Journal of Sports Economics: Playing It Safe? A Fibonacci Strategy For Soccer Betting
Journal of Sports Economics: Playing It Safe? A Fibonacci Strategy For Soccer Betting
Journal of Sports Economics: Playing It Safe? A Fibonacci Strategy For Soccer Betting
com/
Sports Economics
Published by:
http://www.sagepublications.com
On behalf of:
The North American Association of Sports EconomistsThe North American Association of Sports
Economists
Additional services and information for Journal of Sports Economics can be found at:
Subscriptions: http://jse.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://jse.sagepub.com/content/8/3/295.refs.html
What is This?
Downloaded from jse.sagepub.com at TULANE UNIV on September 28, 2014
Research Notes
Playing It Safe?
A Fibonacci Strategy for Soccer Betting
FRAGISKOS ARCHONTAKIS
Universidad Politécnica de Valencia-CSIC
EVAN OSBORNE
Wright State University
Soccer betting in Europe has grown rapidly in the past two decades. It is well known
that such sports betting markets can be analyzed similarly to financial markets. The pre-
sent article differs from others in the literature in two respects. The first is that the outcome
of betting interest is the draw. The second is the betting strategy, which relies on the
Fibonacci sequence to generate bets. Using this approach on International Federation of
Football Association World Cup Finals data, it is possible to earn economic profits
through this strategy in expectation, albeit with fairly large risk.
S occer betting in Europe has grown rapidly in the past two decades. Some betting
firms in the United Kingdom have 1,500 to 2,000 locations with tens of thousands
of employees. More than a decade ago, Jackson (1994) could characterize sports
betting as one of the fastest growing areas in the UK’s betting industry, and
Forrest and Simmons (2001) depict soccer betting in particular as the fastest-
growing type of gambling there.
However, this betting market, like most in sports, appears to be inefficient in the
same way as international financial markets (e.g., Osborne, 2001). Indeed, there is
some evidence that many such markets are systematically far from being efficient.
In an analysis of sports betting markets Sauer (1998) provides—within the context
of an article generally arguing that such markets are efficient—a list of anomalies
that must be resolved to be consistent with standard zero-profit conditions. In a
AUTHOR’S NOTE: This article has benefited from comments by seminar participants at IASE’s 4th
International Conference on Sports, Tourism and Culture: May 31-June 2, 2004 held in Athens,
Greece. Correspondence address: INGENIO (UPV-CSIC); Universidad Politécnica de Valencia;
Edificio i4, Camino de Vera s/n; E-46022 Valencia, Spain. E-mail: fraguis@ingenio.upv.es
JOURNAL OF SPORTS ECONOMICS, Vol. 8 No. 3, June 2007 295–308
DOI: 10.1177/1527002506286775
© 2007 Sage Publications
295
Downloaded from jse.sagepub.com at TULANE UNIV on September 28, 2014
296 JOURNAL OF SPORTS ECONOMICS / June 2007
market based on fixed-odds betting, there are mixed results. Pope and Peel (1989)
examine the efficiency of such markets generally and find that although there was
some evidence of ex post inefficiency, there did not appear to be profitable betting
strategies that could have been implemented ex ante during the sample period. But
Dixon and Coles (1997) propose a Poisson parametric model, similarly motivated
by the possibility of inefficiencies in the soccer-betting market. Their model does
yield positive profits. To the extent that such inefficiencies exist, a natural question
is whether they can be systematically exploited.
The present work differentiates itself from the others in the relevant literature
in that the only random outcome of interest is the draw, which Pope and Peel
(1989) speculate is the most difficult for bookmakers to predict. They conclude
that odds-setting practices among bookmakers (p. 328) “could simply reflect a
general inability to predict draw outcomes with any degree of reliability.”
We also seek to bridge the gap between the theory (i.e., the numerous theoret-
ical papers regarding soccer betting markets) and the practice—that is, the rela-
tively scarce literature regarding simple applied techniques on gambling. We do
this by finding a good estimate of the distribution of soccer-match results and pro-
pose a way of taking advantage of this information. Furthermore, we use the
results of previous researchers regarding the distribution of game results and con-
sider a variation of the martingale casino strategy as the betting method. A
Fibonacci sequence is used to define a consistent betting-quantity strategy. Most
of the sports-betting literature has paid little attention to strategy with respect to
betting amounts, concentrating instead on which way to bet.
Fibonacci numbers are used in finance in technical analysis of stock markets
(Goetzmann, 2004; Murphy, 1986). It will be shown that for fixed odds given for
a draw of at least 2.618, the betting rule proposed is profitable in expectation. The
strategy will then first be applied to actual International Federation of Football
Association (FIFA) 2002 World Cup Finals Tournament (WCFT) data. We also
explore the distribution of actual WCFT draw odds, consider them as a random
variable, and employ Monte Carlo tactics to further explore profit opportunities.
The rest of the article is organized as follows: Section 2 explains how fixed
odds work and briefly describes the data; Section 3 measures draw probabilities
over several iterations of the WCFT and tests their robustness across time; the
betting rule is discussed in Section 4; and Section 5 concludes.
League and do not exist in the National Basketball Association and Major League
Baseball. Thus, for both gamblers and scholars, the main outcome of interest is
whether a particular team wins. But in soccer, draws occur approximately 30% of the
time. Hence, we choose to use this as the uncertain outcome on which bets are placed.
Soccer betting typically occurs on an odds basis. In the more familiar notation
of the probability of various outcomes, if pi is the probability of one event, odds
on the event are then 1/pi. In this instance, the probability space has only two out-
comes (draw, not draw) and is defined as follows:
TABLE 1: Summary Report of Number of Draws Appeared in Each Group and Final
Stages per Tournament Year
Group matches
1 5/6 3/6 0/6 1/6 2/6 3/6
2 0/6 2/6 1/6 2/6 4/6 1/6
3 1/6 1/6 0/6 3/6 2/6 1/6
4 2/6 1/6 1/6 0/6 2/6 1/6
5 3/6 2/6 1/6 2/6 4/6 2/6
6 1/6 2/6 5/6 0/6 1/6 3/6
7 1/6 1/6
8 0/6 2/6
Total groups 12/36 11/36 8/36 8/36 16/48 14/48
Final competition
Round of 16 1/3 1/8 4/8 2/8 2/8 3/8
Quarter finals 2/3 3/4 2/4 1/4 1/4 2/4
(except 1982) 0/3
1/3
Semifinals and finals 1/4 1/4 2/4 1/4 1/4 0/4
Total final competition 5/16 5/16 8/16 4/16 4/16 5/16
Total tournament 17/52 16/52 16/52 12/52 20/64 19/64
each game is the highest (and perhaps risk aversion with respect to making a play
that leads to a loss higher as well) is 32.29%.
p̌ = Σyi/T = –y
From the theorem on the asymptotic normality of the MLE (see, for example,
Rice, 1995, for more details) for large samples p̌ ≈ N(p, p(1–p)/T)—that is, the
MLE of p̌ follows a Gaussian distribution with mean p and variance p(1–p)/T.
Thus, an approximate 95% confidence interval (CI) for p̌ would be
From Table 1, the total number of draws is 100, whereas the total number of
games is T = 336 observations. Hence p̌ = Σyi/T = y–, and as a result p̌ = 0.2976.
This estimate for the WCFT is similar to those reported for European league
soccer by Dixon and Coles (1997) and Stefani (1983).
Figure 1 plots the percentage of draws in each WCFT in the data set and the
associated 95% confidence interval. The 1994 tournament held in the United
States is an obvious outlier, with notably fewer draws. The 1994 finals were dif-
ferent in two respects. First, the travel requirements were much greater owing to
the greater distance across which the tournament was spread. Second, in that year,
there was a significant, one-time rule change.
This rule meant that the goalkeeper could not catch the ball with his hand
when receiving a pass by a teammate. This limits defensive options near the own
goal and hence raises the probability of yielding a score. Figure 1 indicates that
in the next tournament, the number of draws immediately returned to normal
levels. It also suggests that the cumulative draw probability for the WCFT games
is within the approximate range 0.30-0.35—that is, roughly 1/3 of WCFT games
are drawn.
BETTING RULE
The betting rule proposed for draws has similarities with the so-called martin-
gale strategy, used in some casino games (e.g., roulette, in addition to financial
markets). It is simple in the sense that one bets only for a draw, but this also
makes it easy to apply. The strategy is to bet continuously for a draw with
amounts defined by the Fibonacci sequence. The Fibonacci sequence is defined
as the element 1, followed by another 1, with each element thereafter the sum of
the previous two elements. For example, the first few elements of a Fibonacci
sequence are 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, etc. The mathematical sequence is pro-
duced by the recursive formula.
an + 1 = an + an – 1,
where a1 = 1 and a2 = 1.
0.35
0.30
0.25
0.20
0.15
1982 1986 1990 1994 1998 2002
Tournament Year
Figure 1: Average Probability of Draws per Tournament and Cumulative Probability Along
With Its 95% Confidence Intervals
√
(1 ± 5)
λ1,2 = .
2
The positive root (1+√5)/2 is also known as the golden ratio and is approxi-
mately equal to φ = 1.618, because for consecutive Fibonacci terms it is known
that their ratio an + 1/ an → φ as n → ∝. The nth Fibonacci term is given by Binet’s
Formula.
√ √
[(1 + 5)/2]n − [(1 − 5)/2]n
an = √ .
5
Given the assumptions about the distribution of trial results, the gambler has a
probability p of winning on any one of a sequence of bets. A successful bet of an
monetary units given fixed odds of b yields an amount won of ban. The probabil-
ity of winning for the first time at the xth bet is given by p(1 – p)x – 1 (i.e., it fol-
lows the geometric distribution). The mean of x is 1/p and its variance is (1–p)/p².
Formally, the betting rule is to start with a unit bet a1 and then follow it with
another unit bet a2. From the third bet onwards, the nth bet an is the sum of the two
prior ones. Table 2 summarizes the Fibonacci betting rule. In Table 2, we denote
1 1 1 b b-1
2 1 2 b b-2
3 2 4 2b 2 b-4
4 3 7 3b 3 b-7
5 5 12 5b 5 b-12
6 8 20 8b 8 b-20
7 13 33 13 b 13 b-33
n an an+2 –1 an b an b – (an+2 –1)
a current bet as an , the sum of previous invested bets (i.e., the total cost) as Sn and
the revenue and profit if the bet is successful for the first time in a series at time
n as Rn and Pn = Rn – Sn, respectively.
We next calculate the minimum fixed odds number b required to earn profits in
each series of bets. The following equation should hold for the last entry of Table 2.
For profits to occur, the above equation should exceed one. Because an+2 =
an + 1 + an , setting the equation equal to 1 yields
40
30
20
10
0
2.75 3.00 3.25 3.50 3.75 4.00 4.25 4.50 4.75 5.00
N – 1 non-draws followed by a draw with low odds (i.e., with b < 2.618, which
will yield PN < 0). Thus, the Fibonacci betting rule is not profitable for some plau-
sible combinations of draw odds, adding to overall uncertainty.
CONCLUSION
This article considers the unexplored fact in the wagering literature that the
event of a draw between two soccer teams can be thought of as random and has
yet to be modeled. This idea of draw odds as a random variable is especially
appealing when bookmakers set the odds. By using a 20-year set of soccer World
Cup data, we find that the probability of a draw is relatively stable at 29.76%.
The betting strategy suggested is to bet on draws, where bets are placed con-
secutively until victory using the Fibonacci sequence. There are two primary
results. First, for fixed odds of a draw of at least 2.618, the betting rule proposed
is profitable in expectation. This adds to the extensive empirical literature indi-
cating that profitable betting strategies can be consistently applied to sports-gam-
bling markets. In addition, because the average for the fixed odds of a draw are
often greater than 3.0, we consider the odds also as a random variable and the
model is implemented by a Monte Carlo simulation. The simulation experiment
10 7.25 10.27
20 4.79 9.53
30 3.66 6.22
60 5.69 12.22
100 8.19 45.52
200 41.19 305.93
500 34.82 561.44
1,000 17.58 158.13
5,000 49.47 1,786.10
10,000 47.11 2,482.20
100,000 1,054.10 280,200.00
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
–2 0 2 4 6 8 10 12 14 16 18
Figure 3: Example of a Probability Density Function Estimate for the Profit Distribution by
Using the Fibonacci Betting Rule (n = 30)
reveals that the Fibonacci betting rule, although profitable in expectation, is risky.
In particular, there is the possibility of gambler’s ruin, the rapid increase in losses
as draws fail to appear for an extended period.
305
APPENDIX
(continued)
306
Betting Revenue for Betting Revenue for Betting Revenue
Match Half- Strategy Betting Strategy Betting Strategy for Betting
No. Date City Teams Final Time Draw A Strategy A B Strategy B C Strategy C
Total –3 26 222
Revenues
33 June 11 Incheon DEN : FRA 2:0 1:0
34 June 11 Suwon SEN : URU 3:3 3:0 X
35 June 11 Shizuoka CMR : GER 0:2 0:0
36 June 11 Yokohama KSA : IRL 0:3 0:1
37 June 12 Miyagi SWE : ARG 1:1 0:0 X
38 June 12 Osaka NGA : ENG 0:0 X
NOTE: ARG = Argentina; BEL = Belguim; BRA = Brazil; CHN = China; CMR = Camaroon; CRC = Cost Rica; CRO = Croatia; DEN = Denmark; ECU = Ecuador;
ENG = England; ESP = Spain; FRA = France; GER = Germany; IRL = Ireland; ITA = Italy; JPN = Japan; KOR = Korea; KSA = Saudia Arabia; MEX = Mexico;
NGA = Nigeria; PAR = Paraguay; POL = Poland; POR = Portugal; RSA = South Africa; RUS = Russia; SEN = Senegal; SVN = Slovenia; SWE = Sweden; TUN =
Tunisia; TUR = Turkey; URU = Urugray; USA = United States; AET = after extra-time; PSO = penalty shoot-out.
NOTES
1. Another issue, which we do not explore, is whether betting firms will allow an
extended series of bets on draws.
2. The authors contacted one of the biggest betting companies in the United Kingdom
but unfortunately the request for the fixed odds data of the FIFA World Cup 2002 was not
granted.
REFERENCES
Brown, W. O., & Sauer, R. D. (1993). Fundamentals or noise? Evidence from the professional
basketball betting market. Journal of Finance, 48, 1193-1209.
Dana, J. D., & Knetter, M. M. (1994). Learning and efficiency in a gambling market. Management
Science, 40, 1317-1328.
Dixon, M. J., & Coles, S. G. (1997). Modelling association soccer scores and inefficiencies in the soc-
cer betting market. Applied Statistics, 46, 265-280.
Forrest, D., & Simmons, R. (2001). Globalisation and efficiency in the fixed-odds soccer betting market.
Salford, UK: Mimeo, University of Salford.
Goetzmann, W. N. (2004). Fibonacci and the financial revolution. (NBER Working Paper No. 10352).
Cambridge, MA: National Bureau of Economic Research.
Gray, P. K., & Gray, S. F. (1997). Testing market efficiency: Evidence from the NFL sports betting
market. Journal of Finance, 52, 1725-1737.
International Federation of Football Association. (2002). Fédération Internationale de Football
Association web site. Retrieved July 14, 2002, from http://fifaworldcup.yahoo.com/
Jackson, D. (1994). Index betting on sports. The Statistician, 43, 309-315.
Murphy, J. J. (1986). Technical analysis of the futures markets. New York: New York Institute of
Finance.
Osborne, E. (2001). Efficient markets? Don’t bet on it. Journal of Sports Economics, 2, 50-61.
Pope, P. F., & Peel, D. A. (1989). Information, prices and efficiency in a fixed-odds betting market.
Economica, 56, 323-341.
Rice, J. A. (1995). Mathematical statistics and data analysis (2nd ed.). Belmont, CA: Duxbury.
Sauer, R. D. (1998). The economics of wagering markets. Journal of Economic Literature, 36, 2021-2060.
Stefani, R. T. (1983). Observed betting tendencies and suggested betting strategies for European
soccer pools. The Statistician, 32, 319-329.
Evan Osborne conducts research in and teaches about public choice, ethnic conflict, global-
ization, and the economics of development, international trade, sports, and the law. He
received a bachelor's degree in economics from the University of Texas in 1985 and a Ph.D.
in economics from UCLA in 1993. He currently teaches at Wright State University in Dayton,
Ohio, and performed some research for this article while on sabbatical at Osaka University
Institute of Social and Economic Research. He is married and the father of two children.