tiered multi-divisional sports league that enables us to address the
question “what is the optimal number of teams to relegate and promote?” To
achieve this, we require measures of competitiveness, a model for year-to-year
variation in team intrinsic skill levels and a model of how outcomes of individual
games are affected by the relative skill of the two teams. Using these models and
data from the NBA (we will discuss why we chose this as our data source below)
we develop Monte Carlo simulation models and play out thousands of seasons to
determine the effect of the number of teams relegated and promoted, the schedule
(all games within a division, a balanced schedule or a mixed schedule), and
variability of year-to-year intrinsic skill levels on competitiveness. To our
knowledge, this is the first paper to address this question and to propose a
mathematical model for a sports league.
In essence our problem is one of dynamic clustering in which each entity’s
attributes are partially observable and change dynamically. Moreover, the clusters
are ordered and of fixed size and can only be modified through relegation and
promotion between clusters. In that setting the objective might be to minimize the
long run average of within cluster variability. To the best of our knowledge, this
problem has not been studied either.

Related Literature

Competitive balance within sports leagues has been studied in the economics
literature. Moreover, because North American sports leagues are seen as
monopolies or cartels due to their closed structure, individual league decisions
such as the reserve clause, free agency, salary cap, revenue sharing, and win-
oriented owners have been closely monitored on their effects towards competitive
balance. Also, the competitive balance of open league systems that utilize
relegation and promotion has been analyzed and compared to North American
sports leagues (Szymanski and Smith, 2002, Buzzacchi et al., 2003, Szymanski
and Valetti, 2005).
However, these analyses do not isolate relegation and promotion as the
only independent variable, and are highly specific to the sport and other league
characteristics. Theoretical models (Noll, 2002) suggest that the introduction of
relegation and promotion will increase spending, but will not directly translate to
competitive balance.
Medcalfe (2003), in an unpublished paper, studied the effects of
implementing relegation and promotion and found evidence that the introduction
of relegation and promotion does increase competitive balance. However, he did
not study the effect of the number of teams relegated and promoted on
competitive balance of the league, and he did not track team skill level dynamics.
In a surprising article in Nature, Juhas et al. (2006) (see also the comment
by Rehr (2006)) point out that the problem of reconstructing three-dimensional
non-crystalline structures from 'pair distribution function' (PDF) data has
similarities to sports relegation and promotion rules. Juhas refers to their
approach for inverting PDF data as the 'Liga algorithm', because it is modelled on
the rules of relegation and promotion used in most of the world's football leagues,
including La Liga in Spain. Teams correspond to trial clusters of atoms; 'winning'
clusters (those with the smallest errors between the model and the experiment) are
iteratively promoted, whereas losing ones (those with the largest errors) are
relegated, so that an optimal structure is quickly found. The authors show that
their algorithm can determine a number of nanoscale structures with perfect
success rates.

The League and Competition Model

To answer our questions and in order to provide a framework that lets us extend
our analysis to other sports, we propose a mathematical model of an abstract
hierarchical sports league managed by a relegation and promotion system. In
order to generate results that can reasonably reflect a real life sports league, the
following elements are fundamental:

 a measure of team intrinsic skill level (ISL),


 an ISL year-to-year dynamic model (ISDM), and
 a match outcome determination model (MDM).

Our proposed model follows. Let denote a set of sports teams. Each
team possesses an intrinsic “power” or “skill” rating throughout
season , where = 1, 2, …, N. This ISL is not observable but manifests itself
through outcomes of games or matches that take place throughout a season
consisting of games. Teams are divided into a hierarchy of divisions where
division 1 contains the strongest teams and division the weakest teams, with the
relative strength of teams in divisions declining in between. A schedule
determines which teams play each other during a season. Possible schedules
include playing only divisional opponents, playing all teams in the league, and
some mix of the two. We assume for simplicity that all teams in the same division
play the same opponents and that each pair of teams play an even number of
games equally divided between home and away games (this avoids home
advantage effects, a point we elaborate on below). We now expand on our
description of the key model ingredients.

The ISL

In professional sports, a wide range of rating systems have been developed to


determine the relative strengths of different teams or players. Usually they are
based on within season performance, strength of opponents, or team attributes.
And they are highly subjective. From our perspective the source of the rating
system is unimportant, what we require is that it provide a basis for determining
the likelihood of match outcomes and for measuring competitiveness of the league
structure. We will restrict ISLs to be real numbers between 0.0 and 1.0, with
higher values corresponding to stronger teams, however in some sports such as
golf, it might naturally be represented on a different scale.
The ISDM

The ISL for each team varies from year to year. This variability provides the
motivation for relegation and promotion because if ISLs remained constant, no
relegation would be necessary provided the initial assignment of teams to
divisions was correct. In reality, the main sources of variability are player specific
factors including within- and between-season trades, player acquisitions, injuries,
and skill progression or decline; and team factors including coaching,
management changes, and team chemistry plus other intangibles. Unfortunately
such factors cannot be easily modelled, especially in the context of a simulation
model, so instead we propose using a dynamic probabilistic model for the team
ISLs which relates next season’s ISL to this season’s ISL plus other factors that
can be modelled in a simulation. We show below that for NBA data, an AR1
model (Granger and Newbold, 1986) describes the dynamics well, but in other
settings other models may be more appropriate, especially in individual sports
which must take player career trajectories into account. One parameter we will
control for in our study is the variability in the ISDM.
We make the strong assumption that intrinsic skill level changes only after
all games in a season have been played and that it remains constant within a
season. Because both the skill level and the AR1 model are generated through
end-of-season results, the end-of-year skill level change is actually an average of
skill level changes within a season. Therefore, in the long run, we should not
expect skill level changes within a season to affect the overall outcome of our
model.

The MDM

The ISLs of the two teams competing in a match determines the probability of a
result (say team wins) which we restrict to be either a win or loss. The MDM
provides an estimate of this probability. We propose using logistic regression to
model the probability of this outcome although other functional forms may also
be appropriate.

The Simulation Model

We use the Monte Carlo simulation model, schematically represented in Figure 1


to investigate the impact of relegation and promotion (RP) rules, schedules, and
variability in ISLs on competitiveness. One of the primary advantages of using a
simulation model is that the ISLs are known and hence can be used as a true
measure of competitiveness that would be unobservable in the real world.
Figure 1: Schematic representation of league simulation model

The simulation is initialized by assigning an ISL, , to each team in


and setting the season number indicator to zero. The teams are then sorted by ISL
and assigned to the divisions in ISL order. A schedule and RP rule is specified
and the season is played out as follows:

1. Start of New Season - Increment season number indicator by 1 and set


team records to zero-zero.
2. Simulate Season Play - Matches are simulated using probabilities
generated from the MDM and then drawing a randomly generated integer
between 0 and 99 and assigning a win to the team whose assigned range
(represented by its winning probability) contains the generated number
and a loss to the other team. The schedule type and the number of games
to play dictate which teams play which. The simulation ignores the effects
of streaks and trends within a season and assumes that all match outcomes
are independent. Team records are updated after each match.
3. Generate Season Results - After season play is complete the average skill
level of teams within each division, standard deviation of skill level, and
the number of teams in each division that are correctly classified are
computed and stored.
4. Promotion and Relegation - Teams are relegated and promoted based on
the specified RP rules and the final league standings. The top teams are
promoted to a higher division, and lowest teams relegated to a lower
division. No playoff system is used but one could be easily added.
5. Update ISLs - Teams’ skill levels are updated using the AR1 model
(defined formally below) in which estimates of each team’s mean,
autoregressive parameter and previous season ISL are used in the
deterministic part of the formula and the random variation term is
generated from a (0, RMSE) distribution.

The simulation is run through a warm up period of seasons to remove


initial effects and then a large number of seasons are played and performance
measures compiled. We then change the input parameters and rerun the
simulation.

Managerial Levers

League management has three levers to use to achieve competitiveness goals; the
division size, the number of teams to relegate and promote, and the schedule.

Division Size

The teams can be divided into any number of divisions but the divisions should
be large enough to allow a reasonable season length based only on a within-
division schedule and to maintain between-season stability, but small enough to
be competitive. If there are too many teams in a division, the team ISLs will vary
considerably and some matches will become rather one-sided. In a league with 30
teams, we believe three 10-team divisions would be reasonable.

Relegation and Promotion Rules

We assume for simplicity that all divisions are the same size and the same number
of teams will be relegated and promoted from each division. Of course there is no
promotion from the top division and no relegation from the lowest division. As
noted earlier, if team ISLs were constant from year to year, then no relegation and
promotion would be necessary provided the initial assignment of teams to
divisions was correct. On the other hand, if the ISLs varied greatly from year to
year, then it might be optimal to relegate and promote a large number of teams
each year. Reality is somewhere in between.

Schedule

In North American sports leagues, scheduling often has a profound effect on team
performance. In the NFL, the divisional winners in the previous season play
against other division winners in their inter-divisional matches in the subsequent
season. A schedule composed of stronger opponents will naturally impact a
team’s performance making it difficult for that team to perform at the level it did
in the previous season. Meanwhile, weak teams will play against other weak
teams in their inter-divisional matches, allowing for a better chance to achieve
better results. Since the playoff system determines the eventual champion, it
ultimately offsets biases that result from these unbalanced schedules.
The English football league does not use a playoff system to determine the
champion; instead, it uses a within-division schedule in which teams play each
other twice. This way, teams play the same opponents, so that the schedule does
not have a considerable impact on seasonal results. This is important because the
teams that get relegated should be the weakest teams in the division, and the
teams that get promoted should be the best. One drawback of this schedule is that
fans of teams that are not in the higher division cities do not get to see the top
teams in person. This deficiency is lessened due to England’s size. There are 20
teams in the English Premier League, where the population is slightly over 50
million and the area is approximately 130,000 square km. Compare that to the US
where population is about 300 million and the area is about 9.8 million square km.
Therefore, the ratio of Premier League teams to the population and area of
England is much greater, so fans have better accessibility to the top teams.
We will consider three different balanced schedules where by the
expression balanced we mean that each team within the same division will play
against the same opponents. Further, in practice we would prefer an even number
of games against each opponent to balance home advantage. If an odd number of
games are to be played, the schedule should ensure an equal number of home and
away games. We classify the schedules as follows:

 Within-Division - Teams play games against only teams within their


division.
 Half-In-Half-Out - Teams play approximately half of their games against
teams within their division, and the remaining against each other team in
the league twice.
 Inter-Division - Teams play the same number of games against every team
in the league.

The reason for considering three schedule structures is to determine


whether playing outside of the division has any effect on competition levels and
relegation and promotion. Since teams of the same division still play against the
same opponents, we expect there to be none. In a league such as the NBA where
many games are played each season, a within-division schedule would not be
appealing to fans. Therefore, if we conclude that the schedule doesn’t impact
competitiveness then a Half-In-Half-Out schedule might generate the greatest fan
appeal.
Performance Metrics

Two commonly used measures of competitive balance in a sports league are the
standard deviation of winning percentage of its teams and the ratio of actual to
idealized standard deviation of winning percentages (Zimbalist, 2002). Since in
empirical data analysis the team winning percentage is affected by the nature of
the sport and the league (for example the NBA has greater variation than MLB)
independent of league rules, it is difficult to compare competitive balance
between different leagues, leaving appropriate forms of measurements to much
debate. However, in our model the intrinsic skill levels are observable so the
following performance metrics are available.

Mean and Standard Deviation of Within-Division ISLs

For each division in each simulated season, we record the mean and standard
deviation of ISLs within each division and the average of these quantities over the
data generated after the simulation warm up period. Ideally, we would like the
average mean divisional ISLs to be ordered in the same way as the divisions and
the average standard deviation of within-division ISLs to be small. Our
assumption is that low within-division ISL variability corresponds to more
competitive divisions. Since the MDM was found empirically (see below) to
depend on the difference of ISLs, this adds further support to this observation
since if two teams have equal ISLs, each has a probability of .5 of winning a
match. Hence, if all other parameters are constant and relegating and promoting
two teams per division produces a lower within-division ISL variation than
relegating and promoting three teams, we conclude that relegating and promoting
two teams produces a more competitive league. Note that we compute these
measures for each division separately but we could also average them over
divisions to get a composite measure of variability.

Correct Rankings

Another metric we will use is the number of teams in a division that are assigned
to the appropriate division on the basis of the ISL. For example, if Division 1 has
ten teams and only six of them are among the top ten in the league in terms of the
ISL, then we say that six teams are correctly assigned. The purpose of this metric
is to measure how well the relegation and promotion system clusters teams into
divisions. In an ideal system, the best teams will be in the top division and the
weakest in the lowest division. We will focus primarily on values of this metric
for Division 1.
Applying the Model: National Basketball Association

Although our model is intended to be a representative of an abstract sports league,


we used data from the NBA for our empirical study. Ideally, for our study, it
would be best to model our sports league after an existing sports league that
employs the relegation and promotion system. However, in such leagues the
number of official matches played between teams in an inter-divisional setting
was limited so we were unable to gather enough data to accurately model relative
skill level and competition. As a result, we chose to base our analysis on the NBA
because of the readily available data, our own background knowledge of the
league, and the recent reality that the NBA has become a league of elite, mediocre
and poor teams. We used NBA data from 1979-2009 for our analyses. We now
discuss our models for the league components.

Team Intrinsic Skill Level

Deriving reliable skill level models is challenging and the subject of extensive
debate and press. For example http://espn.go.com/nba/hollinger/powerrankings is
one of many such rankings for the NBA. To objectively measure a team’s skill
level requires comparisons between individual players, team chemistry,
performance, schedules, etc. Development and critique of such ratings are beyond
the scope of our study, but nonetheless such ratings are fundamental to our model.
Since a team’s demonstrated performance (measured by its winning
percentage) is correlated with its ISL, we can view the ISL as a function of
performance. Therefore we equate the ISL with the team winning percentage.
Figure 2 summarizes average winning percentages for the 30 NBA teams between
1979-2009. It shows together with Table 1 below, that average historical win
percentage (divided by 100) varies between .35 and .66 and that the league can
naturally be clustered into three divisions of 10 teams each based on these
percentages. As noted above we will use relegation and promotion to modify
cluster membership each season.

The ISL Dynamic Model (The ISDM)

Year-to-year team skill level dynamics is arguably the most important part of
building an accurate simulation. Skill level dynamics differ for different sports,
and can be highly complex. For individual sports such as golf and tennis, skill
level dynamics must take into account a player’s progression and career cycle.
Histogram of Average NBA Winning Percentages
10

Count
5

0
0.30 0.37 0.43 0.50 0.57 0.63 0.70
ave_win_pct

Figure 2: Histogram of NBA historical average winning percentages (divided by


100).

However, for professional team sports, many other factors come into play such as
player acquisition and trades, player progression/career cycle, player injuries,
coaching changes, team chemistry, and ownership/philosophy changes. Especially
in a sport following competition balancing rules of reverse order drafting or the
salary cap, events such as an injury to a star player could have dramatic
consequences. For example, the season ending injury to superstar David Robinson
in 1996-1997 resulted in the San Antonio Spurs going from a top team with a .72
winning percentage to an extremely poor team with a .24 winning percentage, but
also resulted in the drafting of Tim Duncan, extending their run of excellence to
13+ years including four NBA league championships.
The reason this is important for building a realistic model is because of the
effects that it can have on league competition. For example, if we set the variation
of dynamics too high (suggesting that it is common for a team to go from .72 wins
to .24), no amount of relegation and promotion will be able to impact league
competition because quality in a given season will not be an indicator of quality in
the subsequent season.
To build a year-to-year team skill level dynamic model, we gathered
winning percentage data (expressed as decimals between 0 and 1) for all NBA
teams from the 1979-1980 season to the 2009-2010 season. As noted above we
used this as the basis for the ISL. Our statistical analysis showed that year-to-year
variability in individual team winning percentages was well described by the
following first order autoregressive (AR1) model:

where:
 = winning percentage of team in year
 µ = mean winning percentage for team
 = autoregressive parameter with values between -1 and 1
 = a term representing unexplained variation in year . It is assumed
to be normally distributed with mean 0 and standard deviation σ. Note that
σ is a parameter we will vary in our simulations.

We fit time series models to this data using the statistical software package
NCSS. In all cases after fitting the AR1, there was insufficient evidence to reject
the hypothesis that the residuals were white noise (based on the residual
autocorrelation functions, the Box-Ljung test statistic and also fitting either MA1
or AR2 terms). Estimates of the model parameters (where σ is estimated by the
RMSE) appear in Table 1. They show that historical average winning percentages
varied between 0.346 (LA Clippers) and 0.662 (LA Lakers) with a mean of 0.503.
The estimates of varied between 0.087 (Dallas Mavericks) and 0.882 (Utah
Jazz) with a mean of 0.552 and the RMSE’s varied between 0.081 (Utah Jazz) and
0.133 (Chicago Bulls) with a mean of 0.110. This analysis suggests considerable
variability in team underlying quality as measured by the historical average
winning percentage, streakiness or persistence as measured by , and deviations
from the model as measured by the RMSE. For example, the Utah Jazz had a
historical mean winning percentage of 0.573, an estimate of of 0.883, and a
RMSE of 0.081. This means that when their within-season winning percentage
exceeded 0.573 by a large amount, there is a high probability that the next
season’s win percentage will also exceed 0.573 because of the large estimate for
and the small RMSE.
To use the fitted AR1 models in the simulation, we must address the
following subtle issue: do we generate next year’s ISL by substituting the
previous year’s ISL estimate, or the actual winning percentage, into the above
model? We believe the former may be more appropriate because the actual
winning percentage may not reflect the true intrinsic skill level because of the
interaction between the schedule and the divisional structure. For example, if all
matches are played within the division, then a weak team in a strong division
could have a lower winning percentage than a strong team in a weak division, so
that in this case the actual winning percentage might not reflect the ISL.
Conversely, when each team plays all other teams, either method would be
appropriate.
Team Mean Alpha RMSE
Boston 0.575 0.648 0.13
Philadelphia 0.516 0.834 0.097
Washington 0.42 0.419 0.092
New York Knicks 0.49 0.603 0.11
New Jersey 0.404 0.496 0.116
Miami Heat 0.4 0.8 0.126
Orlando * * *
Toronto * * *
Charlotte * * *
Atlanta 0.5 0.596 0.107
Milwaukee 0.507 0.702 0.094
Chicago 0.509 0.693 0.139
Cleveland 0.484 0.69 0.114
Indiana 0.494 0.622 0.0975
Detroit 0.542 0.712 0.121
Denver 0.465 0.594 0.116
Utah 0.573 0.882 0.081
Dallas 0.496 0.0867 0.114
Houston 0.541 0.161 0.113
San Antonio 0.603 0.411 0.133
New Orleans * * *
Minnesota * * *
Memphis * * *
Seattle 0.54 0.559 0.113
Phoenix 0.592 0.311 0.115
Portland 0.564 0.634 0.088
LA Clippers 0.346 0.403 0.105
Golden State 0.41 0.135 0.12
LA Lakers 0.662 0.498 0.094
Sacramento 0.449 0.759 0.095
Mean 0.503 0.552 0.109
Std Dev 0.074 0.216 0.0149

Table 1: Parameters from AR1 model of NBA winning percentages. Teams


indicated by * had insufficient data to fit a reliable AR1 model.

The other challenge in using this model is that over the long run it will
generate winning percentages below 0 and above 1. This can be corrected
analytically by using a logistic transformation of the winning percentage but there
are other problems too. Historically, there has never been a team with a winning
percentage of below 0.10 or above 0.90 in any season (the worst record was the
1972-1973 Philadelphia 76ers who finished 9-73, and the best record was the
1995-1996 Chicago Bulls who finished 72-10). We do not expect more extreme
results than this because of the nature of the business. There is no motivation for a
team that has already secured first place in its division to compete aggressively
and risk injury. Also, professional pride will motivate players on teams with poor
records to compete harder.
From an economic perspective, it would be economically inefficient for a
team to invest so much that it is significantly better than every other team when
the league success is measured by winning the division and not by dominating
every game. If winning a championship can be achieved with a 0.75 win
percentage, there is no need for a team to try for a 0.90 win percentage. Therefore,
we added into our model limits that prevent a team from ever obtaining a skill
level of less than 0.10 or greater than 0.90. To achieve this, when generated skill
levels were below 0.10 or above 0.90 we simply generated new realizations of
until the value fell within this range.
We use the above model as follows. Suppose team has ISL 0.70 in year
, its historical mean equals 0.55, equals 0.6, and that the sample from a normal
distribution with mean 0 and standard deviation equal to RMSE (which is
assumed to be 0.08) is 0.04, then the ISL in year +1 becomes 0.6 * (0.70 – 0.55)
+ 0.04 + 0.55 = 0.68. By applying this equation to our simulation for each team at
the end of a season and resampling when necessary, we have a model that we
believe is a reasonable reflection of skill level dynamics in the NBA.
As noted above, in practice, team ISL dynamics are affected by many
factors. In the AR1 model, the expression , combines a wide range of
intangible and unpredictable factors that cause teams to differ from year to year
including retirements, trades, improvement or deterioration in individual players
skills, new coaches or new management. In our experience and in light of our
purpose in developing an ISDM that can be applied in the simulation, the AR1
model is better suited to our needs and is probably better in forecasting future
ISLs than a model which takes team characteristics into account.

The Match Outcome Determination Model (MDM)

We believe that logistic regression provides a sufficiently rigorous model to


estimate the probability of a match outcome when two teams compete. We use the
team ISLs as explanatory variables and ignored “home advantage” since we are
planning to run our simulation over a large number of seasons and with a
balanced schedule where each team has a home and away game against whichever
team it plays. We assume that the outcome of a match between team and team
in year results in team winning (assuming ties cannot occur) with probability
follows a Bernoulli distribution, where
After extensive empirical analysis of the NBA data, we find that the following
model best fits the data

As a result of this, we can compute by transforming the above equation to


obtain

or

This is a very plausible model for the probability of match outcomes.


When the ISLs of the two teams are equal, the probability that each team wins is
.50. This observation provides theoretical justification to support the empirical
evidence that equals 0. Also the probability of winning depends only on the
difference in ISLs between the two teams which is reasonable when home
advantage is ignored. Further empirical evidence suggests that quadratic terms in
the difference do not improve model fit.
We used NCSS to estimate the parameters and for the above model
from complete NBA match data containing 12,136 observations over the 2000-
2001 through 2009-2010 seasons (data obtained from www.basketball-
reference.com). We found that in the original model was highly insignificant
(p=0.96), that we could not reject the hypothesis that = (p=0.72) and that the
estimate of equalled 4.594.

Divisions, Schedules and RP Rules

We divided the NBA into 3 divisions of 10 teams each because it makes the most
sense from the perspective of the goals of this study, is supported historically by
Figure 2, and reflects reality in 2011. Larger divisions will make scheduling
challenging and be less competitive, while smaller divisions would be more
competitive, less interesting to fans, and vary highly year to year under RP rules.
From a scheduling perspective, several options are available in agreement with
the above three scheduling principles. We consider two possibilities depending on
whether or not we wish to have an equal number of home and away games for
each team. This will not affect the simulation model where home advantage is not
considered in the MDM but will have significant practical implications when the
home advantage is real.

 Within-Division - To balance home and away effects, we consider a 72-


game schedule with eight games against each team. If home advantage is
ignored we could consider an 81-game schedule with 9 games against each
team.
 Half-In-Half-Out - Either: (a) 4 games against each divisional opponent
and two against each opponent outside the division for a 76-game
schedule, or (b) 5 each against teams within their division and 2 each
against the rest of the teams in the league for an 85-game schedule.
 Inter-Division - Teams play either 2 or 3 games each against every other
team in the league, constituting either a 58- or 87-game schedule.

Relegation and Promotion (RP) rules

Since we are considering a three-tiered league with ten teams per division, we
vary the number of teams to relegate and promote between zero and five per
division and assume further that all divisions relegate and promote the same
number of teams. Further, we assume no playoffs, so that the end of the regular
season standings provide the basis for relegation and promotion.

Results

We implemented the computer simulation described above in Java to test the


effects of relegation and other factors on competitiveness. The 30 teams were
divided into three divisions on the basis of historical mean winning percentage or
ISL. Teams were directly modeled after existing NBA teams, with initial skill
levels set equal to their respective winning percentages during the 2009-2010
season. Values for historical mean, AR1 parameter α and RMSE were obtained
from fitted values in Table 1. Teams without sufficient historical data were
assigned values of equal to 0.50, RMSE equal to 0.10 and historical means
close to 0.50 so that the overall historical mean for the league equalled 0.50.
Alternatively, these remaining values could be sampled from the empirical
distributions implicit in Table 1.

The simulation varied the following quantities:

 Schedule (within-division play, half-in-half-out, inter-division play),


 Number of teams to relegate and promote (0 to 5 per division), and
 Variability in year-to-year skill level dynamics denoted by:
o standard: estimated RMSE values
o stable: one half the estimated RMSE values and
o unstable: twice the estimated RMSE values.

There were a total of 54 combinations of factors with each combination replicated


20 times. Each replication consisted of 1000 seasons with the first 100 seasons
removed to allow the league to stabilize. Means of the following performance
measures were computed across the 20 replicates: average ISL (multiplied by
100) within each division, standard deviation of ISL (multiplied by 100) within
each division, and number of teams correctly classified in each division.
Figures A1-A6 (in the Appendix) show that for all three divisions and for
all schedules, relegating and promoting three teams results in the least variation in
ISLs within each division and also the largest number of teams that were correctly
classified. They also show that competitiveness increases as the number of teams
relegated and promoted increases from zero to three, and then decreases as the
number of teams relegated and promoted increases from three to five. This
suggests that it is optimal to relegate and promote three teams. However, the
competitiveness measures for two, three, and four teams were not distinguishable
statistically.
Further on the basis of within-division ISL standard deviation, Division 2
is more competitive than Divisions 1 and 3. This can be justified by noting in the
above histogram that Division 2 was initially the most homogeneous. But also,
Division 2 had the most rebalancing since it was the only division subject to both
relegation and promotion. It would be interesting to determine the true cause of
this effect by either adding more divisions or smoothing out the ISL means in the
ISDM model. Note that Division 2 also had the fewest teams correctly classified,
probably because it was the most homogenous division.

Effect of Changing the Variability of Year-to-year Skill Level Dynamics

Figures A7-A18 in the Appendix provide similar results to those in Figures A1-
A6 for the stable case in which the ISDM standard deviation has been halved and
the unstable case in which the ISDM standard deviation has been doubled. In the
stable case, it is optimal to relegate and promote two teams while in the unstable
case it is optimal to relegate and promote four teams for all schedules. Further, in
the stable case relegating between one and three teams results in similar levels of
competitiveness while in the unstable case relegating between three and five
teams provides similar levels of competitiveness. These results suggest that as the
year-to-year variability increases, it becomes optimal to relegate and promote
more teams.
Figures A19 and A20 in the Appendix show how variability impacts the
Division 1 competitiveness measures for the different RP rules. They show that
for each RP rule, as year-to-year variability increases, the number of teams
correctly classified decreases and the within-division standard deviation increases
suggesting that the divisions are becoming less competitive. They also show how
the optima vary with changing year-to-year variability.

Scheduling Effects

The scheduling system had little effect on competitiveness. This was expected
because all three schedule types required teams to play against the same
opposition.

Discussion

We believe that this paper provides an original and realistic model of a


hierarchical sports league as well as a framework for investigating how to
structure it so as to maximize competitiveness. Our results show that relegating
and promoting different numbers of teams will achieve different degrees of
competitiveness within divisions, and that for a given set of ISDM and MDM
forms such as the AR1 and logistic regression models used herein, or parameter
choices, one can determine the optimal number of teams to relegate and promote.
Further, our results show in the context of our models that as team year-to-year
intrinsic skill levels become more variable, the optimal number of teams to
relegate and promote increases. Finally, we found that these results were
insensitive to schedule structure. Of course, these results were based on an
idealized model of the NBA, but we believe our conclusions make sense and will
hold more broadly.
Relegation and promotion provides sports league management with a lever
to increase competitiveness and to sustain interest throughout the season for fans
of all teams. While it has been adopted widely outside of North America, we and
others believe it has the potential to improve North American sports. For example,
a recent article in the Wall Street Journal proposed a new realignment system for
NCAA Division 1-A football featuring relegation and promotion amidst teams
changing conferences and ongoing issues about the fairness of the Bowl
Championship Series.

“There's no governing authority that can make unilateral decisions


for all of college football—in fact, that's another part of the
problem. But what if there were? Is there a model anywhere in the
sports world for a system that could erase some of the chronic
problems with college football without killing the golden goose?
In a word, yes. It's the English Premier League.” (“Here’s How to
Fix College Football,” Kinkhabwala and Cacciola, Wall Street
Journal, June 15, 2010)

We note also that when relegation and promotion was featured in the 2010
American football video game Backbreaker, fans on discussion boards regarded it
a refreshing change to the existing NFL league structure.

Future Research

To better understand the effects that relegation and promotion have on


competition requires an approach for applying this model to data from sports
leagues that use these levers such as the English football league system. This
requires a model that quantifies the distribution of ISLs across teams in different
divisions. The lack of inter-divisional matches prevents us from using the same
technique as in our model based on NBA data. Also, in the context of the English
football league, we will need to generalize the MDM to account for the possibility
of ties.
Professional golf in North America has been using a dual league system
since 1990. Currently, the Nationwide Tour serves as a developmental league for
the PGA Tour. Rules of entry and exit from the PGA Tour are more complicated
than that of typical relegation and promotion leagues, but the concept remains.
The top 25 players on the Nationwide Tour every year are promoted into the PGA
Tour while all players on the PGA Tour who finish outside the top 125 are
relegated in some sense. This relegation and promotion system is further
augmented by the presence of the end-of-season “Q-School” tournament, in which
the top 25 players and ties receive PGA Tour status. We are currently
investigating the impact of these rules on competitiveness and player development
although we are facing the same problem alluded to above in that we need to
develop an approach to reconciling skill levels between the Nationwide and PGA
Tours since these players don’t often compete head to head. Preliminary analysis
suggests that players promoted from the Nationwide Tour to the PGA score about
1 stroke higher per round on average than they did on the Nationwide Tour in the
previous season. Further, we need to replace the MDM model by a tournament
outcome model. The approach of Connolly and Rendleman (2008) appears
promising.
Conclusion

With increased interest in relegation and promotion and declines in


competitiveness across North American professional sports leagues, it may be
sooner rather than later that one of the major North American sports leagues
considers this feature. If they were to do so, many operational, economic and
experiential factors will need to be considered. Such empirical data will be
difficult to find, and theoretical models may not be sufficient. We believe that the
models and computer simulation analyses proposed in this paper offer a
framework and approach for investigating such changes. Also, we hope that our
model will provide the basis for future formal mathematical analyses of sports
league structure.
Appendix

Figure A1

Figure A2
Figure A3

Figure A4
Figure A5

Figure A6
Figure A7

Figure A8
Figure A9

Figure A10
Figure A11

Figure A12
Figure A13

Figure A14
Figure A15

Figure A16
Figure A17

Figure A18
Figure A19

Figure A20
References

Buzzacchi, L., Szymanski S., and Zimbalist A. (2003). “Equality of Opportunity


and Equality of Outcome: Open Leagues, Closed Leagues and
Competitive Balance”, Journal of Industry, Competition and Trade, Vol.
3, No. 3, pp. 167-186.

Connolly, R.A. and Rendleman, Jr., R.J. (2008). “Skill, Luck and Streaky Play on
the PGA Tour”, Journal of the American Statistical Association 103, 74-
88.

Connolly, R.A. and Rendleman, Jr., R.J. (2010). “Going for the Green: A
Simulation Study of Qualifying Success Probabilities in Professional
Golf”, Unpublished manuscript, September 20, 2010.

Granger, C.W.J. and Newbold, P. (1986), Forecasting Economic Times Series,


Second Edition, Academic Press, San Diego, Ca.

Juhas, P., Cherba, D.M., Duxbury, P.M., Punch, W.F., and Billine, S.J.L. (2006).
“Ab Initio determination of solid state nanostructure”, Nature, 440, pp.
655-658 (30 March).

Medcalfe, Simon. (2003). “Does Promotion and Relegation Improve Competitive


Balance: Evidence from Two Natural Experiments from England,”
Unpublished paper, Lehigh University (May).

Noll, Roger G. (2002). “The Economics of Promotion and Relegation in Sports


Leagues: The Case of English Football,” Journal of Sports Economics,
Vol. 3, No. 2 (May), pp. 169-203.

Rehr, John J. (2006). “Materials Science: Nanostructures in a New League”,


Nature, 440 (30 March), pp. 618-619.

Szymanski, S. and Smith, R. (2002). "Equality of Opportunity and Equality of


Outcome: Static and Dynamic Competitive Balance in European and
North American Sports Leagues" in Transatlantic Sport. Carlos Barros,
Muradali Ibrahim and Stefan Szymanski, eds. Cheltenham, UK: Edward
Elgar.

Szymanski, S. and Valletti, T. (2005). "Promotion and Relegation in Sporting


Contests", Rivista di Politica Economica, 95, 5/6, pp. 3-39.
Zimbalist, Andrew S. (2002). “Competitive Balance in Sports Leagues: An
Introduction”, Journal of Sports Economics, Vol. 3, (May), pp. 111-121.

You might also like