The 123 Theorem and Its Extensions

Abstract It is shown that for every b > a > 0 and for every two independent identically distributed real random variables X and Y, Prob [|X  Y|⩽b] (2⌜ b a ⌝  1) Prob [|X  Y|⩽a] . This is tight for all admissible pairs a, b. Higher dimensional extensions are also considered.