Martingales

Martingale Central Limit Theorem 

The martingale central limit theorem (MCLT) is an extension of the classical central limit theorem to sequences of random variables that exhibit a certain kind of dependency structure—namely, the martingale property.

 

A martingale is a sequence of random variables $\{M_n\}$ where, given the past information, the best prediction for the next value is the current value. In mathematical terms,

$$E[M_{n+1}∣\mathcal{F}_n]=Mn,$$

​where $\mathcal{F}_n$ is the information (or sigma-algebra) available up to time $n$. The increments or differences $X_{n+1}:=M_{n+1}-M_n$ have the property $E[X_{n+1}∣\mathcal{F}_n]=0$. This zero-mean condition (conditionally on the past) is analogous to the “mean zero” condition in the classical central limit theorem (CLT) for independent random variables.

 

 

• Classical CLT:
  The standard CLT tells us that if you sum many independent, identically distributed (i.i.d.) random variables (with some moment conditions), the sum (properly normalized) converges in distribution to a normal (Gaussian) random variable.
• Martingale CLT Insight:
  In a martingale setting, even though the increments $X_i$​ are not independent in the usual sense, they are “conditionally independent” in that each $X_i$ has mean zero given the past. This conditional structure is enough—under the right conditions—to produce a Gaussian limit.