Buying Time: Latency Racing vs. Bidding in Fair Transaction Ordering

We design a practical algorithm for transaction ordering that takes into account both transaction timestamps and bids. The algorithm guarantees that users get their transactions published with bounded delay against a bid, while it extracts a fair value from sophisticated users that have an edge in latency, by moving expenditure from investment in latency improvement technology to bidding. The algorithm creates a score from timestamps and bids, and orders transactions based on the score. We first show that a scoring rule is the only type of rule that satisfies the independence of latency races. We provide an economic analysis of the protocol in an environment of private information, where investment in latency is made ex-ante or interim stages, while bidding happens at the interim stage where private signals have been observed. The algorithm is useful for transaction sequencing in rollups or in other environments where the sequencer has privileged access to order flows.