Law of large numbers vs central limit theorem

By Jude Wallis · Published July 28, 2026

The law of large numbers says that as trials pile up, the sample average settles toward a single value. The central limit theorem says that as the sample size grows, the sampling distribution of the mean becomes approximately normal. One is about where the average lands, the other about its shape.

AP Statistics: Unit 2 (topics 2.3, 2.12). In the Fall 2026 AP Statistics course, the law of large numbers appears in topic 2.3 of Unit 2, where probability is a long-run relative frequency, and the central limit theorem is topic 2.12, where the sampling distribution of a sample mean is approximately normal. The condition n at least 30 for a non-normal population is stated with sampling distributions for means in Unit 4.

Law of large numbers vs CLT: the short answer

Both results describe what happens as your sample gets large, and both are easy to blur together because both feature a growing $n$ . They answer different questions.

The law of large numbers is about a value settling down. As the number of independent trials increases, the long-run relative frequency of an event (or the average of the results) gets closer and closer to a single fixed number. It tells you where the average is headed.

The central limit theorem is about a shape appearing. As the sample size increases, the sampling distribution of the mean of a random sample can be approximated by a normal distribution. It tells you the shape and spread of how that average bounces around from sample to sample. One promises the average converges; the other describes the bell curve the average follows on the way.

What the law of large numbers says

The law of large numbers states that for independent trials, as the number of trials increases, the long-run relative frequency of an outcome or event gets closer and closer to a single value. That single value is the true probability of the event, which is exactly why probability is defined as a long-run relative frequency.

Flip a fair coin. Early on, the fraction of heads can wander well away from 0.5, but after thousands of flips it hugs 0.5 tightly. The law does not say the counts of heads and tails even out; it says the proportion settles toward its true value as trials pile up.

The same idea extends from proportions to averages: as you take more observations, the sample mean $\bar{x}$ (read x-bar) approaches the expected value, the true mean of the distribution. A relative frequency is just the average of a bunch of 0s and 1s, so the coin case is a special case of the mean case. The law of large numbers is why a bigger sample gives an estimate you can trust to be near the truth. It says nothing about shape.

What the central limit theorem says

The central limit theorem (CLT) states that the sampling distribution of a mean of a random sample has a shape that can be approximated by a normal distribution, and the larger the sample is, the better that approximation gets. The remarkable part is that this holds no matter what shape the population has, as long as the sample is large enough.

Start with a population that is skewed or lumpy, take a random sample, and record its mean. Do that over and over and the collection of sample means piles up into an approximately normal, bell-shaped distribution centered at the population mean, with standard deviation $\frac{\sigma}{\sqrt{n}}$ , where $\sigma$ (sigma) is the population standard deviation and $n$ is the sample size. As a rule of thumb, a sample size of $n \geq 30$ makes the approximation work even for non-normal populations.

The CLT is the engine behind nearly every later inference procedure, because it lets you use the normal model for the mean even when the raw data are not normal. For the full mechanics, see the central limit theorem guide and sampling distributions explained.

The differences side by side

Feature	Law of large numbers	Central limit theorem
What it describes	Where the average settles	The shape of the average's distribution
As n grows	The mean approaches the true value	The sampling distribution nears normal
Output	A single number (the truth)	A whole distribution (normal)
Says the data are normal	No	No, it is about the mean, not the data
Main use	Justifies estimates near the truth	Justifies the normal model for inference
AP location	Unit 2, with probability	Unit 2, with sampling distributions

Hold the two apart with one line: the law of large numbers is about the destination of the average, the central limit theorem is about the shape of the path it takes.

When to use which

Lean on the law of large numbers when the question is about convergence to a value: why a long-run relative frequency estimates a probability, why a casino profits over millions of bets, or why a larger sample gives a mean closer to the population mean. It is the justification that estimates get more accurate as data accumulate.

Lean on the central limit theorem when the question is about the distribution of a sample mean: computing a probability for $\bar{x}$ , or checking that inference for a mean is valid because the sampling distribution is approximately normal. The CLT is what you cite to model $\bar{x}$ as normal with center $\mu$ (the population mean) and spread $\frac{\sigma}{\sqrt{n}}$ . You can watch both happen at once in the sampling distribution and CLT interactive: the average steadies (law of large numbers) while its histogram fills in a bell (central limit theorem).

The classic mix-up and how to avoid it

The most common error is thinking the central limit theorem says your data become normal as the sample grows. It does not. The population keeps whatever shape it has; it is the sampling distribution of the mean, the distribution of $\bar{x}$ over many samples, that becomes normal. A large sample from a skewed population is still skewed.

A second error swaps the two entirely, using the law of large numbers to talk about shape or the CLT to talk about convergence. Keep them straight by asking what the theorem delivers: a single value the average approaches (law of large numbers) or a bell-shaped distribution for that average (central limit theorem). A third slip is the gambler's fallacy, expecting past results to be corrected; the law of large numbers promises the proportion converges through more trials, not that earlier outcomes get balanced out.

Rolling a fair die: convergence and shape together

A fair die shows the numbers 1 through 6, each with probability one-sixth, giving a population mean of 3.5. First find the population standard deviation. Then use the central limit theorem to describe the distribution of the mean of 25 rolls, and say what the law of large numbers predicts as you keep rolling.

State the population mean: $\mu = 3.5$ . Find each squared deviation from it: $(1 - 3.5)^2 = 6.25$ , $(2 - 3.5)^2 = 2.25$ , $(3 - 3.5)^2 = 0.25$ , $(4 - 3.5)^2 = 0.25$ , $(5 - 3.5)^2 = 2.25$ , $(6 - 3.5)^2 = 6.25$ .
Add the squared deviations: $6.25 + 2.25 + 0.25 + 0.25 + 2.25 + 6.25 = 17.5$ .
Multiply each by its probability $\frac{1}{6}$ , which here is the same as dividing the sum by 6: variance $= \frac{17.5}{6} \approx 2.9167$ .
Take the square root for the population standard deviation: $\sigma = \sqrt{2.9167} \approx 1.708$ . Check: $1.708^2 = 2.917$ , which rounds back, so it is right.
Apply the central limit theorem for $n = 25$ . The sampling distribution of the mean is approximately normal, centered at $\mu = 3.5$ , with standard deviation $\frac{\sigma}{\sqrt{n}} = \frac{1.708}{\sqrt{25}} = \frac{1.708}{5} \approx 0.342$ .
Apply the law of large numbers. As you roll more and more times, the running average of your rolls gets closer and closer to the true mean 3.5, no matter how it wandered early on.

The population standard deviation is about 1.708. By the central limit theorem, the mean of 25 rolls is approximately normal with center 3.5 and standard deviation about 0.342. By the law of large numbers, the running average of the rolls converges to 3.5 as the number of rolls grows. The CLT gives the shape and spread; the law of large numbers gives the destination.

Frequently asked questions

Does the central limit theorem say my data become normal?

No. The central limit theorem is about the sampling distribution of the mean, not the individual data. A large sample from a skewed population stays skewed; what turns normal is the distribution of the sample mean computed over many samples of that size.

How large does the sample need to be for the CLT?

If the population is already normal, the sampling distribution of the mean is normal for any sample size. If it is not normal, a sample size of about 30 or more usually makes the approximation good, and a strongly skewed population may need more than 30.

Are the law of large numbers and the CLT the same thing?

No. The law of large numbers says the average approaches a single fixed value as trials increase. The central limit theorem describes the shape and spread of that average's distribution across samples. One is about convergence to a value, the other about a normal shape.