In the classical contention‑resolution problem a set of players share a broadcast channel; a player succeeds in a time step only if it is the sole broadcaster. The traditional model assumes every player can act at every step—once it arrives it participates until it succeeds. This work removes that assumption by introducing an adversarial scheduler that, at each step, selects a subset of players to be scheduled, while unscheduled players simply sleep. We measure a protocol by its efficiency, defined as the number of players divided by the expected number of steps in which at least one player is active.\n\nOur first result shows that, provided the players have access to a shared global clock and shared randomness, they can achieve efficiency $\Omega(1/\log n)$ even without any channel feedback and against an adaptive scheduler. We then prove a matching lower bound $O(1/\log n)$ that holds even with ternary feedback and an oblivious scheduler. All upper bounds rely on the shared clock and randomness.\n\nFinally, we demonstrate that this implicit shared state is indispensable: without a global clock or shared randomness, no protocol can attain efficiency better than $O(\log n / n)$, even when the channel offers ternary feedback and the adversary is oblivious.\n\nReview: The paper rigorously identifies the essential role of shared timing and randomness in achieving efficient contention resolution under adversarial scheduling, and provides tight upper and lower bounds that close a gap in the existing literature.