The MEV Carry Trade Just Blew Up: Base’s 37% Deadweight Loss

Consider this the year of our algorithm, 2025, and the blockchain is an accounting ledger more pockmarked than a pirate’s log. 1.4 billion transactions on Ethereum, Optimism, and Base execute a grand ballet of nothing: the state-diff is zero, the result a perfect vacuum save for the fees consumed[^claim_3170]. On Ethereum L1, this is a marginal 2.6% nuisance, a mere scratch on the paint[^claim_3171]. But on the OP Stack, the rot is structural: Optimism at 24%, and Base—the infernal casino of low-cost throughput—at a staggering 37%[^claim_3171]. The stack isn’t scaling Ethereum; it’s amplifying noise at a signal-to-cost ratio that would bankrupt a hedge fund.

The culprits are not some monolithic botnet. Speculative MEV—failed sandwiches, state-neutral arbitrages that collapse at the moment of execution—accounts for 57% of this dead weight on Optimism and 68% on Base[^claim_3172]. Picture a trader in a cold, glass-and-steel office, fingers stabbing at a terminal that echoes with the hollow ring of an empty shell. The algorithm is a jagged line of code, a shiv in the mempool’s dark, flicking out at every perceived whisper of profit. And yet, once fees are tallied, these searchers are not the cunning predators of legend; they are the prey of their own mechanics, subsidizing congestion with every loss-leader trade[^claim_3173]. The rest of the gap comes from other sources—a classification that shatters the old myth that MEV alone is the great demultiplier of L2 value[^claim_3175].

On Ethereum’s mainnet, the profile inverts. Address poisoning campaigns reign, 53% of these state-invariant ghosts[^claim_3174]. Attackers craft addresses that mirror a victim’s past contacts, sending zero-value transfers as bait—a phishing expedition where the chain itself becomes the attack surface. It’s a transparency tax, and the cost is exacted in wasted blockspace, a slow bleed of user trust into the ether.

The paper’s definition of state-invariant transactions—those whose inclusion changes nothing but the ledger’s fee column[^claim_3169]—isn’t merely academic. It’s a computable predicate, a check that any node can run. For sequencer networks like Espresso or Radius, this is a market signal: repricing this waste is no longer a choice but a yield optimization. By filtering out these loss-making speculative MEV trades, a sequencer can boost its own revenue, cut batch costs, and lower the fee floor for every user—a direct transfer of searcher losses into chain economics. Wallet providers can deploy the same filter to hide zero-value poison transactions, cutting phishing off at the knees without waiting for a user to click. And at the protocol layer, an EIP that surcharges zero-state-diff transactions would make spam economically irrational—a short-squeeze on nonsense—without harming legitimate use cases. The net effect? A dramatic reduction in phantom costs and phishing, turning the chain’s deadweight into pure profit[^claim_3176].

The taxonomy reveals a simple truth: a large fraction of blockchain activity is deadweight, and on Base, 37% of all blockspace is provably nothing. The path to cleaner ledgers starts with admitting that a third of the chain’s work is the sound of one hand clapping.