You are on page 1of 10

Generic and specific formulas for the calculation of probability to

predict independent and repeated random outcomes

Panagiotis Tsikogiannopoulos
Physicist - Graduate of Hellenic Open University
e-mail: pantsik@yahoo.gr

Abstract

The following formulas consist a generalised guide on the calculation of the probability for any
number of participants (players) to predict any number of independent random outcomes with each
having the same known probability to occur. This process is known as Bernoulli Trials and the
resulting probability formulas are based on the Binomial Distribution. We can require these
outcomes to occur in exactly, at least and at most any number of times and also we can require the
players to predict them correctly in exactly, at least and at most any number of times.

Keywords: probability formulas, prediction, independent random outcomes, Bernoulli Trials,


binomial distribution

AMS 2000 Subject Classification: 60-01, 60C05

Introduction

Let a random procedure produce a number of ε outcomes of equal probability which are repeated
by a number of κ independent times. This sequence has to be predicted by n participants (players)
or equally by one player with n tries.
The calculation formulas of all the probabilities that can be related to the above procedure are given
in the following sections. For example, we can calculate the probability to predict at least x of κ
outcomes with at most y of n tries.
If more than one outcome is considered correct then we set as ε the quotient of the total outcomes
and the outcomes that are considered correct. For example, if we consider correct prediction for a
die to show a 4 or 5 then ε = 6/2 = 3.

At the end of this paper there is a table for easy reference to the corresponding paragraphs and two
examples of the formulas usage. The formulas proofs follow, which are not entered into the main
text in order to keep its character as a guide.
A spreadsheet in Excel where the following probabilities are calculated automatically is given
separately.

 Probability formulas in one try (n=1)

A) The probability to predict exactly x of κ outcomes (1 ≤ x ≤ κ) is

1
Simplifications:

A.1) The probability to predict correctly all the outcomes is given by letting x=κ in pA and
assuming that 0! = 1

A.2) The probability to predict exactly one of κ outcomes is given by letting x=1 in pA :

A.3) The probability to predict none of κ outcomes is given by letting x=0 in pA :

B) The probability to predict at least x of κ outcomes (1 ≤ x ≤ κ) is

Simplifications:

B.1) The probability to predict at least one of κ outcomes is the complement of the probability
to predict none of κ outcomes, thus

B.2) The possibility to predict at least zero of κ outcomes is certain that will occur, thus its
probability is 1. Formally:

In that case, the corresponding probability formulas P1, P2, P3 introduced bellow are not
applied and are becoming

Exception to this rule consists the case we ask the above certainty to occur zero times, thus
we have

2
C) The probability to predict at most x of κ outcomes (1 ≤ x < κ), that is from 0 to x outcomes,
is

Simplifications:

C.1) The probability to predict at most κ of κ outcomes is like asking the probability to predict
from 0 to κ outcomes, which is a certainty, thus its probability is 1. Formally:

In that case, the corresponding probability formulas P1, P2, P3 introduced bellow are not
applied and the formulas are becoming

Exception to this rule consists the case we ask the above certainty to occur zero times, thus
we have

C.2) The probability to predict at most κ-1 of κ outcomes coincides with the probability not to
predict all the outcomes which is the complement of that in section A.1, thus

C.3) The probability to predict at most one of κ outcomes (meaning one or none) is given by
letting x=1 in pC that eventually becomes

 Probability formulas in many tries (n>1)

For the calculation of probabilities P1, P2, P3 used for many tries that are introduced below, we
insert the probability formulas pA, pB, pC we found in Sections A or B or C into the function p(x,κ)
according to the asked outcomes in these tries.

1) The probability to predict the asked outcomes in exactly y of n tries (1 ≤ y ≤ n) is

3
Simplifications:

1.1) The probability to predict all the outcomes in every try is given by inserting pA(κ,κ) of
Section A.1 into P1 and letting y=n, thus we get

1.2) The probability not to predict any outcome in every try is given by inserting pA(0,κ) of
Section A.3 into P1 and letting y=n, thus we get

1.3) The probability to predict at least one outcome in every try is given by inserting pB(1,κ) of
Section B.1 into P1 and letting y=n, thus we get

1.4) The probability to predict all the outcomes in exactly one try is given by inserting pA(κ,κ)
of Section A.1 into P1 and letting y=1, thus we get

1.5) The probability not to predict the asked outcomes in any try is given by letting y=0 into P1:

where in place of p(x,κ) we insert the corresponding probability pA, pB, pC as we


mentioned above.

2) The probability to predict the asked outcomes in at least y of n tries (1 ≤ y ≤ n) is

Simplifications:

4
2.1) The probability to predict all the outcomes in at least one try, is the complement of the
probability not to predict all the outcomes in any try, which is given in Section 1.5. Thus it
is

2.2) The probability not to predict any outcome in at least one try is the complement of the
probability given is Section 1.3.

2.3) The probability to predict the asked outcomes in at least zero tries is a certainty, thus its
probability is 1. Formally:

3) The probability to predict the asked outcomes in at most y of n tries (1 ≤ y ≤ n), that is from 0
to y tries, is

Simplifications:

3.1) The possibility to predict the asked outcomes in at most n of n tries is a certainty, thus
its probability is 1. Formally:

3.2) The probability to predict the asked outcomes in less than all the tries (until zero tries) is
the complement of the probability to predict the same outcomes in every try, thus

3.3) The probability to predict the asked outcomes in at most one try, is the probability to
predict the same outcomes in one try plus the probability to predict the same outcomes in
zero tries, thus

Sections for finding the conclusive probability P

5
exactly at least at most
y of n y of n y of n
exactly
A and 1 A and 2 A and 3
x of κ
at least
B and 1 B and 2 B and 3
x of κ
at most
C and 1 C and 2 C and 3
x of κ

The above table is used for easy reference to the sections of the corresponding probabilities,
depending on the number of correct outcomes (x of κ) and the number of successful tries (y of n)
for these outcomes.

Example 1

Two players are playing heads or tails with a coin[1]. Each time a head comes up Player A gives one
Euro to Player B and each time a tail comes up the opposite happens. The question is what is the
probability for the two players to end up with their starting money if they toss the coin 40 times.
We will define the variables that will be used in the formulas. First we define ε. The possible
outcomes of a coin are two (head or tail) so we set ε = 2. Next we define κ which is the number of
times the procedure will be repeated or the number of outcomes in total, so we set κ = 40. Finally
we define x which is the number of the asked outcomes for which the probability to occur will be
calculated. At the end of the 40 tosses we want each player to have his starting money, so we ask
the coin to be tossed on heads and tails equally, that is 20 heads and 20 tails. Thus, the number of
the asked outcomes (by heads or tails, doesn’t matter) is x = 20. We want this number of outcomes
to come up exactly that number of times in the 40 tosses, so we will insert the above values into the
formula of Section A and we will find that the asked probability is

or 12.5 %.

In this example, the usage of formulas of Sections 1, 2 or 3 was not needed because the asked
probability was referred only to one game of 40 tosses.
If we were asking the probability for Player A to win 6 Euros at the end of the game then the
variables ε and κ would have the same values but we would set x = 26.

Example 2

Let 5 players play roulette in a casino and each of them bets on 4 numbers in every round
independently from the other players’ bets. The casino manager wants to calculate the probability
not to win any of them for the next 3 rounds, so the casino will not pay any money.
We will define again the variables that will be inserted into the formulas, starting with ε. Each
player bets on 4 numbers from a total of 37 (36 and Zero), so we set ε = 37/4 = 9.25 as we
mentioned in Introduction.
Next we will define κ and x of the inner probability which is referred to one player. We set κ = 3
which is the number of rounds to be considered, for it is the number of outcomes in total. We set
x = 1 because if a player predicts at least one number in the next 3 rounds the casino will pay.

6
Finally we define y and n that are referred to all the players. We set n = 5 which is the number of
players. We set y = 0 because we want the terms for the inner probability described above, not to
happen to any of the 5 players, or to happen in exactly 0 of 5 players (or tries).
From the Table we find that for the probability to occur at least x of κ outcomes in exactly y of
n tries, we will refer to Sections B and 1 and in particular to B.1 and 1.5 from which we find that
the asked probability is

or roughly 18 %.

We would conclude to the same result simpler if we asked for the probability to occur exactly 0 of 3
outcomes in exactly 5 of 5 tries and using just the formula of Section 1.2.

Formulas proofs

In all the proofs below, the probability for the correct prediction of every outcome is 1/ε.
We start by studying the structure of probability formulas for a special case and then we generalize
to every case.

Section A generic formula proof

The objective here is to find the probability formula to predict exactly x of κ outcomes.
Suppose we want to predict 2 of 3 outcomes. In order for this to happen, the first two of three
outcomes or the first and the third or the last two must be predicted. Thus, we must add their three
partial probabilities. In each of them the correct outcome must be predicted twice and not to be
predicted the third time, so each partial probability is the product of the probabilities 1/ε, 1/ε and
(ε–1)/ε. Formally:

The exponent of the second factor indicates the number of correct outcomes (x). The exponent of
the third factor indicates the number of outcomes that were not predicted correctly which is κ – x.
The first factor (number 3) is the number of ways we can distribute 2 identical elements (the factors
1/ε) in 3 places (the total number of factors).
The general form of this factor for distributing x elements in κ places has as follows:
If the elements were not identical, so we could discriminate them between each other, then the first
element could occupy each one of the available places, the second element could occupy each of the
remaining places, and so on, until the last element that could occupy any of the remaining places
that were not occupied by another element and the total number of distributions would be the
product of all these ways. So, if Δμο is the number of non-identical element distributions, we can
formalize it by:

7
However, because of the elements are identical, with the above formula each distribution is repeated
as many times as the number of combinations that the same number of non-identical elements can
be distributed in a number of places equal to the number of these elements. In this case, the first
element can occupy each of the available places, the second one each of the remaining places, and
so on, until the last element that can occupy only one place. Again, the number of these
distributions will be the product of all these ways. So, if Δμο' is the number of distributions that
have equal number of non-identical elements and places, it can be formally expressed as

We can now calculate the number Δο of ways that x identical elements can be distributed in κ
places, which is

The above expression is usually referred as[2]

is called binomial coefficient and although it has more factors that we concluded, has also the
advantage that it gives the correct result in the special case that x=0, assuming that 0! = 1.
Using all that have been mentioned above, we conclude that the generalized formula of the
probability to predict exactly x of κ outcomes is

which is also known as binomial probability distribution for it is composed by distributing two
possible outcomes, that of correct and incorrect prediction.

Section B generic formula proof

The objective here is to find the probability formula to predict at least x of κ outcomes.
Suppose we want to predict at least 1 of 3 outcomes. This case includes the structure of the terms
presented in Section A formula proof for predicting 2 of 3 outcomes but has the additional terms to
predict 1 of 3 and 3 of 3 outcomes. So, it is

8
The exponent of the first factor in the first square bracket indicates the number of correct
predictions in this bracket. The exponent of the second factor in the first square bracket indicates
the number of the remaining outcomes that were not predicted correctly. The exponent of the first
factor in the second square bracket is the previous number plus an additional correct prediction
which is subtracted from the exponent of the second factor in the second square bracket, so that the
sum of the two exponents remains constant and equal to the total outcomes (κ). A similar variation
is repeated for the third square bracket.
The numeric factors at the beginning of each term indicates the number of ways that the factors 1/ε
can be distributed in the places given from the number of outcomes, exactly as calculated in
Section A formula proof.
As we mentioned above, the number of correct predictions in each bracket is increased by 1. If the
symbolism Δο(2 of 3) means the number of distributions of 2 identical elements in 3 places then we
can generalize the above formula as

It becomes clear that the above terms, independent from their number, can be included in one
formula by inserting a series and replacing Δο’s with a binomial coefficient like the one we found
in Section A formula proof. So, the concluding formula is

Section C generic formula proof

The objective here is to find the probability formula to predict at most x of κ outcomes.
The process for this proof is the same with that of Section B generic formula proof, with only the
following difference:
For the probability to predict at least x of κ outcomes we inserted a series whose variable s was
scanning all the outcomes from x, which is the least number of the asked outcomes, to κ, which is
the most number. Here that we ask the probability to predict at most x of κ outcomes, the variable
s of the series should have to scan the outcomes from 0 to x.
So, the respective formula is becoming:

9
Sections 1, 2 and 3 generic formulas proof

Sections 1, 2 and 3 are referred to the probability of predicting the asked outcomes in exactly, at
least and at most y of n tries.
The same methods with the corresponding Sections A, B and C are applied (1 with A, 2 with B and
3 with C) with the only difference being that now we don’t insert the probability 1/ε as a correct
outcome but p(x,κ) which is the probability the asked outcomes to be predicted correctly.
Accordingly, instead of the probability 1–(1/ε) of an incorrect outcome, we use the probability
1– p(x,κ). Also, we change the variables from x to y and from κ to n.
Thus, for example, the probability to predict the asked outcomes in exactly 2 of 3 tries is

which, with the generalizations mentioned in Section A formula proof, is expressed as

References
1
Charles M. Grinstead, J. Laurie Snell, Introduction to Probability, p.15, Example 1.4 (American
Mathematical Society, 1997)
http://www.dartmouth.edu/~chance/teaching_aids/books_articles/probability_book/amsbook.mac.pdf

2
Eric W. Weisstein, Binomial Coefficient, (MathWorld -- A Wolfram Web Resource)
http://mathworld.wolfram.com/BinomialCoefficient.html

10

You might also like