Almost-Sure Convergence¶

Definition¶

Let $X_1,X_2,\ldots$ and $X$ be random variables defined on one probability space, that is, one sample space with one probability function. The sequence $X_n$ converges almost surely to $X$ if

\begin{aligned} \mathbb{P}(\lim_{n\to\infty}X_n=X)=1. \end{aligned}

The set of outcomes on which $X_n$ does not converge to $X$ need not be empty, but it must have probability zero. This is written as $X_n\xrightarrow{a.s.} X$ $X$ . One also says that $X_n$ converges to $X$ with probability one, abbreviated w.p.1.

For each outcome outside the exceptional null set, the Limit of a Sequence is taken for the numerical values $X_n(\omega)$ ; the probability-one condition refers to the set of outcomes where that limit equals $X(\omega)$ .

Interpretation¶

Almost-sure convergence looks at one outcome at a time. For a fixed outcome, $X_1,X_2,\ldots$ is an ordinary sequence of numbers, the sample path of that outcome, and the question is whether it tends to the value of $X$ . Outside a set of outcomes of probability zero, the answer must be yes: this is pointwise convergence on the sample space, with a negligible exceptional set allowed.

Three rows of dots for n = 1 to 20, one row per outcome of U. Each path X n equals 1 for the first few n and then drops to 0 for good, sooner for larger U.

The figure follows $X_n=\indicator_{\{U<1/n\}}$ from the second Example for the three outcomes $U=0.4$ $0.4$ , $0.15$ and $0.06$ . Each path equals $1$ while $n<1/U$ and $0$ from then on: the last $1$ occurs at $n=2$ $2$ , $6$ and $16$ respectively. Smaller $U$ drops later, but every outcome with $U>0$ eventually reaches $0$ and stays there. The only exception, $U=0$ $0$ , has probability zero, so $X_n\xrightarrow{a.s.}0$ $0$ .

Almost-sure convergence is about whole paths: each simulated path of $\bar X_n$ settles at $\mu$ . The red dot marks the last time a path leaves the band.

The word "almost" allows that exceptional set; see almost surely. The probability in the definition is taken once, for the set of outcomes whose whole path converges, not separately for each $n$ . That is the difference with convergence in probability, which only asks that $\mathbb{P}(|X_n-X|>\varepsilon)$ be small for each large $n$ and still allows the paths to keep making occasional large errors forever. Almost-sure convergence is the stronger of the two, and it is the mode of convergence in the strong law of large numbers.

Key Results¶

Almost-sure convergence implies convergence in probability.
Changing any of the variables on a common event of probability zero does not change almost-sure convergence.

Examples¶

If $X$ is finite almost surely and $X_n=X/n$ $X / n$ , then $X_n\to0$ on every outcome where $X$ is finite, so convergence is almost sure.
For a single $U\sim\operatorname{Unif}(0,1)$ and $X_n=\indicator_{\{U<1/n\}}$ $𝟙 1_{{U < 1/ n}}$ , each outcome with $U>0$ eventually has $X_n=0$ $0$ , so $X_n\to0$ almost surely.

Prerequisites¶

Convergence in Probability

Sources¶

Nicky van Foreest · Probability Distributions · Week 7

Aliases: almost sure convergence, convergence with probability one, convergence almost surely, convergence with probability 1 Keywords: a.s. convergence Symbols: epsilon, infinity, mu, omega

Last updated: 20 October 2026 03:02