Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
The Suntory and Toyota International Centres for Economics and Related Disciplines,
The London School of Economics and Political Science, London School of Economics,
Wiley are collaborating with JSTOR to digitize, preserve and extend access to Economica
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
Economica, 54, 289-298
The efficiency and profitability of exotic racetrack bets such as exactas and daily doubles are
examined. Efficiency is understood to mean that above average returns cannot be made in
the long run once risk is appropriately controlled for. The markets in question are found not
to be efficient; the inefficiencies, however, are insufficient to permit simple strategies to show
a consistent profit. Some evidence of "smart money" exists in that holders of inside information
may bet on exactas rather than equivalent standard bets in order to avoid signalling their
actions to the betting public.
INTRODUCTION
W(1-t)
R = p-1.
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
290 ECONOMICA [AUGUST
If the bet is to have positive expected return, the proportion of all dollars bet
on the horse must not exceed pi(I - t). To devise a profitable betting strategy
thus requires the identification not only of winners, but of winners that the
betting public does not regard as winners.
This paper investigates the efficiency of a class of wagers known as 'exotic
bets'. The members of this class are compound bets, and require more compli-
cated estimates of outcomes than do the familiar win, place and show bets.
Because of the track take, all racetrack bets tend to have negative expected
returns.
Even if it is difficult to explain why individuals engage repetitively in such
a money-losing activity, it is interesting and important to examine whether
differential avenues to losing money tend to equalize rates of return. This is
the basic concern of the present paper. In Section I we characterize these bets
and the data to be employed. In Section II we formulate our hypotheses
concerning exactas and present our analysis of the data. Section III proceeds
analogously with respect to daily doubles. Section IV presents an explanation
of a discrepancy between the exacta and daily double findings. Some brief
conclusions are presented in Section V.
The standard racetrack wagers are win, place and show bets. The minimum
bet is usually $2, and a win bet on horse i is successful if and only if horse i
wins the race. In this case, the payoff per dollar bet is W(1 - t)/ Wi, where
the terms are as defined above. A place bet on horse i is successful if horse i
comes in first or second; and a show bet is successful if horse i comes in first,
second or third.2
Exotic bets involve at least two simultaneous wagers on different horses.
In this paper, we investigate the exacta and the daily double. In the exacta,
the bettor must pick two horses: one to win and one to come in second. The
bet is successful if both picks are correct, and is unsuccessful otherwise. In
the daily double, the bettor picks the winners of two consecutive races; the
bet is successful if both picks win, and is unsuccessful otherwise. The racetrack
designates the races in which such betting is available.
It is obvious that the exotic bets are relatively low-probability wagers. Since
the track take also tends to be higher on exotic bets,3 these bets should appeal
to relative risk-lovers. Perhaps the most widely established empirical regularity
in racetrack gambling concerns win bets: horses with low probabilities of
winning ('long shots') are overbet; and horses with high probabilities of
winning ('favourites') are underbet. That is, the proportion of all moneys bet
on long shots is greater than their objective winning probability, with the
reverse holding for favourites (Rosett, 1965, 1971; Snyder, 1978; Ali, 1979;
Asch, Malkiel and Quandt, 1982). As a result, rates of return are higher for
favourites than for long shots.
Whether a similar pattern ought to be expected for exactas or daily doubles
is unclear. In straight win betting, one may easily distinguish between 'high-
probability' bets (with success probabilities of perhaps 03-06) and 'low-
probability' bets (in the range of 002-005). For exactas, however, the prob-
abilities of success may lie largely in the range 0O05-0O001; and we do not
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
1987] EFFICIENCY AND PROFITABILITY 291
If hi, is the objective probability that the (ij)-exacta is successful, the expected
value Eij and the variance Vij of return are
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
292 ECONOMICA [AUGUST
is successful and 0 otherwise, and N is the number of exactas over which the
average is taken.
The average rates of return to exacta bets show an erratic, if not random,
relationship with respect to payoff class.6 It is not surprising that some rates
of return are better than -0 19 (the track take) and not even surprising that
a few are positive; what is significant is that these better-than-expected rates
of return do not appear to be related to payoff class. Thus, we cannot confirm
the existence of a typical underbetting/overbetting bias.
(-1-9965) (31.5177)
(5) h J=1^
1-Pi
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
1987] EFFICIENCY AND PROFITABILITY 293
TABLE 1
Pi Si
0-0030 0-0068
0-0030 0-0110
0-0059 0-0151
0-0119 00201
00178 0-0258
0-0445 0-0323
0-0297 0-0398
0-0386 0-0489
0-0415 0O0581
00804 0-0683
0-0804 0-0787
00923 0-0908
0 1310 0-1036
01161 0-1185
0-1250 0-1368
0-1637 0-1587
01548 01841
0-2024 0-2220
0-3006 02747
0-4554 0-4031
s* = 0-00683 + 1-10267s,j
(186.374) (683-350)
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
294 ECONOMICA [AUGUST
(7) A* = (1 *t).
A 'smart money' hypothesis can be tested by examining the differences A* - A11.
We group the differences for all 41,246 exacta pairs as follows: group I contains
the 51OAO - Aij values for exacta combinations that actually won, and group
II contains the values for the remaining combinations.
Suppose that there are bettors with inside information. To avoid sending
clear signals, they utilize their information by betting on exactas. Whereas, in
the no-inside-information case, we would expect the distribution of differences
V - Aij to be statistically the same for groups I and II, the 'smart money'
hypothesis suggests that actual payoffs Aij will tend to be depressed relative
to A* for exacta winners (group I). The Kolmogorov-Smirnov test rejects the
hypothesis that the two distributions of A* - Aij values were drawn from the
same parent distribution.
The means and variances of the two groups are shown in Table 2. The
differences in means are highly significant and are in the direction predicted
by the 'smart money' hypothesis.
TABLE 2
Winning Losing
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
1987] EFFICIENCY AND PROFITABILITY 295
We now consider only winning daily double pairs. Their actual payoffs are
denoted by Dk (k = 1, . . . , 122). For each of these, we can also compute the
equivalent parlay payoffs, denoted by Lk. If the market is efficient, it should
not, on average, make any difference whether one bets a given pair of horses
via the daily double or a parlay.
This is not the case, however. The average of the daily double payoffs is
$52-54 and the parlay payoffs $41-38, suggesting that it is substantially more
profitable to bet on the daily double.9 The reason may well be that the
commitment to a daily double must be made before the start of the first of
two consecutive races, i.e., before the evolution of odds in race no. 2, which
may play an important role in revealing relevant information.
Since winning daily double combinations appear to be underbet, one must
suspect that a 'smart money' hypothesis is unlikely to be supported here. This
is indeed the case, when we compute the Dk -Lk differences for all daily
double pairs and group them as we did for exactas. The comparison between
the groups suggests the opposite of the 'smart money' hypothesis; that is, the
mean Dk - Lk is smaller for losing pairs than for winning pairs.10
It is interesting to consider why the analysis of exacta and daily double betting
suggests diametrically opposed conclusions with respect to the 'smart money'
hypothesis. We suggest the following answer, based on considerations of
signalling.
If there are n, horses running in the first and n2 in the second race of a
daily double pair, the total number of potential daily double bets is n1n2. As
soon as the first race is over, the number of possible winning bets is reduced
to n2, since the winner of the first race is now known. Racetracks customarily
display, before the beginning of the second race, 'will pay' amounts-i.e., the
sums that would be paid to ticket-holders on each of the still-possible winning
combinations. These figures contain potentially important information.
Suppose that insider information has suggested to some bettors that horse
il in race 1 and horse 12 in race 2 are likely to win. The combination i1J-2 will
then be heavily bet by insiders in the daily double if, as we conjecture, they
attempt to hide their actions from the betting crowd. After the first race is
over, however, the 'will pay' figure allow the public at large to observe that
horse 12 in race 2 has been heavily bet in the daily double relative to the odds
that are evolving in the straight win betting for race 2. The signal that insiders
have hidden in the daily double betting is now revealed.
Racetrack bettors can no longer wager on the daily double; but they can
follow the signal that has emerged from the 'will pay' numbers by backing
horse 12 to win in race 2. Such betting will reduce the odds and winning payoff
to 12, and also will reduce the profitability of a parlay on combination i1j2.
In effect, bettors believe the 'smart money' hypothesis that we have suggested
above, and act accordingly.
There is some evidence that the following type of behaviour exists. Let f
denote the payoff to horse j in straight win betting, mj the payoff to j that
would occur on the basis of morning line odds, and zj some appropriate
measure of the payoff to j in daily double betting relative to the payoff based
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
296 ECONOMICA [AUGUST
on morning line odds such as their ratio. Since the morning line represents a
professional's estimate of the intrinsic quality of a horse, it will be positively
related to the final odds. If imitative (following) behaviour does occur, then
a relatively low z; will induce a relatively low value offj. Thus, in the regression
equation
fj=-16-204 + 2d157mj+10470zj
(-20.503) (35 750) (27-000)
V. CONCLUSION
Are pari mutuel markets for exotic bets efficient? And is inside information
('smart money') a verifiable source of any observed inefficiency? Our analysis
of two types of exotic bets, exactas and daily doubles, yields the following
conclusions. (1) Both of these betting markets exhibit inefficiency in the sense
that their payoffs are not statistically the same as the payoffs to the analogous
bets on individual horses; in addition, the mean-variance locus of exacta
combinations is inconsistent with efficiency for risk-loving bettors. (2) The
differences in payoffs support the 'smart money' hypothesis for exacta betting,
but not for daily double betting.'1
ACKNOWLEDGMENTS
We are grateful to Bruce H. Garland, deputy director of the New Jersey Racing
Commission, Joseph J. Malan, director of mutuels at the Meadowlands Racetrack, and
Joe LaVista, New Jersey State auditor, for their help in providing the data, to the Sloan
Foundation and the National Science Foundation for financial support, and to two
referees for useful comments.
NOTES
1. The track take is the amount taken out of each bet by the racetrack to cover taxes, expenses
and profits, plus the amount ('breakage') resulting from rounding payoffs down to the nearest
10 or 20 cents.
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
1987] EFFICIENCY AND PROFITABILITY 297
2. The payoffs to these bets are more difficult to calculate. The general rule is: first, the original
bets of all successful place (or show) bettors are refunded; then the remaining portion of the
place (or show) betting pool (net of the track take) is divided equally between the two (in the
case of show bets, three) groups of individuals holding successful place (show) tickets. It
follows immediately that the reward to holders of a place ticket on horse i depends on which
other horse i also places. For the exact formulas, which also take breakage into account, see
Ziemba and Hausch (1984).
3. The takes are approximately 0-18 for win, place and show bets, and somewhat over 0-19 for
exactas at the Meadowlands Racetrack.
4. 'Probable payoff' numbers are usually displayed sequentially at certain locations. If the typical
race contains ten entrants, a bettor searching for market information about a two-horse exacta
must view and 'process' 90 possibilities.
5. If W, is the amount bet on horse i, the subjective probability that i will win is Wi/ W. Since
the odds on horse i are defined as Di = W( - t)/ W-1, the subjective probability=
(1- t)/(1 + Di) and is monotone in Di. See Asch, Malkiel and Quandt (1982).
6. We observe in the tables below that rates of return to exacta combinations in various payoff
classes exhibit no systematic tendency. This is consistent with a failure of bettors to distinguish
between high- and low-probability combinations (see table). The prevalence of negative returns
raises the question of whether there are any 'strategies' that will yield positive profit in betting
on exactas. Using the logit model of Asch, Malkiel, Quandt (1984) to predict winners and
runners-up, and simulating various 'reasonable' betting strategies, indicates that on the average
profits are significantly negative.
7. It is interesting to note that this regression is reasonably similar to that obtained by Fabricand
(1965) on the basis of a completely different sample.
8. The simplest intuition for this is as follows: O0 is the probability that i wins; p /(1 -pi) is the
conditional probability that j wins in a race from which i is missing. If the two events are
independent, the product is the required probability.
9. The difference is significant at the 0 05 level. It is almost precisely accounted for by the fact
that daily double bettors pay a single track take of approximately 19 per cent, whereas parlay
bettors pay two takes of about 18 per cent. When the parlay payoffs are adjusted as if parlay
bettors paid a single 'daily double take' (multiply each Lk by 0-8091 and divide by 0.82032),
the mean Lk rises to $49 75, and is no longer significantly different from the mean Dk. It is
as if daily double bettors did not take into account that only one (slightly higher) take is
assessed against their bets.
10. We may also consider the issue of daily double profitability by predicting the winner of each
component of a daily double race pair on the basis of our logit model, and simulating a betting
strategy. The result of this simulation for 122 daily double pairs is an average rate of return
of -0-059, substantially better than the loss implied by the track take, but not profitable.
Overall, we find inefficiency and lack of profitability, as in the case of exacta betting; but we
do not confirm the 'smart money' hypothesis.
11. We have reported elsewhere (Asch, Malkiel and Quandt, 1982, 1984) some evidence suggesting
that 'smart money' may show up in the form of late betting in the 'win bet market'.
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms
298 ECONOMICA [AUGUST
REFERENCES
ALl, MUKTAR M. (1979). Some evidence of the efficiency of a speculative market. Econometrica,
47, 387-92.
ASCH, PETER, MALKIEL, BURTON G. and QUANDT, RICHARD E. (1982). Racetrack betting
and informed behavior. Journal of Financial Economics, 10, 187-94.
(1984). Market efficiency in racetrack betting. Journal of Business, 57, 165-75.
CRAFTs, N. F. R. (1985). Some evidence of insider knowledge in horse race betting in Britain.
Economica, 52, 295-304.
DE BONDT, W. F. H. and THALER, R. (1985). Does the stock market overreact? Journal of Finance,
40, 793-805.
DOWIE, J. (1976). On the efficiency and equity of betting markets. Economica, 43, 139-50.
FABRICAND, BURTON F. (1965). Horse Sense. New York: David McKay.
FIGLEWSKI, STEPHEN (1979). Subjective information and market efficiency in a betting market.
Journal of Political Economy, 87, 75-88.
HARVILLE, D. A. (1973). Assigning probabilities to outcomes of multi-entry competition. Journal
of the American Statistical Association, 68, 312-16.
HAUSCH, DONALD B., ZIEMBA, WILLIAM T. and RUBENSTEIN, MARK (1981). Efficiency of
the market of racetrack betting. Management Science, 17, 1435-52.
KAHNEMAN, DANIEL and TVERSKY, AMOS (1982). Intuitive prediction: biases and corrective
procedures. In Daniel Kahneman, Paul Slovic and Amos Tversky, Judgment Under Uncertainty;
Heuristics and Biases. Cambridge: Cambridge University Press.
RoSETT, RICHARD N. (1965). Gambling and rationality. Journal of Political Economy, 73, 595-607.
(1971). Weak experimental verification of the expected utility hypothesis. Review of Economic
Studies, 38, 481-92.
SNYDER, WAYNE N. (1978). Horse racing: testing the efficient markets model. Journal of Finance,
22, 1109-18.
ZIEMBA, WILLIAM T. and HAUSCH, DONALD B. (1984). Beat the Racetrack. New York: Harcourt,
Brace, Jovanovich.
ZUBER, RICHARD A., GANDAR, JOHN M. and BOWERS, BENNY D. (1985). Beating the spread:
testing the efficiency of the gambling market for National Football League games. Journal of
Political Economy, 93, 800-6.
This content downloaded from 132.77.150.148 on Tue, 17 May 2016 21:31:28 UTC
All use subject to http://about.jstor.org/terms