# Infinite coin flips, etc.

• mag487
In summary, the conversation revolves around the question of whether a certain triplet (X, A, P) satisfies the criteria of a probability space, where X is uncountable and A is the entire power set of X. The conversation explores different constructions and arguments, but ultimately concludes that the problem is not easily solved and may even be unprovable from the axioms of set theory.

#### mag487

Hi, I'm new here. I'm trying to teach myself measure theory and probability and recently wanted to find an example of a probability space (X, A, P) where X is uncountable and the sigma-algebra A is the entire power set of X. Here was my idea: let X be the set of all strings c1c2c3... for $$c_{i} \in \{ 0, 1\}$$, thought of as an infinite sequence of coin flips (with 1 corresponding to heads and 0 to tails), and A = X's power set. For x in X, define the n-restriction xn of x as the string of the first n digits of x. Furthermore, for S in A, define Sn := {xn : x is in S}. Finally, define $$P (S) := lim_{n\rightarrow\infty} card(S_{n})/2^{n}$$, i.e., the probability of S is the long-term proportion of the initial segments of the members of S to all possible initial segments.

I wanted to prove that (X, A, P) is a probability space. I thought I found a (rather torturous) proof of this claim, but got some initial results that led me to think my proof must have gone wrong somewhere. But before I type out my whole line of reasoning, does anyone already know for sure if the above triplet satisfies the criteria of a probability space? Thanks in advance!

Last edited:
The construction of the Vitali set shows that there's no measure m defined on all subsets of R satisfying sigma-additivity, m(T(S)) = m(S) for any translation T, and m([0,1]) = 1. I think you can mimic that construction for your set to show that it too won't work.

Regard X as functions from N to {0,1}. Define a relation ~ on X by:

x~y iff x and y disagree on a finite subset

It's clear that ~ is an equivalence relation. Let [x] denote the equivalence class of x. Invoking the Axiom of Choice, there exists a set R consisting of exactly one representative from each equivalence class. If S is a finite subset of N, x in X, define xS as follows:

xS(n) = x(n) if n is not in S
xS(n) = 1 - x(n) if n is in S

Define RS = {xS : x in R}

It's clear that if S and T are distinct finite subsets of N, then RS and RT are disjoint, and moreover

$$\bigcup _{S \in [\mathbb{N}]^{<\omega}}R_S = X$$

where $[\mathbb{N}]^{<\omega}$ is the set of finite subsets of N. The union above is a countable union since there are countably many finite subsets of N, and it's a union of disjoint sets. So

$$1 = P(X) = P\left (\bigcup _{S \in [\mathbb{N}]^{<\omega}}R_S\right ) = \sum _{S \in [\mathbb{N}]^{<\omega}} P(R_S)$$

It's not hard to see that P(RS) = P(R) for all finite S. So the right hand side is P(R) added to itself countably many times. If P(R) is 0, then the right side is 0, contradicting the fact that the left side is 1. If P(R) is non-zero, then the right side is infinity, contradicting the fact that the left side is 1.

So P isn't defined at R.

Aha, here's the problem.

The key feature of this probability measure is that it's insensitive to what happens 'at infinity'. So, let's do something weird there.

Let S be the set of all eventually constant sequences.

What is P(S)? P(X-S)?

Actually, the problem you're trying to solve, mag487 is not remotely easy. Look up measurable cardinals. I'm just starting to understand this area of set theory, but the existence of "a probability space (X, A, P) where X is uncountable and the sigma-algebra A is the entire power set of X" is something that's either:

i) not known to be provable from ZFC and not known to be disprovable from ZFC,
ii) known to be neither provable nor disprovable from ZFC, or
iii) something like the above two possibilities.

## 1. What is the concept of infinite coin flips?

The concept of infinite coin flips refers to an infinite sequence of coin tosses where each toss has an equal chance of landing on either heads or tails. This means that the probability of getting heads or tails is always 50%, regardless of how many times the coin is flipped.

## 2. Can infinite coin flips result in a specific pattern?

No, since each flip has an equal chance of landing on either heads or tails, there is no way to predict or control the outcome of an infinite number of coin flips. It is possible for a specific pattern to occur, but it is not guaranteed.

## 3. Are there any real-life applications of infinite coin flips?

Infinite coin flips can be used to model and understand probability in different fields such as statistics, economics, and physics. It can also be used in gambling and game theory to analyze and predict outcomes.

## 4. Is it possible for an infinite coin flip sequence to have an equal number of heads and tails?

Yes, it is possible for an infinite coin flip sequence to have an equal number of heads and tails. However, the probability of this happening is extremely low and becomes even lower as the number of flips increases.

## 5. How does the concept of infinite coin flips relate to the law of large numbers?

The law of large numbers states that as the number of trials or experiments increases, the observed results will approach the expected or theoretical probability. In the case of infinite coin flips, this means that as the number of flips approaches infinity, the observed proportion of heads and tails will approach the theoretical 50/50 probability.