Stochastic Calculus

Lesson 1 of 11

Brownian Motion

The same random walk with finer and finer steps: a path that is everywhere continuous and nowhere smooth.

The idea

Start with a random walk that takes a step every Δt\Delta t units of time, where each step is up or down by Δt\sqrt{\Delta t}. Now let Δt\Delta t shrink toward zero. The steps get smaller and more frequent, and in the limit the walk becomes Brownian motion, written WtW_t.

Brownian motion has three defining properties:

  • It starts at zero: W0=0W_0 = 0.
  • Its increments are independent: how far it moves over one time interval tells you nothing about how far it moves over another, non-overlapping interval.
  • Its increments are normal: the change from time ss to time tt is Wt−Ws∼N(0,t−s)W_t - W_s \sim N(0, t-s), so the variance equals the time elapsed.

Because the variance at time tt is tt, the typical distance from the start is t\sqrt t. This is the same square-root scaling as the random walk it came from.