tl;dr: Miner Extractable Value Auction makes centralization worse, not better.
In a previous post, I explained why the conventional thinking about how front-running will affect Ethereum is wrong. To recap: if Fred is skilled at ordering transactions to front-run them and maximize his revenue (his MEV or Miner Extractable Value, in the jargon), a rational Fred will sell his block-construction services to all miners: Front-Running as a Service (FRaaS). The world that would likely emerge then, I argued, is not one in which there’s a dominant miner, but one in which there’s a dominant mining pool.
While a dominant mining pool is clearly better than a single centralized miner, you might wonder whether there is some way to reduce Fred’s power even more.
Let’s focus here on one proposed approach to the front-running problem: MEV Auctions (MEVA), and let’s consider what problems (if any) MEVA solves, whether users are any better off in a world with MEVA (spoiler: they’re not), and whether MEVA actually makes things even worse (spoiler: it does).
The idea behind MEVA is rather simple: instead of trying to prevent front-running, auction off the rights to order transactions, and allow the winner (called the Sequencer) to front-run as much as they’d like. Users have two ways to submit transactions: pay the Sequencer a fee for inclusion in an aggregated block or submit transactions directly to the miners. Either way, the Sequencer chooses the order in which the transactions are executed.
So what’s the argument for MEVA if it doesn’t eliminate MEV, and even actively encourages it? One key argument put forth by MEVA proponents, and voiced by Vitalik here, is that MEV is bad since it leads to mining centralization, and MEVA reduces this centralization risk.
MEVA is not helpful
But this argument is wrong, and it turns out that MEVA wouldn’t actually reduce centralization. To see why, let’s compare two worlds, one without MEVA and one with it. In the non-MEVA world, recall that Fred will set up a mining pool and because of his front-running advantage, his pool will dominate.
Now what happens if we add MEVA to that world, and we assume that virtually all transactions go through the MEVA system. The Sequencer auction will be won by the party who is best able to extract MEV — in other words, Fred will be the Sequencer.
Without MEVA, mining-pool-operator-Fred not only gets to choose the ordering of the transactions in his blocks, but he also gets to choose which transactions to include and which to censor. With MEVA, by contrast, Fred chooses the relative order of transactions, but must include all transactions included by the miners.
But who will control the miners at L1? Fred again. Fred can charge higher fees for his L2 sequencing and inclusion services, and increase his profits, if he controls both paths for getting transactions on-chain. (This is a classic tactic of monopolists: if consumers have an alternative to your monopoly product, buy control of that alternative to make sure it doesn’t undermine your monopoly.) So in a post-MEVA world, Fred will be the highest bidder in the MEVA auction and he will also set up a mining pool. Because of his MEV advantage, Fred will be able to both win the auction and pay out higher rewards to miners than other pools.
It turns out that while MEVA tries to separate transaction ordering from transaction censorship/inclusion, in a world where it’s lucrative to combine these two things, they will end up being controlled by the same party. So “separating” them doesn’t actually reduce any centralization, because of Fred’s incentive to recombine them.
In effect, MEVA auctions both the sequencing rights as well as control of the L1 miners, as the winner will control both. And Fred, with his advanced front-running capabilities, has an advantage over all other bidders.
So does MEVA reduce centralization? Unfortunately not. Both with and without MEVA, Fred will control the transaction ordering as well as transaction inclusion. And in both worlds, Fred will operate a dominant mining pool.
MEVA is harmful
But it’s worse than that; MEVA is actually harmful and makes things worse in several ways. Firstly, as I pointed out previously (and Vitalik conceded), MEVA imposes a tax on users as it increases the amount of MEV that exists, which is ultimately paid for by Ethereum users. Moreover, MEVA auctions off ordering rights for longer periods, but without a monopoly sequencer, Fred has to compete on a per-block basis: at each block miners can choose which mining pool is offering the best reward.
To see this second point, consider a world where there’s another skilled front-runner, Fran. Without MEVA, Fred’s mining pool only buys the ordering rights for one block at a time; at any point Fran can offer a larger reward and lure miners to her pool. With MEVA, by contrast, ordering rights are auctioned off for multi-block periods — the MEVA proposal suggests selling ordering rights for a full day of transactions at a time. This greatly reduces competition and gives the winner an absolute (though time-limited) monopoly on sequencing with significant power to delay, re-order, and censor transactions across large time periods.
So is the current state of affairs bad? Maybe. But does MEVA improve things? No, it actually makes things worse. The community’s energy is better spent on reducing front-running. In our next post, we’ll propose one way to do just that.