The COMB strategy's global optimality range

Let $n$ be the number of experts, and let $\mathbb{S}_n$ denote the sector on which global asymptotic optimality is considered. The COMB strategy is the strategy that orders experts by regret and uses the alternating zeros-and-ones vector.

COMB optimality conjecture. The COMB strategy is globally asymptotically optimal on $\mathbb{S}_n$ only for $n\leq 4$ experts. The paper reports strong numerical evidence against global optimality for $5\leq n\leq 10$, while the assertion for general $n$ remains conjectural.