Why searchers and builders like candle auctions (and why it matters for the builder market post-Glamsterdam)

Introduction

Candle auctions—open‑bidding mechanisms that end at a random moment—have become a favorite playground for MEV searchers and builders. The key feature is a quasi‑lottery: because the exact deadline is unknown, participants can keep nudging the price up by the smallest tick, hoping to be the last to place a bid before the auction stops. This simple dynamic can generate higher expected profits than a traditional sealed‑bid auction, even without any collusion or repeated interaction. Understanding why this happens is crucial for the emerging builder market that follows the post‑Glamsterdam reforms.

How the random deadline creates a bidding advantage

In a stylised model, two risk‑neutral bidders share a common value \(v\) for winning. The auction ends at a random time \(T\) drawn from an exponential distribution (rate \(\mu\)). Each bidder receives opportunities to revise their bid according to an independent Poisson process (rate \(\lambda\)). The higher the ratio \(\lambda/\mu\), the more chances a bidder has to react before the auction closes.

When the deadline is uncertain, the highest standing bid at any moment is a leader and the other participant is a follower. If the follower gets a revision opportunity, the optimal response is to outbid the leader by the smallest possible increment \(\varepsilon\). This “pennying” strategy is a best response because:

1. Small increments preserve surplus – raising the price only a little keeps most of the value \(v\) for the eventual winner.
2. Leader‑follower flips are cheap – each extra bid merely swaps who is in the lead, without changing the expected final price dramatically.
3. Random termination makes the last bid decisive – the probability of being the final bidder is proportional to the number of revision opportunities before \(T\), which favours fast participants.

Mathematically, the expected number of price increases is \(\mathbb{E}[N]=\frac{\lambda}{\mu}(1-q^{M})\) where \(q=\frac{\lambda}{\mu+\lambda}\) and \(M=(v-r)/\varepsilon\) is the maximum number of ticks. The expected winning payment rises only by \(\varepsilon\frac{\lambda}{\mu}(1-q^{M})\), leaving the combined profit of the two bidders at \(S{M}=v-r-\varepsilon\frac{\lambda}{\mu}(1-q^{M})\). Because each bidder has a \(\frac{\lambda}{\mu+2\lambda}\) chance to be the one who opens the auction, the ex‑ante expected profit per participant is \(U=\frac{\lambda}{\mu+2\lambda}S{M}\). When \(\lambda\) is much larger than \(\mu\) but still far below \(M\), the profit split approaches an even 50/50 share of the surplus—far better than the zero‑profit outcome of a sealed bid where the winner pays their full value.

Why builders care in the post‑Glamsterdam world

Glamsterdam introduced proposer‑builder separation (PBS) and a more formal relay market, aiming to reduce the arms race over MEV extraction. However, the candle‑auction dynamics described above give existing builders a strong incentive to preserve an open‑bid environment:

• Higher expected returns – By limiting the pool of participants (e.g., through relay permissioning), a builder can ensure that only a few fast bidders compete, maximising the \(\lambda/\mu\) advantage.
• Reduced need for coordination – The pennying equilibrium arises without explicit collusion; simply being fast and staying in the auction is enough. This makes the strategy robust even if the market later adopts sealed‑bid ePBS mechanisms.
• Latency investment justification – Builders already spend heavily on reducing latency. The model shows that such investment directly translates into a larger \(\lambda\) and therefore a higher probability of being the final bidder, reinforcing the economic rationale for speed.

If the community moves toward sealed‑bid auctions, the expected profit collapses to \(v-b\) where \(b\) is the bid submitted at the deadline—often close to \(v\) itself. Builders would lose the surplus‑splitting benefit that candle auctions provide, potentially reshaping the incentive landscape and discouraging the heavy infrastructure spend that currently underpins the MEV ecosystem.

Practical takeaways for participants

1. Focus on reaction speed – Investing in faster relays, better network positioning, or more aggressive bid‑submission bots raises \(\lambda\) and improves expected profit.
2. Control participation – Restricting who can submit bids in a relay auction (e.g., via permissioned relays) concentrates the competition among a few fast actors, amplifying the pennying advantage.
3. Monitor protocol shifts – Any move toward sealed‑bid or deterministic deadline designs will erode the profit premium of candle auctions; builders should be ready to adapt their revenue models.
4. Consider mixed strategies – In some parameter regimes, a large initial jump (bidding \(v-\varepsilon\) immediately) can dominate pennying. Understanding the \(\lambda/\mu\) ratio helps decide which equilibrium is more profitable.

Conclusion

Candle auctions illustrate how a simple random‑deadline mechanism can turn an open‑bid market into a quasi‑lottery that rewards speed and limited competition. For Ethereum’s builder market after Glamsterdam, this means that preserving an open‑bid, few‑participant environment can sustain higher profits and justify the ongoing latency arms race. As the ecosystem debates sealed‑bid ePBS designs, participants should weigh the loss of the pennying surplus against any gains in fairness or predictability. The choice will shape the economic incentives that drive the next generation of MEV extraction and builder strategies.