Research
Towards Decentralized Searcher Competition in MEV Markets
Existing MEV centralization research concentrates on builders and validators, treating searchers mainly as sources of bundles. This paper flips the lens onto searchers themselves, asking how fair and…

- arXiv
- 2608.05011
- Published
- 2026-08-05
- Authors
- Roozbeh Sarenche, Yunwen Liu
AI summary
Overview
Research area: Blockchain economics and mechanism design — specifically maximal extractable value (MEV) markets, auction design, and cooperative game theory (Shapley value, fairness indices).
Technical level: Intermediate. The paper uses formal game-theoretic notation (Bayesian information structures, Shapley values, Jain's index, HHI) but its central arguments — that winner-take-all rewards track rank rather than contribution, and that reward-sharing mechanisms must be hardened against Sybil and collusion attacks — are accessible to readers with a general crypto background.
Scope (one sentence): The paper builds a heterogeneous model of MEV searcher competition, uses two new metrics to show that the standard first-price winner-take-all auction concentrates rewards, and proposes and security-analyzes an entry-filtered "Shapley-capped" auction as a more decentralized alternative.
What This Paper Is About
Existing MEV centralization research concentrates on builders and validators, treating searchers mainly as sources of bundles. This paper flips the lens onto searchers themselves, asking how fair and decentralized the current first-price, winner-take-all searcher auction actually is when searchers differ in which opportunities they can find and how efficiently they can execute them. The goal is to determine whether an alternative auction design can spread rewards more widely without opening the door to Sybil identities (copying or degrading one's own code under new identities) or to validator–searcher collusion.
Key Contributions
-
A heterogeneous searcher-competition model and evaluation criteria. Searchers differ along two axes: opportunity coverage (a per-class probability α<sub>i,τ(o)</sub> of discovering an opportunity) and execution quality (a certified, normalized net value z<sub>i,o</sub>). The paper defines two economic metrics — a Shapley-weighted Jain fairness index and an expected-reward HHI / effective number of rewarded searchers — and two security criteria — Bayesian copied-code Sybil resistance and Bayesian validator–searcher coalition resistance.
-
A formal analysis of the first-price auction (FPA) as benchmark. The authors prove the FPA has strong (ex-post) copied-code Sybil resistance and validator–searcher coalition resistance, because under monotone searcher-specific bidding, copied or degraded versions of one's own code cannot improve either the searcher's payoff or the searcher–validator joint payoff. Economically, however, FPA rewards "rank dominance" — being the highest-bidding active participant — rather than marginal contribution.
-
The Shapley-Capped Auction. A proposed mechanism that still executes the highest certified-value submission, but distributes retained rewards according to the Shapley value of the validator-dependent max game when the number of admitted submissions is below a cap, switches to a conservative fallback branch and burns the residual when the cap is exceeded, and uses cap and burn parameters to control the incentives created by reward sharing. The paper proves budget feasibility and derives a feasible region for the security parameters.
-
Empirical calibration on Ethereum arbitrage data. An analysis of Ethereum arbitrage data to estimate practical patterns of searcher participation, profit concentration, and opportunity coverage, used to evaluate the mechanism under realistic parameters. The authors explicitly caution that this does not fully identify real-world searcher entities, since one entity may control multiple contracts.
Main Findings
-
First-price auctions are secure but centralizing. The FPA is ex-post copied-code Sybil-resistant and validator–searcher coalition-resistant under monotone searcher-specific bidding, but its rewards are governed by rank dominance rather than marginal contribution.
-
Heterogeneity makes concentration worse. When execution quality is unequal, first-price rewards can concentrate on a small dominant class; in the limiting case of a single dominant searcher, the effective number of rewarded searchers approaches one.
-
Real markets are already skewed. The paper cites Heimbach et al., who find that eleven searchers account for more than 80% of identified arbitrage volume in centralized–decentralized exchange arbitrage, with the two largest searchers alone accounting for 49.2%. It also notes that the leading builder, Titan, accounts for around 50% of blocks in May 2026.
-
Sybil defense is the core design obstacle. Any mechanism that naively rewards participation invites searchers to split one execution strategy across multiple identities submitting copied or degraded code, and invites validators to collude when splitting raises joint payoff. The paper's security analysis shows a bounded range of Sybil sizes suffices to consider, yielding a feasible parameter region for copied-code Sybil resistance and coalition resistance.
-
The Shapley-capped mechanism improves fairness where it matters most. The theoretical benchmarks show the proposed mechanism is especially effective in centralized opportunity classes, where first-price competition concentrates rewards among a small set of searchers, and is a potential improvement over first-price competition in empirically motivated opportunity types.
-
Budget feasibility holds. The mechanism is proven budget-feasible, with residual rewards burned rather than redistributed when the cap is exceeded.
-
Metric behavior. The Shapley-weighted Jain index ranges from 1/n to 1, equaling 1 under contribution-fair allocation (R<sub>i</sub> = λφ<sub>i</sub><sup>VS</sup>) and 1/n when only one searcher has a positive reward-to-contribution ratio. The expected-reward HHI ranges from 1/n to 1, with its reciprocal giving the effective number of rewarded searchers.
Methodology in Plain English
The authors build an abstract model of a single MEV opportunity in which each searcher may or may not discover it (a coin flip with a searcher- and opportunity-class-specific probability) and, if it does, submits execution code whose value and cost are certified under a canonical simulation rule and normalized by a public opportunity-scale parameter. Searchers know their own code's value but not rivals' submissions — an information structure the authors say can be implemented with an encrypted mempool or commit–reveal — so security is evaluated in a Bayesian (expected-payoff) sense rather than state by state.
For fairness, they treat each searcher's contribution as its Shapley value in a cooperative game whose value function is the expected maximum certified value achievable from a coalition that includes the validator. They then apply Jain's index to the ratio of expected reward to that Shapley contribution, and separately apply the Herfindahl–Hirschman Index to expected reward shares.
They analyze the standard first-price auction first, deriving closed-form expected rewards under a proportional-bidding benchmark (bid = ρ·z, so the winner keeps fraction 1−ρ of certified net value), then design the alternative: admit submissions up to a cap, split retained rewards by Shapley value, and burn the remainder beyond the cap. Cap and burn parameters are then tuned by checking two attack scenarios — a searcher submitting k ≥ 2 copied or degraded versions of its own code, and a searcher–validator coalition doing the same — and finally the model is calibrated against Ethereum arbitrage data.
Why This Matters
Impact on research. The paper shifts attention from builder concentration to the searcher auction itself, and supplies reusable measurement tools (Shapley-weighted Jain index, expected-reward HHI) plus a formal security vocabulary (copied-code Sybil resistance, validator–searcher coalition resistance, in both ex-post and Bayesian forms) for evaluating any proposed searcher mechanism. It also sits directly against the result of Pan et al. that the second-price auction with symmetric tie-breaking is the only non-wasteful, incentive-compatible, Sybil-proof direct mechanism in single-parameter settings — the authors' capped, burning design is a route around that constraint.
Real-world applications:
- Designing order-flow auctions and searcher auctions on rollups and app-chains that want broad searcher participation rather than a few dominant bots.
- Evaluating whether a proposed MEV redistribution or rebate scheme accidentally pays for identity splitting.
- Stress-testing encrypted mempool and commit–reveal designs, whose timing assumptions the paper's Bayesian security model explicitly relies on.
- Auditing existing bundle markets for the rank-dominance effect, where a single high-quality searcher's rewards approach full concentration.
Industry relevance. Searcher concentration propagates through the supply chain: dominant searchers become preferred counterparties, builders with access to their flow gain an advantage, and validators face a block-building market shaped by privileged order-flow channels. For builders, relays, and protocol designers, the paper offers a concrete parameterized alternative and a warning that "reward more participants" is not automatically pro-decentralization.
Future Directions
- Empirical identification of searcher entities. The authors acknowledge their on-chain analysis cannot fully separate real searchers from multiple contracts controlled by one entity; resolving this would sharpen calibration of coverage and execution-quality primitives.
- Opportunity-class clustering. The paper explicitly leaves outside the formal model how opportunities are grouped into classes and how active sets are empirically identified, offering only a suggestive rule (group submissions when executing the best one pushes the residual profitability of the rest below a threshold).
- Extending beyond on-chain, atomic settings. The model abstracts away from off-chain and cross-domain factors such as centralized-exchange price movements and non-atomic, multi-chain MEV, which the related work identifies as a growing share of searcher activity.
- Parameter selection in practice. The security results give a feasible region for the cap and burn parameters, but the paper does not report a concrete recommended parameterization or quantitative benchmark results (the supplied content is truncated before the numerical section, so specific benchmark magnitudes and on-chain data statistics beyond the cited figures are not reported here).
Target Audience
Protocol and mechanism designers working on MEV supply chains, order-flow auctions, and encrypted mempools; game-theory researchers interested in Sybil-proof and collusion-resistant auction design; and blockchain economists and analysts who study market concentration. Readers need comfort with expected-value reasoning and basic cooperative game theory to follow the formal sections, but the framing and conclusions are legible to anyone familiar with the searcher–builder–validator pipeline.
Authors’ abstract
Centralization in maximal extractable value (MEV) markets is a significant concern for blockchain systems, as persistent concentration of economic power can weaken competition, reduce openness, and undermine the decentralization goals of permissionless protocols. While much of the existing analysis has focused on builders, validators, and block-building markets, this paper studies centralization from the perspective of searcher competition. We develop a heterogeneous model…
The full abstract on arXiv