Random Walks

A random walk is a cumulative sum of random increments. In the basic independent, identically distributed finite-variance case, expected position equals the start plus step count times mean increment, while position variance equals step count times increment variance.

A random walk accumulates random increments. In the basic independent, identically distributed model, the mean increment determines drift and the increment variance determines how uncertainty grows. With finite variance, position variance grows in proportion to the number of steps. Its standard deviation grows with the square root of that number; the full possible range need not do so.

The model is useful when cumulative movement looks purposeful even though its increments follow a simple random mechanism. Specify the process before interpreting a streak. A long upward segment can occur without a changing rule, while persistent bias in the increments produces real drift. Failure to find predictability is not proof that a process is a random walk.

When to use it

When distinguishing between predictable and unpredictable processes in a domain; when financial markets exhibit random-walk characteristics that undermine pattern-based strategies; when understanding the diffusion property of random processes (outcomes spread wider over time); when the unpredictability of innovation processes needs to be acknowledged in planning.

How it can help

Use a specified random walk as a comparison model for cumulative fluctuations. Separate drift, uncertainty, and possible outcomes, then test whether its assumptions fit the process.