Sample size
How many past weeks does a comparison actually rest on, and what does a thin sample hide?
In plain words
Sample size is the number of past observations behind a statistic. Comparing this week's return or volatility with history requires enough earlier weeks for the comparison to mean anything; with only a few, the reference point is unstable.
Why it matters
Statistics drawn from a thin sample swing sharply when one more week is added, so percentiles, averages and volume comparisons computed on a young market can mislead. A careful desk checks how many weeks of history underlie a ranking before treating it as normal or extreme.
In a real week: ETH, week of 26 February 2018
ETH in the week of 26 February 2018 returned 2.99% on dollar volume of 0.29 billion, but the data carried only 2 weeks of history, so the return percentile against history was left empty and the volume comparison against the prior 12 weeks was unavailable. The two stored prior weeks were 12.67% and -7.66%, a sample too thin to call this week ordinary.
A common mistake
Reading an empty percentile as an error. It is often an honest sign that the market has not existed long enough for a fair comparison.