Skip to content
Crypto.world

The Ledger

Ethereum Research Models How Candle Auctions Preserve Builder Rents

A November 2 Ethereum Research post models how open, unpredictable-deadline auctions can let Ethereum block builders retain surplus through small, repeated bids. The author cautions that the result depends on stylized assumptions and may change under Glamsterdam’s sealed in-protocol bidding channel.

Ethereum Research Models How Candle Auctions Preserve Builder Rents

Two bidders can repeatedly raise Ethereum block bids by the smallest permitted increment and still retain most of the auction’s available surplus, according to a model published November 2 on Ethereum Research. The result challenges the assumption that visible, frequent bidding necessarily squeezes builders’ profits toward zero. The author argues that an uncertain closing time can turn open bidding into a lottery over who places the final bid.

The analysis is theoretical, not a measurement of current Ethereum auctions. Its central claim is narrower: under a particular set of assumptions, a “pennying” strategy can be an equilibrium even without collusion, repeated interaction or punishment for deviating.

Two bidders, many small raises

The model starts with two risk-neutral bidders who share the same known value for winning. The auction ends at a random time, while each bidder gets independent chances to revise a bid; bids are visible immediately, cannot be withdrawn or lowered, and the winner pays the amount bid.

When one bidder leads, the other raises the price by one tick at the next opportunity. The incumbent holds its bid. Because the auction might close before the next response, each successive raise changes who is in front without forcing the winner to bid the full value of the block.

The author describes the finding this way: Open-bidding with random deadline (candle auctions) and few auction participants can lead to higher bidder profits than sealed-bid auctions. The post’s author, writing under the username jcschlegel, says the model’s surplus-preserving outcome requires neither explicit coordination nor repetition.

The result depends on the auction’s timing

The model’s key condition is that bidder response opportunities are frequent compared with the rate at which the auction ends, but not so frequent that the available price increments are exhausted. In the post’s notation, that range is 1 ≪ λ/μ ≪ M, where λ represents the rate of bid-revision opportunities, μ the auction’s closing rate, and M the number of available bid ticks.

In that range, the bidders can trade the lead many times while the price rises only modestly relative to the surplus available. They split the combined profit evenly in expectation. The paper’s result therefore describes incentives inside the model; it does not establish that real builders earn those profits or that they follow the modeled strategy.

The author also constructs another equilibrium in which a bidder jumps almost to the maximum immediately. When the response rate is high enough relative to the number of bid ticks, that strategy can be optimal instead. The existence of multiple equilibria means the pennying result is not a unique prediction of the model, and the post does not claim it is the bidders’ best possible outcome.

Ethereum’s deadlines are not memoryless

The model assumes an exponential closing time, so the remaining time has the same probability distribution regardless of how long the auction has already run. The post says that assumption suited proof-of-work block timing better than today’s proof-of-stake auctions, where proposer behavior, relay policies and network latency shape when bidding ends.

Builders can learn those timing patterns and aim bids closer to an expected deadline. That changes the strategic trade-off: bidding early can invite counterbids, while waiting risks missing the auction. The author says better deadline prediction does not have a clear, one-directional effect on bidder profits; a late bid that rivals cannot observe and answer could instead behave like a sealed bid.

There is another limitation: the model treats response speed as fixed and equal between bidders. It does not model the latency investments, geographic differences or practical delays involved in sending, observing and responding to bids. The post says frequent bidding may help explain infrastructure spending, but it does not solve how builders choose those investments.

Glamsterdam puts open bidding under pressure

The post connects the result to Glamsterdam and enshrined proposer-builder separation, or ePBS. In the author’s account, builders could watch bids on an open relay feed and then submit a competing bid through the protocol’s default sealed channel. A relay seeking to preserve open bidding might need to restrict who can access its feed and enforce participation rules, or rely on proposers to forward bids.

That creates opposing incentives. Builders may value open auctions if they preserve higher rents, while proposers may prefer sealed bids and a credible commitment not to disclose them. The model does not determine which arrangement will prevail, or how much real-world bidder profits would change under either design.

Its concrete warning is that rapid bids and intense latency competition do not, by themselves, prove that builders are competing away their margins. Whether ePBS can keep bids confidential while relays continue to offer open auctions remains unresolved in the post.

Sources

  1. Original source Ethereum Research ethresear.ch

Related stories

Explore

More articles