Odds and Probabilities


Odds and probabilities are important (and related) concepts in sports analytics. They appear in several contexts, such as (the related ones of) prediction, bookmaking, and betting.

Despite their importance and wide usage, there is [1] common confusion about these terms. This results in (and is compounded by) inaccurate statistical language in the dissemination of results. Information reported also often neglects subtle but important details. Understanding these issues is necessary to describe (and understand) the likeliness of events.

In this Article, odds and probabilities and their relationship are considered in detail. As an application, book odds are discussed.

Odds and Probabilities

Odds and probabilities are both statistical terms that are used to describe the likeliness of an event. How they each describe this, however, are different.


In statistics, odds are defined [2] from two ways: odds in favor of (or odds on) and odds against an event A. In both ways, odds describe the likeliness of A.
One definition of odds is mathematically summarized by the following. Consider that for event A that there are N total outcomes and k of which favor that event (and, hence, N - k of which do not). Based on this definition, the odds are therefore

    \begin{eqnarray*} \text{odds in favor of A} = k : N - k \\ \text{odds against A} = {N - k} : k \end{eqnarray*}

where : is used to denote against.


Probability can also be used to describe the likeliness of an event.

The probability of event A, P(\text{A}), is expressed as

(1)   \begin{equation*} P(\text{A}) = \frac{k}{N} ~~~ ; \end{equation*}

and that not occurring is

(2)   \begin{equation*} P(\text{A}^\text{c}) = \frac{N - k}{N} \end{equation*}

where \text{A}^\text{c} is the complement of A.

Note that an important equality implied by Eqs. (1) and (2) is

(3)   \begin{equation*} P(\text{A}) + P(\text{A}^\text{c}) = 1 ~~~ . \end{equation*}

The Relationship Between Odds and Probabilities

Odds and probabilities are clearly related (as explicitly stated above).

In fact, they can straightforwardly be defined in terms of each other:

    \begin{equation*} \text{odds in favor of A} = \frac{P(\text{A})}{1 - P(\text{A})} ~~~ ; \end{equation*}

all other definitions (odds against, or definitions of probabilities) follow similarly.

Book Odds

In the context of sports betting, the consideration of (book) odds must be made with care. This is because there is unknown information (to the bettor) that the bookmaker has.

In making a book (the mathematics of bookmaking will be considered in a separate article), a bookmaker will reduce (from the true) odds, in order to ensure an expected profit.

Consider an example of a baseball game with moneyline odds reported as {+130} [away (A)]and {-150} [home (H)]. Converting these “odds” to “probabilities” gives

    \begin{eqnarray*} P'(\text{A}) = \frac{100}{100 + 130} = 0.4348 \\ P'(\text{H}) = \frac{150}{150 + 100} = 0.6 \end{eqnarray*}

(where quotations ” and primes ‘ have been added, for the reasons to follow). Adding these “probabilities” gives

    \begin{equation*} P'(\text{A}) + P'(\text{H}) = 1.0348 ~~~ . \end{equation*}

As this sums to more than one, these “probabilities” cannot be considered as (true) probabilities [and satisfy Eq. (3)].

The amount by which a book exceeds a “probability” of 1 (and which represents the bookmaker’s expected profit) is known as the overround, bookmaker margin, or the vigorish (vig).

Without additional information (about the odds reduction), it is impossible to determine the bookmaker’s (true) probability estimates.

Assume, however, that the bookmaker does not want to accept a negative expected value for any bet. It can therefore be concluded that

  • 0.4348 and 0.6 are upper bounds to the (true) probability P(\text{A}) and P(\textH}), respectively;
  • 0.4348 - 0.0348 = 0.4 and 0.5652 lower bounds.

Precise estimates can only be made by further assuming (or knowing) the model for reducing. Consider the simplest one, which preserves the relative “probabilities”. In this case,

    \begin{eqnarray*} P(\text{A}) o = P'(\text{A}) \\ P(\text{H}) o = P'(\text{H}) \end{eqnarray*}

where o is the overround. Using Eq. (3), the above equation can be used to solve for o, and then P(\text{A}) and P(\text{H}). For the example above, this results in

    \begin{eqnarray*} P(\text{A}) = 0.4202 \\ P(\text{H}) = 0.5798 ~~~ . \end{eqnarray*}


[1] L. V. Fulton, F. A. Mendez, N. D. Bastian, and R. M. Muzal, “Confusion Between Odds and Probability, a Pandemic?” Journal of Statistics Education 20 (3) (2012)

[2] D. A. Berry and B. W. Lindgren, Statistics: Theory and Methods, 2nd ed., p. 15 (Duxbury Press, 1996)


About Author

statshacker is an Assistant Professor of Physics and Astronomy at a well-known state university. His research interests involve the development and application of concepts and techniques from the emerging field of data science to study large data sets. Outside of academic research, he is particularly interested in such data sets that arise in sports and finance. Contact: statshacker@statshacker.com

Leave A Reply