Profiting from arbitrage and odds biases of the European football gambling market

Abstract

A gambling market is usually described as being inefficient if there are one or more betting strategies that generate profit, at a consistent rate, as a consequence of exploiting market flaws. This paper examines the online European football gambling market based on 14 European football leagues over a period of seven years, from season 2005/06 to 2011/12 inclusive, and takes into consideration the odds provided by numerous bookmaking firms. Contrary to common misconceptions, we demonstrate that the accuracy of bookmakers' odds has not improved over this period. More importantly, our results question market efficiency by demonstrating high profitability on the basis of consistent odds biases and numerous arbitrage opportunities.

Full text

- 1 -
Corrected version (with corrections highlighted), version 2013.10.16.
To appear in the Journal of Gambling Business and Economics.
Profiting from arbitrage and odds biases of the European football gambling market
ANTHONY C. CONSTANTINOU1 AND NORMAN E. FENTON
Risk and Information Management (RIM) Research Group,
Department of Electronic Engineering and Computer Science,
Queen Mary, University of London, UK, E1 4NS
THIS IS A PRE-PUBLICATION DRAFT OF THE FOLLOWING CITATION:
Constantinou, A. C. & Fenton, Norman E. (2013). Profiting from arbitrage and odds biases of the European football
gambling market. The Journal of Gambling Business and Economics, Vol. 7, 2: 41-70.
DOI 10.5750/jgbe.v7i2.630:
CONTACT: Dr. Anthony Constantinou, anthony@constantinou.info, a.constantinou@qmul.ac.uk
ABSTRACT: A gambling market is usually described as being inefficient if there are one or more betting strategies that
generate profit, at a consistent rate, as a consequence of exploiting market flaws. This paper examines the online European
football gambling market based on 14 European football leagues over a period of seven years, from season 2005/06 to
2011/12 inclusive, and takes into consideration the odds provided by numerous bookmaking firms. Contrary to common
misconceptions, we demonstrate that the accuracy of bookmakers' odds has not improved over this period. More importantly,
our results question market efficiency by demonstrating high profitability on the basis of consistent odds biases and
numerous arbitrage opportunities.
Keywords: betting market, favourite-longshot bias, football betting, profit margin, soccer betting, sports betting, sports
gambling
1 Introduction
The enormous popularity of Association Football (hereafter referred to as simply football), along with
increasing interest in gambling (particularly after its introduction online) means that great attention is
now paid to its betting odds. While betting interest in horse racing has decreased, betting on football
has increased so that is now by far the biggest sport in terms of turnover through the online
bookmakers (Finnigan & Nordsted, 2010). For example, in 1998 the turnover for British football
betting was close to £2bn (Global Betting and Gaming Consultants, 2001), and even then football
betting was described as the fastest growing sector in British gambling (Mintel Intelligence Report,
2001). In comparison, the turnover reported by just a single bookmaker (bwin Group, 2009) for 2008
was around £2.92bn, which represented an astonishing 31.4% increase from the previous year.
For any large scale gambling market (and this includes financial markets) the question of
efficiency is paramount. If there is a betting strategy that is consistent in generating profit against a
gambling market, then such a market is normally described as being inefficient. The possibility of
profiting because of market flaws is clearly both important and exciting. Because of the explosion of
interest in football betting, increasingly researchers have turned their attention to evaluating the
efficiency of this particular betting market. (Sauer, 1998) suggested that if a betting market is to be
considered efficient, then such a market should encompass all the relevant information available in
order to eliminate bettors exploiting opportunities within the market and achieving profits. Intuitively,
many assume that the market is efficient. Yet, indications of an inefficient football betting market are
becoming increasingly popular within the academic literature.
(Dixon & Coles, 1997) concluded that the UK football betting market is inefficient after a
rather simple bivariate Poisson distribution model was able to earn positive returns under specific
high-discrepancy trading rules during the English Premier League (EPL) season 1995/96. Similar
1 Corresponding author. E-mail addresses: anthony@constantinou.info (Constantinou, A. C.), norman@eecs.qmul.ac.uk
(Fenton, N. E.)

- 2 -
conclusions have been reported in (Rue & Salvesen, 2000; Kuypers, 2000; Dixon & Pope, 2004).
Further, in 2004 (Goddard & Asimakopoulos, 2004) found that the betting market is inefficient at the
start2 and, most notably, at the end of a football season, whereas (Forrest & Simmons, 2008)
concluded that published odds appear to be influenced by the number of fans of each club in a match
after observing that popular teams are offered more favourable terms on their wagers. More recently,
(Constantinou et al., 2012) demonstrated how a Bayesian network model was able to generate
profitability against the various published market odds by incorporating subjective information along
with relevant historical data, whereas in (Constantinou et al., 2013) an improved Bayesian network
model questioned market efficiency by demonstrating even greater profitability. In (Constantinou &
Fenton, 2013) a novel and simple rating approach, called the pi-rating, provided further evidence of
market inefficiency by demonstrating profitability over a period of five English Premier League
(EPL) seasons. Yet, perhaps the primary reason why the football betting market is considered by
many to be inefficient is the strong evidence of a favourite-longshot bias (see Section 4) as reported in
(Cain et al., 2000, Forrest & Simmons, 2001; 2002).
In contrast to the studies above, other researchers came to opposite conclusions. In 1989
(Pope & Peel, 1989) investigated the ex post inefficiency of the fixed odds provided between
bookmaking firms and concluded that no profitable betting strategies could have been implemented ex
ante at that time. (Forrest et al., 2005) demonstrated how the efficiency of the market increased over a
five year period with the help of an ordered profit model and showed that their model was unable to
make profitable returns against the bookmakers. More recently, (Graham & Stott, 2008) introduced
two forecast models; one based on football results, which is similar to that of (Forrest et al., 2005),
and another based on past bookmaking odds in an attempt to compare the bookmaking opinion of
various UK teams with the ratings generated by the football results based model. They showed that
bookmaking prices were rational and not significantly different than those generated by the model,
even though in some cases systematic bookmaking odds biases were observed which could not have
been explained. Possibly the strongest evidence of efficiency are reported in studies in which
researchers have attempted to outperform bookmakers' odds by introducing their own forecast models
(ranging from very simple to rather sophisticated models), but failed to do so. As a result, other
relevant studies have concluded and/or assumed that the betting market is efficient (Peel & Thomas,
1988; 1992; 1997; Vecer et al., 2009).
While this paper is focused on fixed-odds football betting markets, it is worth noting that
there are various other studies within the academic literature which focus on sport betting markets that
encompass significant differences in betting behaviour3. Discussions regarding such distinct betting
markets can be found in (Vergin & Scriabin, 1978; Hausch et al., 1981; Asch et al., 1984; Zuber et al.,
1985; Sauer et al., 1988; Thaler & Ziemba, 1988; Golec & Tamarkin, 1991; Shin H., 1991; Shin R. E.,
1992; Shin H., 1993; Woodland & Woodland, 1994; Peel & Thomas, 1997; Vaughan Williams &
Paton, 1997; Golec & Tamarkin, 1998; Henery, 1999; Jullien & Salanie, 2000; Woodland &
Woodland, 2001; Levitt, 2004; Paton & Vaughan Williams, 2005).
In this paper we evaluate the modern behaviour of the online football gambling market on the
basis of the odds provided by numerous bookmaking firms for 14 European football leagues and over
a period of seven football seasons (i.e. from season 2005/06 to season 2011/12 inclusive). The data
used for this study is available at www.football-data.co.uk. Throughout this paper (to be consistent
with other studies) we restrict our attention to just the three possible outcomes of any match: H (home
win), D (Draw) and A (Away win). The paper is structured as follows: Section 2 analyses profit
margin variability and resulting arbitrage opportunities, Section 3 assesses bookmakers' accuracy,
Section 4 examines predetermined biases in published odds, and we provide our concluding remarks
and implications in Section 5.
2 This agrees with (Forrest et al., 2005), in which authors showed that over a five-year period, their benchmark statistical
model was outperforming bookmaking odds at the very start of the season. However, in all cases the model eventually failed
to outperform bookmakers' odds. No claims were made of an inefficient market.
3 Notably, other markets include pari-mutuel betting where published odds are determined solely by the behaviour of the
bettors (e.g. horse racing), spread betting where the returns are based on the accuracy of the bettor (e.g. NFL); betting
exchange where one bettor can bet against another bettor (e.g. horse racing – this has also recently emerged in UK football
betting (betfair, 2000)).

- 3 -
2 PROFIT MARGIN & ARBITRAGE
The bookmakers' profit margin, also known as 'over-round', refers to the margin by which the sum of
the probability odds of the total outcomes exceeds 1 (thus, making the odds unfair for the bettor). A
lower profit margin results in less-unfair published odds. In short, the profit margin indicates the
precise profit a bookmaker expects to receive if bets are distributed such that the bookmaker pays
identical amounts of winnings whatever the match outcome. Since it is highly unlikely that the bets
are distributed as specified above, the profit margin is just an approximation of a bookmaker's
expected profit.
On the other hand an arbitrage opportunity is simply an opportunity whereby profit is
guaranteed on the basis of a negative profit margin which results by combining the odds published by
the various bookmaking firms. In particular, arbitrage opportunities depend on two factors: a) the
divergence in outcome probabilities between bookmaking firms and b) the profit margin by each
bookmaker. Negative profit margin is simply a scenario where a set of  probabilities is found (for
a single match instance) in which the sum of the probabilities within that set is . Hence, profit for
the bettor can be guaranteed if the bets are placed such that the return is identical whatever the
outcome.
For example, if we find that the best (lowest) probabilities for the bettor for a specific match
instance, over a number of bookmaking firms, are ,  and , the
sum of probabilities is just ; corresponding to the respective decimal odds of 2.222, 3.4482
and 4. For this scenario the arbitrage size is 1% and we can guarantee a profit of 1.0101% 
. If we
want to bet , then the bet has to be distributed on the three outcomes as follows (rounded to
the nearest penny):
 £45.45 on outcome 
 £29.29 on outcome 
 £25.25 on outcome 
using the equation 

 for each case of , where  represents the odds of  while calculating the
bet to be placed on outcome  and so forth.
Academic evidence that demonstrate arbitrage opportunities date back to the 1980s, where
(Pope & Peel, 1989) reported many such cases by considering the odds offered by four bookmakers,
on a pre-tax basis, from 1980 to 1982. However, more recent studies (Dixon & Pope, 2004; Forrest et
al., 2005) found no such opportunities in modern betting and concluded that there have been far less
divergences in odds in recent years than in earlier periods. In particular, (Forrest et al., 2005)
performed similar tests for the EPL seasons 1998 to 2003 using information from five bookmakers
with introduced profit margins within the range of 10% and 12%. They showed that the minimum
possible profit margin attained over the five seasons was averaging close to 6.6% and as expected,
they reported no cases of positive arbitrage returns during that period.
2.1. Results
For determining the profit margin we have considered the average odds provided by the different
bookmaking firms, whereas for determining arbitrage opportunities we have considered the maximum
odds (best available for the bettor). The number of bookmaking firms taken into consideration for
computing averages and maximums varies between 25 and 60 for each match instance (for details
refer to Table A.1).
Table 1 presents the average profit margin introduced, from average bookmaking odds, for
the specified football league and season. The results are not surprising since they clearly indicate that
the profit margin has decreased over time and this outcome is in agreement with (Hvattum & Arntzen,
2010) who concluded that the competitiveness of the football betting market has increased during the
period 2000 to 2008 on the basis of similar market behaviour.

- 4 -
It is interesting that the reduction in profit margin for the most recent seasons (i.e. 2009/10
onwards) is much stronger. Table 1 also demonstrates how lower divisions suffer from increased
profit margins relative to the top divisions. It is also interesting to note that the odds provided for
match instances of the English League II (i.e. 4th division in England) suffer from a lower profit
margin than those published for top divisions in Holland, Belgium and Greece. (Rue & Salvesen,
2000) suggested that it is natural for the bookmakers to provide better odds for the Premier League
than for the lower divisions as the majority of the bettors bet on the Premier League; if their
hypothesis is correct then we should conclude that more bettors bet on the 4th division in England
rather than on top divisions in Holland, Belgium and Greece. But with no further evidence it is very
difficult to accept such a conclusion and hence, even though we believe that the market-forces
assumption of (Rue & Salvesen, 2000) is rational, it is possible that other important factors that would
allow us to formulate such a conclusion are not available for assessment. For example we do not have
sufficient information whether the following kind of scenarios occur: if 100 bookmakers publish odds
for league A and 500 bookmakers publish odds for league B, then it may be possible that the
published odds of league B suffer from a lower profit margin than those of A due to the higher
competitiveness between bookmaking firms for league B.
Table 1. Average profit margin introduced, from average bookmaking odds,
for the specified league and season.
League
Season
2005-06
2006-07
2007-09
2008-09
2009-10
2010-11
2011-12
Overall
Eng. Premier Leag.
9.45%
8.98%
8.29%
7.26%
6.67%
6.09%
6.08%
7.55%
Eng. Champ.
10.97%
10.42%
9.69%
8.51%
8.13%
7.45%
7.38%
8.94%
Eng. League I
11.17%
10.85%
10.30%
9.08%
8.63%
7.89%
8.07%
9.43%
Eng. League II
11.19%
10.96%
10.48%
9.24%
8.74%
7.99%
8.13%
9.53%
Eng. Conf. League
11.47%
11.20%
11.02%
10.51%
9.98%
9.41%
9.40%
10.43%
Germ. Bundesliga I
10.15%
9.60%
8.79%
8.03%
7.45%
6.62%
6.52%
8.17%
Germ. Bundesliga II
11.20%
10.56%
10.03%
9.37%
8.86%
8.48%
8.30%
9.54%
Italian Serie A
10.17%
9.80%
8.96%
8.26%
7.63%
7.02%
6.65%
8.36%
Italian Serie B
11.15%
11.00%
10.72%
10.17%
9.74%
9.68%
9.13%
10.23%
Sp. La Liga Primera
10.19%
9.74%
8.94%
8.03%
7.36%
6.63%
6.59%
8.21%
Sp. La Liga Segunda
11.76%
11.68%
11.03%
10.29%
9.53%
9.22%
9.04%
10.36%
Dutch Eredivisie
11.15%
10.98%
10.17%
9.49%
8.82%
8.32%
8.22%
9.59%
Belgian Jupiler
11.29%
11.19%
10.32%
9.82%
9.25%
9.09%
9.00%
9.99%
Greek Ethniki Kat.
11.40%
11.26%
10.87%
10.48%
9.51%
8.85%
8.98%
10.19%
Table 2 presents the average profit margin in the EPL per specified bookmaker over the same
period. The results here show that the profit margin can be significantly different between
bookmaking firms; suggesting that the published odds of one bookmaker cannot be representative of
the whole market. This outcome contradicts the assumptions of (Forrest & Simmons, 2002; Forrest et
al., 2005) who based their conclusions on the notion that the odds of a single bookmaker are
representative of the whole market and that little information is lost by concentrating on just one
bookmaker.
Figures A.1 and A.2 provide further evidence that discredit the notion of a single
representative bookmaker by demonstrating how distinct profit margins per match of the EPL seasons
2010/11 and 2011/12 can further vary considerably. The results show that the profit margin for
distinct match instances of the same league and bookmaker can still vary considerably.
Table 2. Average profit margin introduced per specified bookmaker and EPL season.
Bookmaker
Season
2005/06
2006/07
2007/08
2008/09
2009/10
2010/11
2011/12
William Hill
12.49%
12.49%
12.37%
7.01%
7.35%
6.50%
6.70%
BET365
7.91%
7.95%
5.98%
5.31%
5.43%
5.44%
5.46%
Bwin
10.13%
10.07%
10.06%
10.07%
8.30%
8.01%
6.42%
Gamebookers
8.11%
8.04%
7.45%
7.29%
7.75%
7.68%
7.67%

- 5 -
Interwetten
12.33%
11.35%
11.39%
10.21%
8.36%
10.13%
10.07%
Ladbrokes
12.27%
12.32%
12.19%
9.26%
7.48%
6.49%
6.65%
Sportingbet
8.14%
10%
10.13%
10.14%
10.12%
10.12%
7.66%
By taking into consideration a relatively small number of different bookmaking firms,
previous studies (Dixon & Pope, 2004; Forrest et al., 2005) demonstrated no evidence of arbitrage
opportunities in recent years. However, that situation has now changed with the increase in number of
bookmakers. Indeed, this increase, together with the reduced bookmaker profit margins, means there
are extensive arbitrage opportunities and high return rates.
Figure 1(a) demonstrates the significant increase in arbitrage size per match instance as we
move towards season 2011/12; in particular, for the two most recent seasons the potential profit from
arbitrage has doubled. Figure 1(b) demonstrates the average profit per match instance generated
through arbitrage over all leagues and seasons for each country and shows that profitability from
arbitrage can vary significantly between leagues of different countries. Figure A.3 presents the
number of arbitrage instances discovered per 100 match instances for the specified league and season,
clearly demonstrating how arbitrage has become common over the more recent seasons. Table A.1
provides detailed information regarding a) the number of arbitrage opportunities discovered, b) the
average number of bookmaking firms taken into consideration, c) the average negative profit margin
attained and d) the average profit generated, for each football league and season.
Figure 1. Average profit generated through arbitrage per match instance over all leagues and per football season (graph (a)),
and over all leagues and seasons for each country (graph (b)).
When it comes to different levels of football divisions, Figure 2 demonstrates weak evidence
of increased profitability from arbitrage as we move towards lower divisions. However, the number
of bookmaking firms taken into consideration for computing maximums cannot explain this
behaviour. In fact, we can observe from Table A.1 that the average number of bookmaking firms
taken into consideration, per match instance, for the English divisions were (starting from top
division) 47.03, 41.26, 37.73, 36.96 and 32.69 respectively. In reality, we would have expected the
profitability generated for the lower divisions to decrease since the number of bookmaking firms
taken into consideration for arbitrage in those divisions is lower than in top divisions. As a result, we
can only infer that the divergence in odds provided by the various bookmaking firms is much higher
for lower divisions than it is for higher divisions, which would explain the higher profitability
generated via odds combination.

- 6 -
Figure 2. Average profit generated through arbitrage per match instance over all leagues and per division level (graph (a)),
and for the different division levels in England (graph (b)). Note that the results for divisions 3, 4 and 5 are practically
identical for both graphs since no data for weaker divisions was available in countries other than England.
3 BOOKMAKERS' ACCURACY
Since bookmakers increase profitability by encouraging bettors to place as many bets as possible,
their profit is not only determined by the profit margin, but also by the accuracy of their published
odds which should, therefore, represent a good approximation of the ‘true’ probabilities of any
particular match without introducing biases. In this section we examine the degree of variation
between bookmakers with regards to the accuracy of their normalised4 odds. For this assessment we
have considered the average bookmaking odds.
3.1. Methodology
For forecast assessment we make use of the Rank Probability Score (RPS), a scoring rule introduced
in (Epstein, 1969), and which has been described to be particularly appropriate in assessing both
interval and ordinal scale probabilistic variables (Murphy, 1970). (Strumbelj & Sikonja, 2010) have
already used the RPS to assess bookmakers' football match odds and (Constantinou & Fenton, 2012)
explain why RPS is the most rational scoring rule of those that have already been proposed and used
for assessing probabilities assigned to football outcomes. In brief, this scoring rule represents the
difference between the observed and forecasted cumulative distributions in which a higher difference
leads to a higher penalty (Wilks, 1995) which is subject to a negative bias that is strongest for small
ensemble size (Jolliffe & Stephenson, 2003). RPS is both strictly proper and sensitive to distance
(Murphy, 1969; Murphy, 1970) and for a single match instance is defined as
 


 


where r is the number of potential outcomes (three in this case), and pj and ej are the forecasts and
observed outcomes at position j (so, for example, pj is the forecast probability of  and ej is 1 if the
outcome was a home win and 0 otherwise). A lower score indicates a more accurate forecast (less
error). This methodology is defined in detail, along with relevant examples, in (Constantinou &
Fenton, 2012).
3.2. Results
Table 2 presents the accuracy scores, over all bookmakers, for each league and season. The
results are rather surprising. Whereas (Forrest et al., 2005) concluded that bookmakers provided more
4 The odds are normalised such that the profit margin is eliminated and the sum of the probabilities over the possible events
is equal to 1.

- 7 -
accurate odds over time, our results show no indications of forecast improvement in bookmaking
performance over the period of seven seasons and this is true for almost all of the 14 distinct leagues.
Overall, the results tend to suggest that the accuracy of the odds provided for top division leagues is,
to some extent, higher than lower division leagues.
Table 2. RPS assessment for each league over the seven seasons. Lower score indicates greater accuracy.
League
Season
2005-06
2006-07
2007-08
2008-09
2009-10
2010-11
2011-12
Overall
Eng. Premier Leag.
0.1949
0.1953
0.178
0.1917
0.1828
0.2006
0.2014
0.1921
Eng. Champ.
0.2083
0.2233
0.2178
0.215
0.2075
0.22
0.2176
0.2156
Eng. League I
0.2209
0.2206
0.2158
0.2195
0.2007
0.2203
0.2082
0.2151
Eng. League II
0.2186
0.2185
0.2289
0.2166
0.223
0.2176
0.2167
0.22
Eng. Conf. League
0.2161
0.2254
0.2105
0.2092
0.2094
0.2047
0.2105
0.2123
Germ. Bundesliga I
0.1918
0.2169
0.2055
0.2049
0.2068
0.228
0.2032
0.2082
Germ. Bundesliga II
0.2235
0.2094
0.2012
0.2072
0.2192
0.2071
0.2007
0.2098
Italian Serie A
0.1783
0.1829
0.1912
0.1954
0.1931
0.2008
0.1986
0.1915
Italian Serie B
0.1897
0.1896
0.1962
0.2041
0.2084
0.2026
0.2105
0.2002
Sp. La Liga Primera
0.2101
0.2101
0.2135
0.2025
0.1812
0.1907
0.1897
0.1997
Sp. La Liga Segunda
0.2148
0.2176
0.2106
0.2001
0.2085
0.2103
0.2083
0.21
Dutch Eredivisie
0.1956
0.1863
0.2048
0.1939
0.1709
0.1846
0.1893
0.1893
Belgian Jupiler
0.2028
0.1893
0.1993
0.2011
0.2042
0.1925
0.1901
0.197
Greek Ethniki Kat.
0.1639
0.1846
0.1818
0.1803
0.1784
0.2016
0.1929
0.1834
Table 3 presents the average RPS values generated for each specified bookmaker over this
period. Clearly, the accuracy of the odds between bookmakers turns out to be extremely consistent.
These results are also in agreement with (Forrest et al., 2005) who claimed that the variation in the
predictability of match results from season to season is much larger than the variation in forecasting
performance between bookmakers for the same season.
Table 3. Average RPS for each bookmaker and EPL season. Lower score indicates greater accuracy.
Bookmaker
Season
2005/06
2006/07
2007/08
2008/09
2009/10
2010/11
2011/12
William Hill
0.1952
0.1953
0.1799
0.1914
0.1832
0.2002
0.2010
BET365
0.1949
0.1955
0.1769
0.1915
0.1824
0.2003
0.2013
Bwin
0.1956
0.1960
0.1782
0.1921
0.1843
0.2010
0.2022
Gamebookers
0.1953
0.1949
0.1777
0.1918
0.1833
0.2014
0.2016
Interwetten
0.1967
0.1963
0.1798
0.1916
0.1832
0.2008
0.2014
Ladbrokes
0.1957
0.1953
0.1799
0.1927
0.1846
0.2004
0.2010
Sportingbet
0.1951
0.1954
0.1786
0.1921
0.1836
0.2006
0.2013
4 ODDS BIASES
In gambling markets, the favourite-longshot bias refers to the tendency for bets at short odds to
generate a higher return than bets at long odds. Possibly the strongest hypothesis behind this
phenomenon is the preference of the bettor in backing risky outcomes, which are also referred to as
longshots, and this phenomenon is observed in many different markets (Ali M., 1977; Quandt, 1986;
Thaler & Ziemba, 1988; Shin H., 1991, Shin R. E., 1992; Shin H., 1993; Woodland & Woodland,
1994; Vaughan Williams & Paton, 1997; Golec & Tamarkin, 1998; Jullien & Salanie, 2000), and
various theories exist, such as risk-loving behaviour, on why people are willing to bet on such
uncertain propositions (Sobel & Raines, 2003; Snowberg, 2010).
For example, consider a game between a top team playing at home against a very weak team.
Suppose the 'true' probability of a home win is 0.9. Then even if a bookmaker offered the perfectly
fair odds of 1.11 for a home win a typical bettor is reluctant apparently to place £100 bet to win only
£11. However, at the other extreme, if the 'true' probability of an away win in this match is 0.05 then a
typical bettor would be prepared to bet at less than fair odds of, say 15, because for a very small bet

- 8 -
£10 they stand to win £150. Bookmakers are believed to exploit these types of behaviour and increase
profitability by offering more-than-fair odds for 'safe' outcomes, and less-than-fair odds for 'risky'
outcomes.
However, while some researchers have focused their analysis on different bookmaking prices
for determining favourite-longshot biases (Cain et al., 2000), others have considered all home and
away wins serving as favourite and longshot outcomes respectively (Graham & Stott, 2008). When it
comes to home and away match instances, it might be true that away wins can be seen as longshots on
average (and vice versa), but the same does not hold for a significant proportion of matches.
Therefore, while the analysis in (Cain et al., 2000) helps us to understand market behaviour when it
comes to the favourite-longshot bias, the analysis in (Graham & Stott, 2008) could be best described
as a home-away bias. To avoid confusion between the different types of biases, in this section we
define and investigate the three following types of odds biases:
1. Favourite-longshot bias: as in (Cain et al., 2000), low probability outcomes can be
considered as longshots (and vice versa). In this paper we focus our longshot bias
analysis in grouped outcomes with less than 20%, 15% and 10% chance of occuring,
and our favourite bias analysis in groups with more than 60%, 70% and 80% chance
of occuring.
2. Home-away bias: as in (Graham & Stott, 2008), we assess home and away team
biases by focusing on home and away wins respectively.
3. Most-likely/least-ikely bias: unlike the favourite-longshot bias which only considers
high and low probability outcomes, here we investigate potential odds biases when it
comes to the most likely and the least likely outcomes for each match instance (and
therefore this takes into consideration all of the match instances).
Throughout this section we have considered the average match outcome odds, as provided by
numerous bookmaking firms, for the top divisions of England, Germany, Italy and Spain and over the
period of seven seasons; from 2005/06 to 2011/12. To investigate the biases we simulate £1 bets on
the relevant outcomes and compare the respective cumulative/final returns.
4.1. Favourite-longshot bias
Figure 3 presents the average profit per bet as generated by simulating bets on match
outcomes which satisfy the specified probabilistic conditions for favourite and longshot bets as
specified above. The results clearly demonstrate how loss is increased while bets move towards
longshot outcomes and returns break even while moving towards favourite outcomes. This implies
that the bias was not strong enough to overcome the averaged bookmaking profit margin (i.e. most of
the bookmaking firms falling between 5.5% and 12.5% from average bookmaking odds as
demonstrated in Section 2).

- 9 -
Figure 3. Profit rates as generated by simulating £1 bets on match outcomes which satisfy the specified probabilistic
favourite-longhshot conditions, given all of the top division match instances of England, Germany, Italy and Spain over the
seven football seasons (from 2005/06 to 2011/12).
Figures B.1 and B.2 demonstrate how this outcome differs for each country-league and season
respectively. With the exception of the German league, this phenomenon appears to be highly
consistent. When it comes to different seasons, however, the bias appears to have diminished in recent
seasons. In particular, the overall losses during season 2010-11 appear to be greater when betting on
favourites, whereas for season 2011-12 the greater overall losses return back to longshot bets but still
the losses seem to even out when compared to previous seasons. It should be noted that the
unexpected outcome observed during season 2010-11 is due to a high number of unexpected longshot
outcomes observed in the EPL.
4.2. Home-away bias
Figure 4 demonstrates the cumulative returns generated by simulating £1 bets on all home
wins, draws and away wins, of match instances of England, Germany, Italy and Spain. Seasons are
ordered such so that Season 1 represents season 2005/06 and Season 7 represents season 2011/12. The
results demonstrate bias on returns generated from bets on the outcomes of a home win. In particular,
betting on home wins yields the highest cumulative return under most of the scenarios. This
phenomenon does also not appear to be particularly profitable but still, with the exception of the
German league, home-away bias appears to be equally as strong as the favourite-longshot
phenomenon. This is because, while bets on away wins generate a high cumulative loss, bets on home
wins generate minimal cumulative loss over the whole period of seven seasons.
Figure 4. Cumulative returns generated by simulating £1 bets on all home wins, draws and away wins, of match instances of
England, Germany, Italy and Spain and over the period of seven seasons (from 2005/06 to 2011/12).

- 10 -
4.3. Most-likely/least-likely bias
Figure 5 demonstrates the profitability generated over the period of seven seasons by
simulating bets on all most-likely and least-likely outcomes for each country. Overall, the results
appear to be in agreement with the favourite-longshot bias, whereby most-likely bets generate
considerably higher returns than least-likely bets, and the results appear to be fairly consistent for the
cases of England, Italy and Spain.
Figure 6 presents the profit rates from bets on most-likely and least-likely outcomes when
partitioned into  outcomes. From this, we observe that:
a) For the cases of England and Germany, betting on a home win when home win is the least
likely outcome appears to be highly profitable with respective overall profit rates of 6.51% and
5.12%, and the rate of profitability appears to be consistent. In contrast, for the cases of Italy
and Spain this betting strategy generates significant losses with respective profit rates of -
17.61% and -14.08%, and which are also rather consistent.
b) Overall, betting on a home least-likely win, rather than an away least-likely win, generates
considerably higher returns. In particular, the average home win bet on least likely outcomes
generated profits of 6.51%, 5.12%, -17.61%, and -14.08%, whereas the average away win bet
generated profits of -26.22%, -3.66%, -25.28%, and -11.08%.
Figure 5. Cumulative returns generated by simulating £1 bets on all most-likely and all least-likely outcomes of match
instances of England, Germany, Italy and Spain over the period of seven seasons (from 2005/06 to 2011/12).

- 11 -
Figure 6. Profit rates generated by simulating £1 bets on all most-likely and least-likely outcomes, as partitioned into 
outcomes and each country over the period of seven seasons.
5 SUMMARY AND CONCLUSIONS
Overall, the evidence presented in this paper, based on recent market trends, casts further doubt on the
notion that the football betting market is efficient. We summarise our findings as follows:
Bookmakers' accuracy, profit margin and arbitrage
a) The market nowadays consists of a higher number of bookmaking firms who publish odds
with significantly lower profit margins than even before. This has led to frequent arbitrage
opportunities. The emergence of software to make it easier to spot arbitrage opportunities (for
example, websites such as www.oddschecker.com5) and evolving systems that perform
automated internet analysis in real time to spot arbitrage opportunities make the whole
process much easier. If this trend continues then in the near future bookmakers may be
exposed to substantial arbitrage risks.
b) Primarily based on English football data, evidence suggest that the profitability from arbitrage
increases when betting on match instances of lower divisions and the results tend to indicate
that the explanation for this is because the divergence in odds between bookmakers is much
higher for lower divisions.
c) We have demonstrated that profit margins can vary significantly between bookmaking firms,
and also for match instances of the same league offered by a single bookmaker. These
observations contradict the popular assumption that the published odds of one bookmaker are
representative of the whole market (Forrest & Simmons, 2002; Forrest et al., 2005).
d) Contrary to popular belief (Forrest et al., 2005; Graham & Stott, 2008) and based on a
suitable measure of forecast accuracy, our results show no evidence of forecast accuracy
improvement by bookmakers over a period of seven seasons and with 14 football leagues
taken into consideration.
5 A website that gives an overall view of the market and informs visitors about the best available odds by considering a large
number of various online bookmakers.

- 12 -
Odds biases:
a) The presence of the widely accepted favourite-longshot bias is still strong. However, while
(Cain et al., 2000) demonstrated this bias back 1991/92 with no evidence of potentially
profitable strategies on favourites, our results in many cases show evidence of high
profitability. For example, in Figure B.1 note the cases of England and Spain, and in Figure
B.2 the cases of seasons 2005/06 and 2009/10, whereby bets on outcomes with >80% chance
generated average profits of approximately 2%, 2%, 8% and 10% respectively. This outcome
does not necessary imply that the favourite-longshot bias is now stronger than what it was two
decades ago. The profitability could be explained by the recently reduced bookmaker profit
margins. In fact, Figure B.2 demonstrates how this bias is diminished over the seven seasons
under analysis, in which the average returns from each longshot bet with less than 10%
chance were -55%, -64%, -68%, -33%, -83%, 20%6 and -19% respectively (starting from
season 2005/06 to 2011/12). Similar behaviour is observed for longshot outcomes with 15%
and 20% chance of success. This reduction in loss, especially for the last two seasons, might
also be explained by the recently reduced bookmaker profit margins; in particular, it may be
possible that the bookmakers reduce the bias such so that the chance of profit when betting on
favourites is not increased due to the continuous decrease in profit margins.
b) Based on match instances of England, Italy and Spain, our results demonstrate a clear home-
away bias whereby the returns from bets on home win outcomes generate considerably higher
returns, when compared to bets on away wins, and this result is in agreement with (Graham &
Stott, 2008) who reported this outcome as evidence to support the favourite-longshot bias. In
fact, our results show that this type of bias is almost equally as strong as the favourite-
longshot bias.
c) Betting on the most likely outcome generates notably higher returns than betting on the least
likely outcome. The bias in this case, however, does not appear to be profitable on the whole.
d) When considering all of the least likely outcomes, the returns from bets on home wins
generate higher returns than returns from bets on away wins. While exploiting this bias,
evidence of consistent profitability have been observed for the cases of England and
Germany. In particular, over the period of seven seasons, the average profit from home win
bets on least likely outcomes for England and Germany were 6.51% and 5.12%. In the case of
bets on most-likely outcomes not strong bias has been discovered.
Due to the favourite-longshot bias observed over the previous seasons many have suggested
that this bias was due to bookmakers taking dynamic positions against the presumed tendency of the
bettors to underbet on favourites and to overbet on risky outcomes (Rossett, 1971; Snyder, 1978; Ali
M. M., 1979; Asch et al., 1984; Levitt, 2004; Graham & Stott, 2008). On the basis of this assumption,
which was also the case in various other gambling markets, a mixture of theories have been
formulated such as risk-loving behaviour on why people are willing to bet on such uncertain
propositions (Sobel & Raines, 2003; Snowberg, 2010). If the assumption of having bookmakers
taking positions against bettors for maximising profit is correct, then bookmakers' odds are prices
published with the intention of maximising profit, which would also support the above assumptions.
On the basis of the above theory, the additional odds biases demonstrated in this paper can
only be explained by the theory as follows:
1. On the basis of evidence from (b) above we can infer that bettors prefer betting on an away
win rather than on a home win.
2. On the basis of evidence from (c) above we can infer that bettors prefer betting on the least
likely outcome, and this is in agreement with the favourite-longshot bias reasoning.
3. On the basis of evidence from (d) above we can infer that, when home and away wins are the
least likely to occur, bettors still have preference betting on an away win rather than a home
win.
6 During season 2010/11 the returns from longshots with <10% chance generated highly profitable returns due to a large
number of surprising results in the English Premier League.

- 13 -
Although we have suggested possible explanation for the phenomena that we have observed,
the fact that the only data that we have available to work with is price data constrains our ability to
infer the true underlying causal influences. Such inferences require important pieces of data on such
(certainly unobserved) factors such as the proportion of bets placed on each outcome. Such data are
not available to the public. Planned extensions of this research will examine how published football
models which have already demonstrated profitability against market odds (Constantinou et al., 2012;
2013) might further benefit in betting effectiveness by formulating betting patterns which take into
consideration the different types and degrees of biases presented in this and other recent relevant
papers.
Acknowledgements
We acknowledge the financial support by the Engineering and Physical Sciences Research Council
(EPSRC) for funding this research. We would also like to thank Dr. Xiao-Lu Yu from the Department
of Physics and Astronomy, Mississippi State University, for indicating an error in an earlier version of
our analysis.

- 14 -
APPENDIX A: PROFIT MARGIN & ARBITRAGE OPPORTUNITIES
Figure A.1. Distinct profit margins, as introduced by the specified bookmaking firms, per match instance and over the EPL
season 2010/11.

- 15 -
Figure A.2. Distinct profit margins, as introduced by the specified bookmaking firms, per match instance and over the EPL
season 2011/12.

- 16 -
Figure A.3. Number of arbitrage instances discovered per 100 match instances for the specified league and season.

- 17 -
Table A.1. Arbitrage opportunities discovered per football league and season.
League
Season
Match
instances
considered7
Bookmaking
firms
considered8
Arbitrage
instances
found
Average profit
per match
instance
English
Premier
League
2005-06
380
57.84
33
2.29%
2006-07
380
44.70
26
2.63%
2007-08
380
43.39
25
0.54%
2008-09
380
40.16
75
0.53%
2009-10
380
36.02
53
0.57%
2010-11
380
35.73
62
0.55%
2011-12
380
41.56
70
0.52%
English
Champ.
2005-06
462
55.30
13
2.87%
2006-07
462
41.39
6
2.67%
2007-08
462
38.42
25
2.74%
2008-09
462
37.58
39
0.51%
2009-10
462
34.67
50
0.60%
2010-11
166
41.33
26
0.41%
2011-12
462
40.07
71
0.40%
English
League I
2005-06
462
49.10
8
1.07%
2006-07
461
37.86
6
0.21%
2007-08
462
36.81
16
1.06%
2008-09
462
34.38
51
0.83%
2009-10
462
32.74
52
0.71%
2010-11
462
34.33
45
0.55%
2011-12
462
38.88
70
0.47%
English
League II
2005-06
462
47.13
9
3.34%
2006-07
462
37.22
1
4.05%
2007-08
461
35.72
11
1.33%
2008-09
462
34.80
37
0.91%
2009-10
462
31.68
32
0.94%
2010-11
462
33.50
47
0.63%
2011-12
462
38.68
82
0.63%
English
Conference
League
2005-06
462
36.38
7
1.63%
2006-07
461
31.17
7
3.45%
2007-08
462
32.17
5
0.50%
2008-09
455
31.55
29
0.81%
2009-10
462
28.47
36
1.28%
2010-11
462
32.01
29
0.69%
2011-12
453
37.05
86
0.74%
German
Bundesliga
I
2005-06
306
56.49
34
0.92%
2006-07
306
44.46
12
0.56%
2007-08
306
41.93
25
0.62%
2008-09
306
38.55
67
0.72%
2009-10
306
34.74
34
0.54%
2010-11
306
36.14
61
0.83%
2011-12
306
41.56
70
0.68%
German
Bundesliga
II
2005-06
306
40.65
12
1.29%
2006-07
303
30.75
9
2.36%
2007-08
305
36.59
15
0.54%
2008-09
306
30.81
43
0.71%
2009-10
306
30.53
40
0.90%
2010-11
306
32.46
43
0.97%
2011-12
306
38.82
47
1.07%
2005-06
380
55.74
50
1.62%
7 The number of match instances considered for discovering arbitrage opportunities. For instance, if there were 10 missing
entries in an EPL dataset (which consists of a total of 380 match instances over a single season), the reported number of
instances considered would be 370.
8 Average bookmaking firms considered per match instance.

- 18 -
Italian
Serie A
2006-07
380
42.16
9
0.70%
2007-08
380
39.71
24
0.43%
2008-09
379
37.72
40
1.09%
2009-10
380
34.75
41
0.84%
2010-11
380
35.24
47
0.67%
2011-12
380
41.50
77
0.52%
Italian
Serie B
2005-06
460
31.81
16
2.52%
2006-07
462
30.2
10
1.30%
2007-08
461
32.52
32
0.69%
2008-09
461
29.94
44
1.07%
2009-10
456
25.24
39
1.24%
2010-11
462
27.81
20
0.91%
2011-12
462
37.69
28
2.15%
Spanish La
Liga
Primera
2005-06
380
56.96
44
1.29%
2006-07
380
44.15
17
0.49%
2007-08
380
42.26
16
0.71%
2008-09
380
37.28
53
0.66%
2009-10
380
36.19
42
1.17%
2010-11
380
36.15
59
0.79%
2011-12
380
41.54
89
0.51%
Spanish La
Liga
Segunda
2005-06
462
32.75
5
1.40%
2006-07
461
28.35
6
4.45%
2007-08
462
33.71
20
0.94%
2008-09
453
28.81
13
0.62%
2009-10
462
30.19
19
0.82%
2010-11
462
32.16
33
1.06%
2011-12
462
38.80
36
0.74%
Dutch
Eredivisie
2005-06
306
48.92
13
1.93%
2006-07
306
39.42
4
1.01%
2007-08
306
33.49
7
0.47%
2008-09
306
34.84
27
1.98%
2009-10
306
33.67
21
0.71%
2010-11
306
34.28
23
0.43%
2011-12
306
40.74
45
0.68%
Belgian
Jupiler
2005-06
306
45.12
11
1.47%
2006-07
306
36.32
5
0.24%
2007-08
306
37.08
16
1.20%
2008-09
306
33.41
17
1.40%
2009-10
210
30.76
11
0.34%
2010-11
240
31.24
11
0.65%
2011-12
240
37.21
23
0.53%
Greek
Ethniki
Katigoria
2005-06
204
25.44
9
2.01%
2006-07
200
24.93
1
0.22%
2007-08
204
29.40
15
1.46%
2008-09
204
29.13
11
1.42%
2009-10
204
25.30
5
0.57%
2010-11
204
28.45
16
0.83%
2011-12
204
35.88
38
0.85%

- 19 -
APPENDIX B: ODDS BIASES
Figure B.1. Profit rates as generated by simulating £1 bets on match outcomes which satisfy the specified probabilistic
favourite-longshot conditions, given all of the top division match instances over the seven football seasons (from 2005/06
to 2011/12) and as partitioned for England, Germany, Italy and Spain.

- 20 -
Figure B.2. Profit rates generated by simulating £1 bets on match outcomes which satisfy the specified probabilistic
conditions, given all of the top division match instances of England, Germany, Italy and Spain, and as partitioned for each of
the seven football seasons.

- 21 -
References
[1] Ali, M. M. (1979). Some evidence of the efficiency of a speculative market. Econometrica, 47, 387-392.
[2] Ali, M. (1977). Probability and utility estimates for racetrack bettors. Journal of Political Economy, 85, 803-815.
[3] Asch, P., Malkiel, B. G., & Quandt, R. E. (1984). Market efficiency in racetrack betting. Journal of Business, 57,
165-175.
[4] betfair. (2000). About Us: How does Betfair work? Retrieved November 22, 2011, from
http://help.betfair.com/contents/itemId/i65767197/index.en.html
[5] bwin Group. (2009). Annual Report 2008. Vienna, Austria: bwin Interactive Entertainment AG.
[6] Cain, M., Law, D., & Peel, D. (2000). The Favourite-Loghshot Bias and Market Efficiency in UK Football
Betting. Scottish Journal of Political Economy, 47, 25-36.
[7] Constantinou, A. C., & Fenton, N. E. (2012). Solving the problem of inadequate scoring rules for assessing
probabilistic football forecast models. Journal of Quantitative Analysis in Sports, Vol. 8: Iss. 1, Article 1.
[8] Constantinou, A., Fenton, N., & Neil, M. (2012). pi-football: A Bayesian network model for forecasting
Assocciation Football match outcomes. Knowledge-Based Systems, 36: 322, 339.
[9] Constantinou, A., & Fenton, N. (2013). Determining the level of ability of football teams by dynamic ratings based
on the relative discrepancies in scores between adversaries. Journal of Quantitative Analysis in Sports, Vol. 9, Iss.
1, 37-50.
[10] Constantinou, A., Fenton, N., & Neil, M. (2013). Profiting from an inefficient Association Football gambling
market: Prediction, Risk and Uncertainty using Bayesian networks. Knowledge-Based Systems, 50: 60-86.
[11] Dixon, M., & Coles, S. (1997). Modelling association football scores and inefficienies in the football betting
market. Applied Statistics, 46, 265-80.
[12] Dixon, M., & Pope, P. (2004). The value of statistical forecasts in the UK association football betting market.
International Journal of Forecasting, 20, 697-711.
[13] Epstein, E. (1969). A Scoring System for Probability Forecasts of Ranked Categories. Journal of Applied
Meteorology, 8, 985-987.
[14] Finnigan, M., & Nordsted, P. (2010). The Premier Football Betting Handbook 2010/11. Hampshire, Great Britain:
Harriman House.
[15] Forrest, D., & Simmons, R. (2001). Globalisation and efficiency in the fixed-odds soccer betting market.
University of Salford, Salford: Mimeo.
[16] Forrest, D., & Simmons, R. (2002). Outcome uncertainty and attendance demand in sport: the case of English
soccer. The Statistician, 2, 241-291.
[17] Forrest, D., & Simmons, R. (2008). Sentiment in the betting market on Spanish football. Applied Economics, 40,
119-126.
[18] Forrest, D., Goddard, J., & Simmons, R. (2005). Odds-setters as forecasters: The case of English football.
International Journal of Forecasting, 21, 551-564.
[19] Global Betting and Gaming Consultants. (2001). 1st annual review of the global betting and gaming market, 2001.
West Bromwich: Global Betting and Gaming Consultants.
[20] Goddard, J., & Asimakopoulos, I. (2004). Forecasting Football Results and the Efficiency of Fixed-odds Betting.
Journal of Forecasting, 23, 51-66.
[21] Golec, J., & Tamarkin, M. (1991). The degree of inefficiency in the football betting market. Journal of Financial
Economics, 30, 311-323.
[22] Golec, J., & Tamarkin, M. (1998). Bettors love skewness, not risk, at the horse track. Journal of Political
Economy, 106, 205-225.
[23] Graham, I., & Stott, H. (2008). Predicting bookmaker odds and efficiency for UK football. Applied Ecomonics,
40, 99-109.
[24] Hausch, D. B., Ziemba, W. T., & Rubinstein, M. (1981). Efficiency of the market for racetrack betting.
Management Science, 27, 1435-1452.
[25] Henery, R. J. (1999). Measures of over-round in performance index betting. The Statistician, 48, 435-439.
[26] Hvattum, L. M., & Arntzen, H. (2010). Using ELO ratings for match result prediction in association football.
International Journal of Forecasting, 26, 460-470.
[27] Jolliffe, I. T., & Stephenson, D. B. (2003). Forecast Verification. A Practitioner's Guide in Atmospheric Science.
West Sussex, England: Wiley.
[28] Jullien, B., & Salanie, B. (2000). Estimating preferences under risk: the case of racetrack bettors. Journal of
Political Economy, 108, 503-530.
[29] Kuypers, T. (2000). Information and efficiency: an empirical study of a fixed odds betting market. Applied
Economics, 32, 1353-1363.
[30] Levitt, S. D. (2004). Why are gambling markets organised so differently from financial markets? The Economic
Journal, 114, 223-246.
[31] Mintel Intelligence Report. (2001). Online betting. London: Mintel International Group Ltd.
[32] Murphy, A. (1969). On the "ranked probability score". Journal of Applied Meteorology, 8, 988-989.
[33] Murphy, A. (1970). The ranked probability score and the probability score: A comparison. Monthly Weather
Review, 98, 917-924.
[34] Paton, D., & Vaughan Williams, L. (2005). Forecasting Outcomes in Spread Betting Markets: Can Bettors Use '
Quarbs' to Beat the Book? Journal of Forecasting, 24, 139-154.

- 22 -
[35] Peel, D. A., & Thomas, D. A. (1988). Outcome uncertainty and the demand for football. Scottish Journal of
Political Economy, 35, 242-249.
[36] Peel, D. A., & Thomas, D. A. (1992). The demand for football: some evidence on outcome uncertainty. Empirical
Economics, 4, 567-570.
[37] Peel, D. A., & Thomas, D. A. (1997). Handicaps, outcome uncertainty and attendance demand. Applied Economic
Letters, 4, 567-570.
[38] Pope, P., & Peel, D. (1989). Information, prices and efficiency in a fixed-odds betting market. Economica, 56,
323-341.
[39] Quandt, R. E. (1986). Betting and equilibrium. Quarterly Journal of Economics, 101, 201-207.
[40] Rossett, R. N. (1971). Weak experimental verification of the expected utility hypothesis. Review of Economic
studies, 38, 481-492.
[41] Rue, H., & Salvesen, O. (2000). Prediction and retrospective analysis of soccer matches in a league. The
Statistician, 49, Part 3, pp. 339-418.
[42] Sauer, R. D. (1998). The Economics of Wagering Market. Journal of Economic Literature, 36, 2021-2064.
[43] Sauer, R., Brajer, V., Ferris, S., & Marr, M. (1988). Hold your bets: another look at the efficiency of the gambling
market for national football league games. Journal of Political Economy, 96, 206-213.
[44] Shin, H. (1991). Optimal betting odds against insider traders. Economic Journal, 101, 1179-1185.
[45] Shin, H. (1993). Measuring the incidence of insider trading in a market for state-contingent claims. Economic
Journal, 103, 1141-1153.
[46] Shin, R. E. (1992). Prices of state contingent claims with insider traders, and the favourite longshot bias. Economic
Journal, 102, 426-435.
[47] Snowberg, E. (2010). Explaining the Favorite-Long Shot Bias: Is it Risk-Love or Misperceptions? Journal of
Political Economy, 118, 4, 723-746.
[48] Snyder, W. N. (1978). Horse racing: testing the efficient markets model. Journal of Finance, 33, 1109-1118.
[49] Sobel, R. S., & Raines, S. T. (2003). An examination of the empirical derivatives of the favourite-longshot bias in
racetrack betting. Applied Economics, 35, 371-385.
[50] Strumbelj, E. & Sikonja, M. R. (2010). Online bookmakers' odds as forecasts: The case of European soccer
leagues. International Journal of Forecasting, 26: 482-488.
[51] Thaler, R., & Ziemba, W. (1988). Parimutuel betting markets: Racetracks and lotteries. Journal of Economic
Perspectives, 2, 161-174.
[52] Vaughan Williams, L., & Paton, D. (1997). Why is there a favourite-longshot bias in British racetrack betting
markets? Economic Journal, 107, 150-158.
[53] Vecer, J., Kopriva, F., & Ichiba, T. (2009). Estimating the Effect of the Red Card in Soccer: When to Commit an
Offense in Exchange for Preventing a Goal Opportunity. Journal of Quantitative Analysis in Sports, 5: Iss. 1,
Article 8.
[54] Vergin, R. C., & Scriabin, M. (1978). Winning strategies for wagering on National Football League games.
Management Science, 24, 809-818.
[55] Wilks, D. (1995). Statistical Methods in the Atmospheric Sciences. International Geophysics Series, Vol. 59,
Academic Press, 467 pp.
[56] Woodland, L., & Woodland, B. (1994). Market efficiency and the favourite-longshot bias: the baseball betting
market. The Journal of Finance, 49, 269-279.
[57] Woodland, L., & Woodland, B. (2001). Market efficiency and profitable wagering in the National Hockey League:
can betting score in longshots? Southern Economic Journal, 67, 983-995.
[58] Zuber, R., Gandar, J., & Bowers, B. (1985). Beating the spread: testing the efficiency of the gambling market for
national football league games. Journal of Political Economy, 93, 800-806.