Sampling frequency
Why can hourly, daily and weekly data describe the same market week so differently?
In plain words
Sampling frequency is how often prices are recorded: every hour, once a day, or once a week. Finer sampling reveals intraperiod swings that coarser sampling compresses into one net change, so the same move can appear calm or violent depending on the window used.
Why it matters
Risk taken within a week is exposed to the path, not the ending value. Weekly data can hide the depth of a fall and the size of a rebound, leaving a position sized on coarse data unprepared for the swings that actually occurred between its checkpoints.
In a real week: ETH, week of 5 August 2024
In the week of 5 August 2024, ETH closed 4.97% lower, a modest weekly figure. Hourly data showed a different picture: a low of 2111.00 on Monday 01:00 UTC, a largest hourly move of -8.21%, a 29.04% range, and annualised realised volatility of 127.0%. Daily data added a worst day of -10.02% on 5 August and a best day of 14.5% on 8 August. The three frequencies describe one week but stress different things.
A common mistake
Reading the weekly net change as the week's risk. A small weekly return can sit atop large hourly and daily swings, and only the finer sampling shows them.