Not All Volatility Means Risk

The goal here: build quant finance intuition from first principles.
What's actually being measured by "volatility"?
- Take this price process
- $P(t) = P(t-1)\cdot\left(1 + r\cdot(-1)^t\right)$; $r > 0$
- Watch it alternate ($t \ge 2$)
- Up $100r\%$, down $100r\%$, up $100r\%$, down $100r\%$, forever. Nothing random about it.
- Run the sample standard deviation on it (assuming an even number of return observations)
- Mean of returns = 0 (they split evenly between $+r$ and $-r$).
- Every return is exactly $\pm r$, so the sum of squared deviations is $n\cdot r^2$ — giving $s = r\sqrt{n/(n-1)} > 0$, which is about $r$, and gets closer to it the higher $n$.
- See what that number can't tell you: sample standard deviation reflects dispersion, it doesn't allow for apt diagnostic of the data-generating process.
- It answers "how spread out are the values around the mean?" and nothing more.
- It cannot answer "is this spread coming from noise or from a rule?" — and thus it cannot itself distinguish true risk from a hidden, exploitable pattern.
Bottom line: obviously, this is an overkill as far as real financial time series go.
It's put deliberately so extreme to make the point that sample standard deviation doesn't explain the nature of a series unambiguously — and can therefore be misleading, especially for those at an early stage of familiarizing themselves with financial time series analysis.

Dr Krzysztof Ozimek
Quantitative Investment & Trading Research Educator
Related methodological resources
Continue Exploring Quantitative Finance Methodology
Explore PDF textbooks, implementation-oriented bundles, and methodological frameworks in quantitative investing, trading research, and scientific backtesting.
Explore PDF Textbooks & Bundles
