Adjusting Winning-Percentage Standard Deviations and

a Measure of Competitive Balance for Home Advantage

Gregory A. Trandel
Department of Economics
University of Georgia
Joel Maxcy
Department of Kinesiology
University of Georgia
Athens, GA 30602
March 2009
Abstract: One measure of sports-league competitive balance uses a ratio: the standard
deviation of team winning percentages is divided by the so-called ideal standard deviation,
which assumes either team is equally likely to win any game. In fact, teams playing at
home win more than they lose. The extent of this advantage diers across sports. Ignoring
this pattern makes the traditional ideal standard deviation too large; the ratio of standard
deviations is therefore improperly small. The authors derive a standard-deviation formula
that accounts for home advantage, and consider how the adjustment aects comparisons of
competitive balance across leagues.
We thank, without implicating for any remaining aws, Craig Depken for comments on an
earlier version of this paper.
Introduction
The degree of competitive balance across the teams in a sports league can be an important
issue to both team and league executives and outside analysts. Both groups may, for example,
want to know how the degree of competitive balance among teams aects game attendance,
or how imbalance in a given league has changed (perhaps in response to changes in league
rules) over time. Analysts may also want to compare competitive balance across various
leagues.
Many methods of measuring competitive balance have been proposed; both Quirk and
Fort (1992) and Humphreys (2002) review many of these methods.
1
One common approach
uses a measure of the dispersion (typically, the standard deviation) of the end-of-season
winning percentages across the teams in a league.
One complication in using this method arises from the fact that the number of games
that make up a season diers across leagues. In North American professional leagues, for
example, Major League Baseball (MLB) teams currently play a 162-game schedule, while
National Football League (NFL) teams play 16 games. It would therefore be inappropriate
to directly compare winning-percentage standard deviations between such leagues the
shorter a leagues season, the greater the likelihood that random variation alone will produce
winning percentages that substantially dier from .500.
To address this concern, researchers can compare a leagues actual winning-percentage
standard deviation (typically averaged over a number of years) with the standard deviation
that would theoretically exist over the same number of games in (one particular version
of) a perfectly-balanced league. This approach rst appeared in the work of Noll (1988)
and Scully (1989), and was made popular by Quirk and Fort (1992). Using Forts (2007)
notation, a leagues actual standard deviation (ASD) of winning percentage is divided by
a (so-called) ideal standard deviation (ISD) to produce a ratio of standard deviations (i.e.,
RSD = ASD/ISD).
2
A larger value for RSD indicates a wider spread of outcomes, adjusted to
control for the dispersion that could be expected from random variation, and thus a greater
degree of competitive imbalance.
In the articles noted above, the ideal standard deviation is dened as that expected from
a league in which any two opposing teams have an equal probability of winning. With a
binomial outcome a team either wins or loses this assumption means that probability of
any team winning (or losing) any given game equals P(W) = P(L) = 1P(W) = 0.5. With
a season consisting of g games, the calculation of the standard theoretical ideal standard
deviation simplies to 0.5/

g. For example, the ISD in a 162-game MLB season equals

0.5/

162 = 0.0393, while in a 16-game NFL season it equals 0.5/

16 = 0.1250. Since the

National Basketball Association (NBA) and National Hockey League (NHL) both now have
82-game schedules, the standard ISD when those leagues play their full schedules equals
0.0552.
1
For other methods, see Depken (1999) and Mizak, Neral, and Stair (2007), as well as the discussion and
citations in Sanderson and Siegfried (2003).
2
In this usage, the ideal standard deviation is dened as one that arises from a league in which all
teams have equal skill levels. It is unclear whether such a characteristic would in fact be an ideal outcome
from, for example, the perspective of league-wide prot maximization.
1
Normalizing standard deviation by using RSD allows competitive balance to be compared
across leagues with diering season lengths. Doing so also makes it possible for competitive-
balance comparisons to account for changes in season length. For instance, MLB increased
(in 1961) its number of games from 154 games to 162, while the NFL has increased its
season from 12 to 14 (in 1961) and then to 16 contests (in 1977). Calculations using RSD
also facilitate comparisons of a leagues winning-percentage dispersion during periods in
which some season was shortened due to a strike or lockout (e.g. MLB in 1994 and 1995 and
the NBA in 1998-99).
In this paper, we argue that existing calculations of ISD and therefore of RSD are
awed because they fail to account for the fact that even two perfectly-balanced athletic
teams are not equally likely to win any particular contest. Rather, the results of nearly all
sporting events show that teams (or, in many cases, individuals) playing in the location to
which they are accustomed (i.e., playing at home) are more likely to win than are teams
playing on the road (see, for example, the results summarized in Courneya and Carron
(1992).)
Since actual sporting results arise from a world in which home advantage exists, it is
improper to compare empirical outcomes to a baseline in which that advantage is ignored. As
is shown below, neglecting home advantage means that traditional ISDs are inappropriately
large, which implies that calculated RSDs are improperly small. The size of this bias increases
with the extent of a leagues home advantage. Since home advantage diers by sport, the
errors also dier a league in which teams have a relatively large home advantage will
appear, in standard RSD calculations, to have a relatively greater degree of competitive
balance than is really the case.
This paper proceeds by rst briey describing the sizable literature that considers home
advantage. It next details how the standard deviation calculation can be modied to account
for the presence of home advantage. Finally, the paper shows the extent to which the
correction aects computed measures of competitive balance.
Home Advantage
Courneya and Carron (1992) dene home advantage as the consistent nding that sports
teams playing in their regular surroundings (and assuming that each teams schedule is
balanced between home and road games) win more games than they lose. This pattern
implies that a game played between two evenly-matched teams in a location in which only
one is playing at home is not a 50/50 proposition. Rather (assuming that a team either
wins or loses), the home teams probability of winning h exceeds .5, while the visiting teams
probability of winning 1 h < .5.
Following the seminal formal work on the subject by Schwartz and Barsky (1977), the
concept and existence of home advantage in sports contests has been well documented in the
sport-science and sport-psychology literatures.
Courneya and Carron (1992) synthesize the ndings on the major team sports and report
historical home winning percentages of 53.5% for baseball, 57.3% for football, 61.1% for
ice hockey, 64.4% for basketball, and 69.9% for soccer. Leifer (1995) calculates the home
advantage for each of the four major North American professional team-sport leagues from
their inceptions to 1990 and reports similar results. Data from Pollard and Pollard (2005)
2
who consider trends in home advantage in the four major North American team sports and
English soccer from 1876 to 2003 can be used to compute the following home advantages
over (for example) the period 1990-2002: 53.5% in MLB, 55.7% in the NHL, 58.7% in the
NFL, 61.0% in the NBA, and 60.8% in the (English) Premier League.
3
The above authors, and many others, have also sought to determine why home advantage
exists and why it varies systematically across sports. Courneya and Carron (1992) and Leifer
(1995) provide thorough reviews of this literature, with an update by Carron et al (2005).
4
Courneya and Carron (1992) organize their discussion of home advantage around what
they call game-location factors and psychological or behavioral states. Location factors that
may aect home and road teams dierently include the eects on participants of crowds, fa-
miliarity with the playing eld and surroundings, travel, and rules that explicitly dierentiate
between home and road teams.
5
These four factors can in turn inuence the psychological
or behavioral states of the participating athletes, coaches, and/or ocials.
It is not surprising that the degree of home advantage varies across sports. Crowds sit
closer to players and ocials in arena sports such as basketball and ice hockey than they
do in (primarily) outdoor sports. Football teams travel to a game on average once every
other week, while basketball or hockey players may play three or four road games in a week.
Playing surfaces and typical weather conditions dier across football and baseball stadiums,
while baseball stadiums also have diering playing-eld dimensions.
Table 1 shows the home advantage measured as the league-wide home winning percent-
age for each of the four major North American professional sports leagues from 19902007.
Over all games in these seasons, the home winning percentages are MLB 53.7%, NHL
55.6%, NFL 58.4%, and NBA 60.8%.
6
Note that in some sports leagues, games can end in a tie (also known as a draw). Such
leagues often rank teams on a point system, in which a teams reward for tying a game is
3
The ve most signicant European soccer leagues appear to have relatively similar degrees of home
advantage. Pollard (2006) shows that the home advantage in the highest league in the major European soccer
nations (over the 1996/972001/02 seasons) are 60.7% (England), 63.3% (Germany), 63.9% (Spain), 64.2%
(Italy), and 65.0% (France). Less prestigious soccer leagues, however, can display much wider variation.
Among European soccer leagues, Pollard (2006) notes home advantages that range (over a six-season period)
from 53% (Estonia and Latvia) to 7779% (Bosnia and Albania).
4
While most studies of home advantage focus on game results, Jones (2007) considers how home advantage
evolves over the course of NBA games. He shows that home teams do particularly well in the rst quarter,
and that a home team that enters a quarter with the lead does not on average extend its lead, but that a
home team that enters a quarter trailing on average gains.
5
In addition to the rigors of travel itself, Entine and Small (2008) consider the fact that NBA teams
playing on the road typically have fewer days of rest between games than do teams in the midst of a home
stand. On the issue of game rules, Turocy (2008) who models baseball as a zero-sum Markov game
investigates the requirement that road teams bat rst. He concludes (given his modeling assumptions) that
there is no signicant quantitative advantage conferred by the order in which teams bat. Note though that
Turocys simulations show that the introduction of one particular strategy the sacrice bunt, a strategy
that can be modeled in a straightforward way does give the team batting last a slight advantage.
6
These overall percentages are not the averages of the yearly home winning percentages; rather, they are
the percentages found by totalling all wins, losses, and ties in the league over the eightteen seasons.
3
one-half or one-third as large as the reward for winning. All the results presented in this
paper treat ties as half-wins.
7
Over this period, home teams record more wins than do road teams in each year in each
league. The home advantages are remarkably consistent the standard deviation of the
yearly home winning percentages over the period is less than 0.01 for one league, less than
.02 for two others, and under .03 for the fourth.
In spite of the many investigations into home advantage, few researchers have considered
any link between home advantage and issues of competitive balance. Koning (2000) is one
who has. His approach which incorporates the trinomial outcomes possible in soccer
uses an adjustment for home advantage when employing an ordered probit specication to
estimate a strength parameter for each team in Dutch professional soccer. In one of his
measures of competitive balance, Koning measures the standard deviation of these strength
parameters. However, when Koning directly calculates the standard deviation of season-end
team point totals, he does not appear to make any direct adjustment for home advantage.
Forrest, et al (2005) take a dierent approach by describing a process by which an increase
in what would commonly be called competitive balance i.e., a reduction in the range of
skill dierences across a leagues teams could possibly decrease fan interest in a sport.
The rationale is that if teams become more closely balanced in skill levels, the expected
winner of some games would be more likely to be determined by home advantage. For some
fraction of a leagues contests, this would reduce outcome uncertainty, which could cause
aggregate attendance to fall.
The present paper links the issues of home advantage and competitive balance in a
very dierent way. It argues that the traditional standard-deviation measure of competitive
balance, which ignores home advantage, is awed. The following section describes a way to
modify the measure of ideal standard deviation to account for home advantage, and explains
how this adjustment aects judgments about relative competitive balance.
Home-Advantage-Corrected Ideal Standard Deviation
As noted above, many previous researchers have used as a benchmark the so-called ideal
standard deviation of the winning percentage that would be observed in a league in which
all games are equally likely to be won by either team.
7
Note that rule changes in the NHL have caused some adjustments in the calculation of game results. For
much of its history, NHL teams played a sixty-minute game, and the league awarded two points to winning
teams, one point to each team for a draw, and no points to a losing team. Starting with the 19831984
season, however, NHL teams that are tied at the end of regulation time play a ve-minute overtime period.
Points were originally still awarded on 2W-1T-0L basis. [Most world soccer leagues instead use a 3W-1D-0L
system.] Since the start of the 19992000 season, however, a team that loses either in overtime or (starting
with the 200506 season) in a shootout is awarded one point. Teams that win under any circumstances
continue to receive two points. In some games, therefore, the teams share a total of three points rather
than two. Through the 20032004 season, games could still end in ties. After a labor dispute cancelled the
20042005 season, however, league rules were altered by the addition of a shootout (which takes place after
the overtime period); as a result, every game now has a winner. To maintain data consistency throughout
the relevant time period, and so that all NHL games are valued equally, the empirical work presented in this
paper in based on NHL results in which (since the 19992000 season) the point awarded for an overtime or
shootout loss has been removed. In the data used in this paper, in other words, a team either wins, ties
(prior to 20052006), or loses; a tie is counted as half a win, and any kind of loss is counted as having no
value.
4
The evidence cited above, however, implies that a perfectly-balanced league is more
correctly viewed as one in which (assuming no ties) every team has a probability h > .5 of
winning a game played at home and a probability (1 h) < .5 of winning an game on the
road.
Since a sport leagues actual results occur in a world in which home teams have an ad-
vantage, comparing those results to a baseline which assumes no such advantage is improper.
Note that some authors have proposed a modication of the ISD method for sports where
trinomial outcomes win, loss, and draw are common. In the major North American
sports, the binomial model is proper for studies of MLB, the NBA, and the NFL as league
rules require that ties be settled through extended play.
8
In the NHL (through the 2003-2004
season) and soccer, however, matches frequently end in a draw. Cain and Haddock (2006)
argue that that the various point systems used by leagues in which draws can occur produce
diering ideal standard deviations. They suggest ways to revise ISD for soccer leagues given
dierent point assignment schemes. Fort (2007) critiques their conclusions and counters that
when point totals are converted to winning percentages, with a tie dened as half a win,
their method oers no innovation. The calculations reported in the present paper treat ties
as half-wins.
Allowing h to be greater than .5 alters the calculated value of the standard deviation
of winning percentage in a perfectly-balanced league. In particular, the home-advantage-
corrected ideal standard deviation (HISD) will be smaller than is the traditionally-calculated
ideal standard deviation. Intuition for this pattern can be developed using an extreme
example. Consider a case in which every one of a leagues games is won by the home team
with probability .95. The expected standard deviation of season-long winning percentage
in such a league would be very small since the winning percentage of every team would be
extremely likely to be quite close to .500.
The bias in the traditional ideal standard deviation means that ignoring home advantage
produces a value for a leagues ratio of standard deviations that is improperly small. The
larger a leagues home advantage, the greater the extent of this bias. Analysts who have
ignored this eect have therefore concluded that leagues have a greater degree of competitive
balance than is truly justied. Because the extent of home advantage diers across leagues,
the bias will also aect any cross-league comparison of competitive balance.
Fortunately, the standard deviation of an equally-balanced league in which a home advan-
tage exists can be computed with a relatively straightforward, albeit somewhat complicated,
formula. Suppose that the teams in a sports league play n (an even number of) games,
playing n/2 games at home and the rest on the road. Furthermore, assume that teams in
the league have equally-balanced talent levels, but that the probability of the home team
winning any particular game equals h > .5.
Computation of such a leagues expected end-of-season standard deviation proceeds by
considering the probability of a team achieving any particular season-long record. For exam-
ple, a team can win all of its games only by winning every game at home (probability h
(n/2)
)
8
The NFL instituted an overtime policy in 1974. Games are extended by an additional fteen minute
period. The sudden death method determines that the team scoring rst the winner. However, if neither
team scores in the overtime the game is recorded as a tie. In practice, tie game are extremely rare under this
system. Between the institution of the overtime rule and 2008, only 13 games were tied after the overtime
period (including only four from 1990-2008).
5
and every game on the road (probability (1h)
(n/2)
). Theres only one way of winning every
game. Since this performance results in a winning percentage of 1 (which is compared to the
league-average winning percentage, which must equal .5), computing the expected standard
deviation of winning percentage requires that the above term be multiplied by (1 .5)
2
.
A team has two ways to win all but one of its games. The rst method is to win all but
one of its home games (probability h
((n/2)1)
(1 h) there are (n/2) dierent ways to do
this and to win all of its road games ((1 h)
(n/2)
). The second method is to sweep at
home (h
(n/2)
) and win all but one on the road ((n/2) (1h)
((n/2)1)
h). The total probability
that the team will win all but one of its games is thus (n/2)h
((n/2)1)
(1 h) (1 h)
(n/2)
+
h
(n/2)
(n/2)(1 h)
((n/2)1)
h. This term is then multiplied by (((n 1)/n) .5)
2
.
A team has three ways to win all but two of its games. It could win all but two games at
home (probability h
((n/2)2)
(1h)
2
; there are
(n/2)!
((n/2)2)!2!
ways to do this (where x! represents
x factorial (or 1 2 . . . x 1 x)). The team could win all but one at home and all
but one on the road; the probability is (n/2)h
((n/2)1)
(1 h) (n/2)(1 h)
((n/2)1)
h. Or,
the team could win every game at home and all but two on the road; the probability is
h
(n/2)

(n/2)!
((n/2)2)!2!
(1 h)
((n/2)2)
h
2
. In the standard-deviation calculation, these three terms
are added and multiplied by (((n 2)/n) .5)
2
.
Note that in the league being considered, the probability of a team winning no games
equals the probability that it wins every game, the probability of winning one game equals
the probability of winning all but one, etc.
Considering all possible outcomes results in the following formulae for the home-advantage-
corrected standard deviation (variance) of an evenly-balanced league:

2
= 2
n/2

k=1
__
k

i=1
_
n/2
k i
_
h
((n/2)k+i)
(1 h)
(ki)

_
n/2
i 1
_
h
(i1)
(1 h)
((n/2)+1i)
__
n + 1 k
n
.5
_
2
_
_

2
= 2
n/2

k=1
__
k

i=1
(n/2)!
((n/2) k +i)! (k i)!
h
((n/2)k+i)
(1 h)
(ki)

(n/2)!
((n/2) + 1 i)! (i 1)!
h
(i1)
(1 h)
((n/2)+1i)
__
.5
k 1
n
_
2
_
_
=

_
2
n/2

k=1
_

_
k

i=1
((n/2)!)
2
h
((n/2)k+2i1)
(1 h)
((n/2)+1+k2i)
_
.5
k1
n
_
2
((n/2) k +i)! (k i)! ((n/2) + 1 i)! (i 1)!
_

_
.
6
In the rst formula, the notation for a binomial coecient, such as
_
n/2
k i
_
, represents
(n/2)!
((n/2)k+1)!(ki)!
, which is the number of ways of picking ki (unordered) outcomes from n/2
possibilities.
The above formulae can be easily computed using a mathematical computer program such
as (the one used by the authors) Mathematica. Table 2 shows the home-advantage-corrected
ideal standard deviations for various combinations of h and n. The table also reports an
approximate value for the average percentage bias created by implicitly assuming away home
advantage. That bias depends, obviously, on the extent of the advantage. Calculations using
the numbers in the table reveal, however, that the bias is relatively unaected by the number
of games in a season.
Home-Advantage-Corrected Ratio of Standard Deviations
In this section, 19902007 results from the four major North American sports leagues
are used to produce home-advantage corrected versions of both the ideal standard deviation
and the ratio of standard deviations. In so doing, the league average historical home-team
winning percentages over those years are used as a measure of the theoretical parameter h.
9
Remember from above that those empirical home winning percentages are: MLB 53.7%,
NHL 55.6%, NFL 58.4%, and NBA 60.8%. Assuming for now that each league
plays its current standard number of games, the home-advantage corrected ideal standard
deviations (HISD) based on these empirical results are: NFL .123221; NBA .053920;
NHL .054872; MLB .039174.
Using the seasons in the 1990-2007 period, the average standard deviations of the season-
end standings in the major North American sports have been: NFL .19075; NBA
.15856; NHL .10187; MLB .06996; As noted above, the traditionally measured ideal
standard deviations for these leagues (assuming each league plays its now standard number
of games) are: NFL .12500; NBA and NHL .055216; MLB .039284.
Dividing ASD by ISD for each year, incorporating the fact that leagues sometimes do
not play their standard number of games, and then averaging the results over the period,
produces the following values for the traditionally-measured ratio of standard deviations:
NFL 1.5260; NBA 2.8365; MLB 1.7597; NHL 1.8177.
10
Dividing yearly ASD by the HISD given above, incorporating shortened seasons, and
averaging produces the following values for the home-advantage-corrected ratio of standard
deviations (HRSD): NFL 1.5485; NBA 2.9046; MLB 1.7646; NHL 1.8291.
The adjustment for home advantage has a greater impact for leagues whose teams have
a larger home advantage. As a result, the NBA, compared to the other three leagues, looks
9
Note, though, that the empirical winning percentages dont perfectly capture the theoretical h, which
is dened for a league of perfectly-balanced teams. The use of home-team winning percentage to measure
home advantage is common, but there are other ways to measure the size of the advantage. Fair and Oster
(2007), to cite one example, use the results of college football games to estimate that playing at home is, on
average, worth the equivalent of 4.3 points.
10
Because of labor disputes, NHL teams played a 48-game season in 199495, NBA teams played 50 games
in 199899, and MLB teams played a 144-game schedule in 1995, and averaged 114 games in 1994. In
addition, the NHL regular season twice consisted of 80 games and twice of 84 games over the four seasons
spanning 199095. Finally, the NHL did not play at all during what would have been its 200405 season.
7
even more competitively imbalanced. In addition, since the NFL (for example) displays a
stronger home advantage than does MLB, the dierence in the ratio of standard deviations
which captures the measured degree of competitive imbalance between MLB and the
NFL falls by between 89%.
Conclusion
One method that researchers have used to measure the competitive balance of a sports
league involves dividing the standard deviation of team winning percentages by the standard
deviation that would be expected if each game was equally likely to be won by either team.
Dividing by the so-called ideal standard deviation adjusts for dierences in season length
across leagues.
In no major team sport, however, are games between equally-skilled teams actually 50/50
propositions. Rather, teams playing in their regular locations win (on average) more than
fty percent of games. The extent of this home advantage diers across sports.
Ignoring home advantage means that the typical calculation of the ideal standard devi-
ation is biased: it is larger than it would be if the advantage was taken into account. As a
result, the ratio of standard deviations is inappropriately small. In other words, the typical
procedure overestimates a leagues degree of competitive balance. Because a larger home
advantage produces a greater bias, comparisons of competitive balance across leagues are
also aected.
This paper describes a formula that can be used to calculate a home-advantage-corrected
ideal standard deviation, and therefore a corrected measure of competitive balance.
Since the NBA has the largest home advantage among the major North American sports
leagues, correcting for home advantage increases the extent to which the NBA is shown to
be competitively imbalanced. The home-advantage adjustment also modestly narrows the
gap in measured competitive imbalance between MLB and the NFL.
8
Table 1: Percentage of Games Won By Home Team
Year MLB NFL NHL NBA
1990 .537 .585 .589 .659
1991 .538 .589 .605 .631
1992 .552 .607 .557 .611
1993 .538 .549 .540 .612
1994 .517 .571 .583 .597
1995 .532 .600 .561 .604
1996 .541 .621 .550 .575
1997 .535 .608 .536 .595
1998 .538 .629 .543 .623
1999 .521 .597 .551 .611
2000 .540 .556 .551 .598
2001 .524 .548 .553 .591
2002 .542 .580 .549 .628
2003 .550 .613 .549 .614
2004 .535 .566 n/a .605
2005 .537 .590 .574 .603
2006 .546 .531 .550 .591
2007 .543 .574 .537 .601
Overall .537 .584 .556 .608
Stnd Dev .009 .027 .019 .019
Table 2: Home-Advantage-Corrected Ideal Standard Deviation of Winning Percentage
number of home-team winning probability (h)
games (n) .50 .55 .60 .65
10 .158114 .157321 .154919 .150831
20 .111803 .111243 .109545 .106654
40 .0790569 .0786607 .0774597 .0754155
80 .0559017 .0556215 .0547723 .0533268
160 .0395285 .0393303 .0387298 .0377078
average
bias from
using h = .5 0.5% 2.1% 4.8%
9
References
Cain, Louis P., and David D. Haddock (2006). Measuring parity: Tying into the idealized
standard deviation. Journal of Sports Economics 7, 330338.
Carron, Albert V., Todd M. Loughhead, and Steven R. Bray (2005). The home advantage
in sport competitions: Courneya and Carrons (1992) conceptual framework a decade
later. Journal of Sports Sciences 23(4), 395407.
Courneya, Kerry S. and Albert V. Carron (1992). The home advantage in sport competi-
tions: A literature review. Journal of Sport and Exercise Psychology 14, 2839.
Depken, Craig (1999). Free Agency and the Competitiveness of Major League Baseball.
The Review of Industrial Organization 14(3), 205217.
Entine, Oliver A. and Dylan S. Small (2008). The Role of Rest in the NBA Home-Court
Advantage. Journal of Quantitative Analysis in Sports 4(2), Article 6. Available at:
http://www.bepress.com/jqas/vol4/iss2/6
Fair, Ray C. and John F. Oster (2007). College Football Rankings and Market Eciency.
Journal of Sports Economics 8(1), 318.
Forrest, David, James Beaumont, John Goddard, and Robert Simmons (2005). Home
advantage and the debate about competitive balance in professional sports leagues.
Journal of Sports Sciences 23(4), 439445.
Fort, Rodney (2007). Comments on measuring parity. Journal of Sport Economics 8(6),
642651.
Humphreys, Brad R. (2002). Alternative measures of competitive balance in sports leagues.
Journal of Sport Economics 3(2), 133148.
Jones, Marshall B. (2007). Home Advantage in the NBA as a Game-Long Process. Journal
of Quantitative Analysis in Sports 3(4), article 2. Available at: http://www.bepress.
com/jqas/vol3/iss4/2
Koning, R. H. (2000). Balance in competition in Dutch soccer. The Statistician 49(3),
419-431.
Leifer, Eric M. (1995). Making the Majors. Cambridge, MA: Harvard University Press.
Mizak, Daniel, John Neral, and Anthony Stair (2007). The adjusted churn: an index of
competitive balance for sports leagues based on changes in team standings over time.
Economics Bulletin 26(3), 17.
Noll, R. G. (1988). Professional basketball (Studies in Industrial Economics Paper No.
144). Stanford University, Stanford, California.
Pollard, Richard (2006). Home advantage in soccer: variations in its magnitude and a
literature review of the inter-related factors associated with its existence. Journal of
Sport Behavior 29(2), 169189.
Pollard, R. and Pollard, G. (2005). Long-term trends in home advantage in professional
team sports in North America and England (18762033) Journal of Sports Sciences
23(4), 337350.
Quirk, James and Rodney D. Fort (1992). Pay dirt: The business of professional team
sports. Princeton, NJ: Princeton University Press.
Snderson, Allen R., and John J. Siegfried (2003). Thinking about competitive balance.
Journal of Sport Economics 4(4), 255279.
Schwartz, Barry and Stephen F. Barsky (1977). The home advantage. Social Forces 55(3),
641661.
10
Scully, Gerald W. (1989). The Business of Major League Baseball. Chicago: University of
Chicago Press.
Turocy, Theodore L. (2008). In Search of the Last-Ups Advantage in Baseball: A Game-
Theoretic Approach. Journal of Quantitative Analysis in Sports 4(2), article 5. Avail-
able at: http://www.bepress.com/jqas/vol4/iss2/5.
11