Optimal-order mixing of the Thorp shuffle
The Thorp shuffle on cards mixes in complete shuffles: as , , while for every integer , , yielding and .
Both the support-based lower bound and the asymptotic upper bound are rigorously proved in Sections 2 through 7.
“Theorem 1.1 (Optimal-order mixing). As d → ∞, ∥q ∗(1600d) d − Un∥TV −→ 0. For every integer t ≥ 0, ∥q ∗t d − Un∥TV ≥ 1 − 2 tn/2 n! . Consequently the mixing time at distance 1/4 satisfies tmix(d) ≥ 2 n log2 3n! 4 = 2d − O(1), tmix(d) = Θ(d).” · Show in paper