Central Limit Theorem
The central limit theorem says averages of many independent values tend toward a normal distribution. Learn what it means for trading statistics and when it fails.
The central limit theorem (CLT) is one of the most important results in statistics. It says that if you add up or average many independent random values, the result tends toward a normal (bell shaped) distribution, even if the individual values are not normal. This is why the normal distribution appears so often in statistics and why standard errors and confidence intervals work for averages. For traders, the CLT explains when statistical tests of strategy returns are reasonable and when they can mislead.
The statement#
If X1, X2, ..., Xn are independent random variables with the same distribution, mean μ and finite standard deviation σ, then as n grows:
sample mean ≈ Normal(μ, σ / √n)The spread of the sample mean shrinks with √n, which is the standard error. See Sampling and Standard Error.
An intuitive example#
Why it matters for traders#
| Application | How the CLT helps |
|---|---|
| Confidence intervals for average returns | Lets us use normal based intervals. See Confidence Intervals |
| Hypothesis tests on strategy returns | t tests and similar rely on approximate normality of means. See Hypothesis Testing and P-Values |
| Portfolio returns | Sums of many positions' returns tend toward normal, if positions are not too correlated |
| Monte Carlo results | Averages of simulations converge to normal shapes |
Returns over longer horizons#
Daily returns are fat tailed, but monthly or annual returns, which are sums of many daily returns, look somewhat more normal. This is partly the CLT. However, because daily returns are not fully independent (volatility clusters) and have very heavy tails, the convergence is slower and incomplete. See Fat Tails and GARCH.
When the CLT fails or works poorly#
| Situation | Why |
|---|---|
| Very fat tails | If variance is infinite or extremely large, averages converge slowly or not to a normal shape |
| Strong dependence | Correlated observations reduce the effective sample size. See Autocorrelation and Partial Autocorrelation |
| Small samples | The approximation needs enough observations, especially with skewed data |
| Changing distributions | If the data generating process shifts, there is no single mean to converge to. See Structural Breaks and Regime Changes |
| Rare extreme events | Strategies with occasional huge losses (short options) can look safe in samples that miss the rare event |
Practical takeaways#
- The CLT justifies standard errors for averages, but only with enough independent observations.
- Be careful with skewed and fat tailed strategies; use larger samples or bootstrap methods. See Bootstrap and Permutation Tests.
- Account for correlation between observations.
- Do not assume individual returns are normal just because averages are.
Frequently asked questions#
What is the central limit theorem?#
A result stating that the average of many independent random values with finite variance tends toward a normal distribution as the number of values grows.
Why is the central limit theorem important in trading?#
It justifies using normal based confidence intervals and tests for average returns, which help judge whether a strategy's performance is real.
Does the central limit theorem make returns normal?#
No. It applies to averages and sums of many independent values; individual returns remain fat tailed, and dependence slows convergence.
Next, learn to express uncertainty as a range in Confidence Intervals.
3 quick questions on this lesson. Get them all right to finish it.
Turn on JavaScript to take the quiz.