Same idea as Pearson, but computed on ranks instead of raw values. Catches any monotonic relationship — linear or curved — and shrugs off outliers.
Pearson r—
Spearman ρ—
Raw values
Ranks (1 = smallest, n = largest)
Picking which one: Pearson assumes the relationship is linear; Spearman just needs it to be monotonic. On the "Monotonic curve" tab, Pearson under-reports the strength because the curve isn't a line — Spearman gets it right.