pith. sign in

arxiv: 2605.22667 · v1 · pith:RFFJLXAMnew · submitted 2026-05-21 · 💻 cs.GT · q-fin.TR

Imperfect Commitment in Maximal Extractable Value Auctions

Aleksei Adadurov , Sergey Barseghyan , Anton Chtepine , Antero Eloranta , Andrei Sebyakin , Arsenii Valitov This is my paper

Pith reviewed 2026-05-22 04:01 UTC · model grok-4.3

classification 💻 cs.GT q-fin.TR
keywords MEV auctionsimperfect commitmentblock builderssearchersEthereumdefectionarbitragesandwich opportunities
0
0 comments X

The pith

Imperfect commitment allows Ethereum block builders to capture additional surplus from MEV opportunities after seeing bids.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper studies how block builders in MEV auctions can defect after observing searcher bids and bundles, since the protocol does not force them to honor the outcome. Builders are modeled as defecting with fixed probability ε and then replicating a type-specific fraction γ(τ) of the winning opportunity. Searchers anticipate this behavior and choose between a risky first-price bid or a safe deterrence bid that removes the profit from defection. The resulting equilibrium bidding is piecewise and the cost of imperfect commitment rises with both replicability and competition intensity. Data from the libmev dataset shows clear heterogeneity: sandwich opportunities are already highly competitive with little exposed surplus, while naked arbitrage and liquidations leave substantially more value open to builder defection.

Core claim

The equilibrium in these auctions with imperfect commitment produces a piecewise outcome in which the cost depends jointly on replicability γ(τ) and competition; empirical decomposition of observed auction revenue against the surplus a defecting builder could capture reveals sharp heterogeneity across MEV types, with sandwiches showing high competition and low exposed surplus while arbitrage and liquidations show the opposite.

What carries the argument

The defection model in which a builder defects with probability ε and replicates a type-specific fraction γ(τ) of the winning MEV opportunity, with searchers choosing between risky first-price bids and safe deterrence bids.

If this is right

  • The cost of imperfect commitment depends jointly on replicability and competition.
  • Sandwich opportunities are already highly competitive with minimal exposed surplus.
  • Naked arbitrage and liquidations leave substantially more surplus exposed to builder defection.
  • Credible MEV auctions require constraints on the builder's ability to use observed bid and payload information ex post.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Hiding bid or payload details from builders until after the block is built could lower effective defection incentives.
  • Comparing auction outcomes before and after any information-hiding rules would test whether the exposed surplus shrinks.
  • Allowing the defection probability to vary with opportunity size might sharpen the predicted heterogeneity across types.

Load-bearing premise

Builders defect with a fixed probability ε and replicate a type-specific fraction γ(τ) of the opportunity upon defection, and searchers correctly anticipate this when choosing between risky and safe bids.

What would settle it

Direct measurement of actual builder defection rates together with a check of whether the observed revenue decomposition matches the predicted surplus exposure for each MEV type.

Figures

Figures reproduced from arXiv: 2605.22667 by Aleksei Adadurov, Andrei Sebyakin, Antero Eloranta, Anton Chtepine, Arsenii Valitov, Sergey Barseghyan.

Figure 1
Figure 1. Figure 1: FIG. 1: Timing of the game [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: Bribe percentage vs extracted value by MEV [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Bergemann disclosure benchmark by effective [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Auction revenue (blue) vs foregone frontrun [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
read the original abstract

Ethereum block builders run sealed auctions among searchers, but nothing in the protocol forces a builder to honor the auction outcome after observing submitted bundles. This paper studies the commitment problem. We model a builder who defects with probability $\varepsilon$ and, upon defection, replicates a type-specific fraction $\gamma(\tau)$ of the winning MEV opportunity. Searchers anticipate this behavior and choose between a risky first-price bid and a safe deterrence bid that makes frontrunning unprofitable. The resulting equilibrium is piecewise, with the cost of imperfect commitment depending jointly on replicability and competition. Using the \texttt{libmev} dataset, we estimate $\gamma(\tau)$ from right-tail bribe plateaus and decompose observed auction revenue against the surplus a defecting builder could capture. The results show sharp heterogeneity across MEV types: sandwich opportunities are already highly competitive, while naked arbitrage and liquidations leave substantially more surplus exposed to builder defection. Credible MEV auctions, therefore, require not only an auction format, but also constraints on the builder's ability to use observed bid and payload information ex post.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper models imperfect commitment in MEV auctions on Ethereum, where a builder defects with fixed probability ε and, upon defection, replicates a type-specific fraction γ(τ) of the winning opportunity. Searchers anticipate this and select between a risky first-price bid and a safe deterrence bid. The resulting equilibrium is piecewise and depends on replicability and competition. Using the libmev dataset, the authors estimate γ(τ) from right-tail bribe plateaus, decompose observed auction revenue into surplus exposed to defection, and report sharp heterogeneity: sandwich opportunities are already highly competitive while naked arbitrage and liquidations leave substantially more surplus exposed.

Significance. If the identification of γ(τ) holds, the results demonstrate that auction formats alone are insufficient for credible MEV extraction and that ex-post constraints on builder information use are required. The theoretical equilibrium derivation provides a clean framework for analyzing commitment in repeated auctions with private information, and the data-driven decomposition offers a novel empirical lens on MEV-type heterogeneity.

major comments (2)
  1. [§5] §5 (Estimation of γ(τ)): The identification strategy locates right-tail bribe plateaus in the libmev data and interprets them as the replicable fraction γ(τ) a defecting builder can capture. This assumption is load-bearing for the heterogeneity claims, yet the paper does not provide robustness checks against alternatives such as equilibrium bid shading under varying searcher counts or intrinsic shapes of MEV profit functions (e.g., sandwich saturation). If plateaus arise from these factors instead, the subsequent decomposition of revenue into “surplus exposed to defection” misattributes differences across MEV types.
  2. [§6] §6 (Decomposition results): The reported surplus-exposure figures for each MEV type are constructed by subtracting a quantity that depends on the fitted γ(τ) estimated from the same auction data. This creates a circularity risk between the parameter and the heterogeneity conclusions; the paper should clarify whether the decomposition uses out-of-sample validation or alternative identification to break the dependence.
minor comments (2)
  1. [§3] Notation for the safe deterrence bid and the piecewise equilibrium thresholds could be clarified with an explicit table summarizing the four regimes.
  2. [§6] Figure 4 (or equivalent revenue decomposition plot): axis labels and legend entries should explicitly state that the exposed-surplus bars are computed using the estimated γ(τ) rather than raw data.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the identification strategy for γ(τ) and the potential circularity in our decomposition. We address each point below and outline revisions to strengthen the empirical sections.

read point-by-point responses

  1. Referee: §5 (Estimation of γ(τ)): The identification strategy locates right-tail bribe plateaus in the libmev data and interprets them as the replicable fraction γ(τ) a defecting builder can capture. This assumption is load-bearing for the heterogeneity claims, yet the paper does not provide robustness checks against alternatives such as equilibrium bid shading under varying searcher counts or intrinsic shapes of MEV profit functions (e.g., sandwich saturation). If plateaus arise from these factors instead, the subsequent decomposition of revenue into “surplus exposed to defection” misattributes differences across MEV types.

    Authors: We appreciate the referee's point on the load-bearing nature of the plateau interpretation. Our model predicts that right-tail stabilization occurs precisely when additional bidding no longer offsets the expected loss from replication, which aligns with the observed flat tails across opportunity sizes. While we acknowledge that searcher count variation or saturation could contribute, the theoretical framework distinguishes replicability-driven plateaus from pure shading because the latter would not produce type-specific heterogeneity consistent with the data. In revision we will add explicit robustness checks, including controls for searcher participation and comparisons of profit-function curvature across MEV types, to isolate the replicability channel. revision: partial

  2. Referee: §6 (Decomposition results): The reported surplus-exposure figures for each MEV type are constructed by subtracting a quantity that depends on the fitted γ(τ) estimated from the same auction data. This creates a circularity risk between the parameter and the heterogeneity conclusions; the paper should clarify whether the decomposition uses out-of-sample validation or alternative identification to break the dependence.

    Authors: We agree there is a dependence risk when γ(τ) and the decomposition draw from the same observations. To break this, the revised manuscript will implement a split-sample procedure: γ(τ) will be estimated on one partition of the data (by time or MEV subtype) and applied to the held-out partition for the surplus-exposure calculations. We will also report the resulting heterogeneity under this out-of-sample approach and discuss remaining limitations of the identification. revision: yes

Circularity Check

1 steps flagged

Revenue decomposition and MEV-type heterogeneity rest on γ(τ) fitted from the same libmev bribe data

specific steps
  1. fitted input called prediction [Abstract]
    "Using the libmev dataset, we estimate γ(τ) from right-tail bribe plateaus and decompose observed auction revenue against the surplus a defecting builder could capture. The results show sharp heterogeneity across MEV types: sandwich opportunities are already highly competitive, while naked arbitrage and liquidations leave substantially more surplus exposed to builder defection."

    γ(τ) is fitted from right-tail plateaus in the libmev data; the surplus a defecting builder could capture is then defined using that fitted γ(τ). The decomposition of observed revenue into exposed surplus therefore uses the same data-derived parameter, so the reported type-specific heterogeneity is statistically tied to the estimation procedure rather than an independent prediction or external benchmark.

full rationale

The paper estimates the replicability parameter γ(τ) directly from right-tail bribe plateaus observed in the libmev dataset for each MEV type τ. It then defines the surplus a defecting builder could capture in terms of this same γ(τ) and decomposes observed auction revenue against that quantity to produce the headline heterogeneity result (sandwiches competitive, arbitrage and liquidations more exposed). Because the exposed-surplus metric is constructed from the fitted γ values drawn from the identical data, differences across types are partly determined by the estimation step itself rather than emerging independently from raw revenue patterns. This matches the fitted-input-called-prediction pattern and raises the circularity score to 6 while leaving room for independent content in the equilibrium model and auction-format conclusions.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on a behavioral model of builder defection and searcher anticipation plus empirical extraction of replicability from observed bids; these are not derived from first principles but introduced to match the institutional setting.

free parameters (2)
  • ε
    Probability that a builder defects after observing the winning bid; introduced as a modeling parameter without independent calibration shown in abstract.
  • γ(τ)
    Type-specific fraction of MEV opportunity replicable by a defecting builder; estimated from right-tail bribe plateaus in the libmev dataset.
axioms (1)
  • domain assumption Searchers correctly anticipate the builder's defection probability and replicability when choosing bids.
    Invoked to derive the piecewise equilibrium between risky first-price and safe deterrence bids.

pith-pipeline@v0.9.0 · 5745 in / 1346 out tokens · 53421 ms · 2026-05-22T04:01:28.760695+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages · 1 internal anchor

  1. [1]

    Imperfect Commitment in Maximal Extractable Value Auctions

    honest-disclosure benchmark, interpreting the honest- disclosure threshold as an upper bound on user-beneficial arXiv:2605.22667v1 [cs.GT] 21 May 2026 2 disclosure once strategic builder behavior is admitted. The contribution is twofold. First, the paper isolates the builder-defection channel in MEV auctions from the auction-format question studied in [1]...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  2. [2]

    shows how public transaction information creates priority-gas-auction races and consensus risk, while [13] formalizes the AMM sandwich channel through which or- der details such as size and slippage become directly ex- ploitable. These papers establish replication-based fron- trunning as the relevant manipulation channel in our set- ting: once the builder...

    work page

  3. [3]

    The builder’s defection rateεbecomes known

    work page

  4. [4]

    Each ofNpotential searchers decides whether to enter, and the number of entrantsn≤Nis realized

    work page

  5. [5]

    Each entrantiobserves her own private signalz i and submits a (sealed) bidb i to the builder

    work page

  6. [6]

    Resolution of the game: with probabilityε the builder defects and uses the observed bids (b1, . . . , bn) together with the winning transaction payload to frontrun the winner; with probability 1−εthe builder honors the auction outcome, i.e., the highest bidder wins and pays their own bid. Builder’s defection rateεis known Searchers decide whether to enter...

    work page

  7. [7]

    Ifγ(τ)v > β(v)for∀v∈ V(high extractability regime), thenε ∗ = 1

    work page

  8. [8]

    Proof.By Theorem 3.4, ε(v, τ) is strictly decreasing in v, implying its inversev ∗(ε) is strictly decreasing inε

    Ifγ(τ)v < β(v)for∀v∈ V(low extractability regime), thenε ∗ = 0. Proof.By Theorem 3.4, ε(v, τ) is strictly decreasing in v, implying its inversev ∗(ε) is strictly decreasing inε. Hence, dv∗ dε <0. Becauseε, 1−γ(τ), andf (1) are all non- negative, the term− dv∗ dε v∗ε(1−γ(τ))f (1)(v∗) is strictly positive. In the high extractability regime, the integrand in...

    work page 2024

  9. [9]

    and Valitov, A

    Adadurov, A., Barseghyan, S., Chtepine, A., Eloranta, A., Sebyakin, A. and Valitov, A. (2026) ‘Open vs. sealed: Auction format choice for maximal extractable value’, arXiv preprint arXiv:2603.16333

    work page arXiv 2026

  10. [10]

    and Weinschelbaum, F

    Arozamena, L. and Weinschelbaum, F. (2009) ’The effect of corruption on bidding behavior in first-price auctions’, European Economic Review, 53(6), pp. 645–657

    work page 2009

  11. [11]

    and Roughgarden, T

    Bahrani, M., Garimidi, P. and Roughgarden, T. (2024) Centralization in block building and proposer–builder separation. arXiv preprint arXiv:2401.12120

    work page arXiv 2024

  12. [12]

    and Winter, E

    Bergemann, D., Heumann, T., Morris, S., Sorokin, C. and Winter, E. (2022) ’Optimal information disclosure in auctions’. Working paper, Yale University, MIT and the Hebrew University of Jerusalem

    work page 2022

  13. [13]

    and Pesendorfer, M

    Bergemann, D. and Pesendorfer, M. (2007) ’Informa- tion structures in optimal auctions’, Journal of Economic Theory, 137(1), pp. 580–609

    work page 2007

  14. [14]

    (2009) ’Revealing information in auctions: The allocation effect’, Economic Theory, 38(1), pp

    Board, S. (2009) ’Revealing information in auctions: The allocation effect’, Economic Theory, 38(1), pp. 125–135

    work page 2009

  15. [15]

    and Che, Y.-K

    Burguet, R. and Che, Y.-K. (2004) ’Competitive pro- curement with corruption’, RAND Journal of Economics, 35(1), pp. 50–68

    work page 2004

  16. [16]

    and Verdier, T

    Compte, O., Lambert-Mogiliansky, A. and Verdier, T. (2005) ’Corruption and competition in procurement auc- tions’, RAND Journal of Economics, 36(1), pp. 1–15

    work page 2005

  17. [17]

    and Juels, A

    Daian, P., Goldfeder, S., Kell, T., Li, Y., Zhao, X., Ben- tov, I., Breidenbach, L. and Juels, A. (2020) ’Flash boys 2.0: Frontrunning, transaction reordering, and consensus instability in decentralized exchanges’, in Proceedings of the 2020 IEEE Symposium on Security and Privacy (SP). IEEE, pp. 910–927

    work page 2020

  18. [18]

    and Schummer, J

    Eso, P. and Schummer, J. (2004) ’Bribing and signaling in second price auctions’, Games and Economic Behavior, 47(2), pp. 299–324

    work page 2004

  19. [19]

    and Szentes, B

    Eso, P. and Szentes, B. (2007) ’The price of advice’, RAND Journal of Economics, 38(4), pp. 863–880

    work page 2007

  20. [20]

    and Schertenleib, E

    Heimbach, L., Pahari, V. and Schertenleib, E. (2024) Non-atomic arbitrage in decentralized finance. arXiv preprint arXiv:2401.01622

    work page arXiv 2024

  21. [21]

    and Wattenhofer, R

    Heimbach, L. and Wattenhofer, R. (2022) ’Eliminating sandwich attacks with the help of game theory’, in Pro- ceedings of the 2022 ACM Asia Conference on Computer and Communications Security. ACM, pp. 153–167

    work page 2022

  22. [22]

    (1999) ’Auction theory: A guide to the literature’, Journal of Economic Surveys, 13(3), pp

    Klemperer, P. (1999) ’Auction theory: A guide to the literature’, Journal of Economic Surveys, 13(3), pp. 227–286

    work page 1999

  23. [23]

    (2010)Auction theory

    Krishna, V. (2010)Auction theory. 2nd edn. Burlington, MA: Academic Press

    work page 2010

  24. [24]

    and Tirole, J

    Laffont, J.-J. and Tirole, J. (1991) ’Auction design and favoritism’, International Journal of Industrial Organiza- tion, 9(1), pp. 9–42

    work page 1991

  25. [25]

    and Wolfstetter, E

    Lengwiler, Y. and Wolfstetter, E. (2010) ’Auctions and corruption: An analysis of bid rigging by a corrupt auc- tioneer’, Journal of Economic Dynamics and Control, 34(10), pp. 1872–1892

    work page 2010

  26. [26]

    and Shi, X

    Li, H. and Shi, X. (2017) ’Discriminatory information disclosure’, American Economic Review, 107(11), pp. 3363–3385

    work page 2017

  27. [27]

    and Yao, D

    Ma, R., Tang, W. and Yao, D. (2025) Analysis of the order flow auction under proposer–builder separation on blockchain. arXiv preprint arXiv:2502.12026

    work page arXiv 2025

  28. [28]

    Menezes, F. M. and Monteiro, P. K. (2006) ’Corrup- tion and auctions’, Journal of Mathematical Economics, 42(1), pp. 97–108

    work page 2006

  29. [29]

    Milgrom, P. R. and Weber, R. J. (1982) ’A theory of auctions and competitive bidding’, Econometrica, 50(5), pp. 1089–1122

    work page 1982

  30. [30]

    Myerson, R. B. (1981) ’Optimal auction design’, Mathe- matics of Operations Research, 6(1), pp. 58–73

    work page 1981

  31. [31]

    and Matthes, F

    ¨Oz, B., Sui, D., Thiery, T. and Matthes, F. (2024) Who wins Ethereum block building auctions and why? arXiv preprint arXiv:2407.13931

    work page arXiv 2024

  32. [32]

    Pai, M. M. and Resnick, M. (2023) Structural advantages for integrated builders in MEV-Boost. arXiv preprint arXiv:2311.09083

    work page arXiv 2023

  33. [33]

    (1961) ’Counterspeculation, auctions, and competitive sealed tenders’, Journal of Finance, 16(1), pp

    Vickrey, W. (1961) ’Counterspeculation, auctions, and competitive sealed tenders’, Journal of Finance, 16(1), pp. 8–37

    work page 1961

  34. [34]

    and Zhang, F

    Yang, S., Nayak, K. and Zhang, F. (2024) Decentral- ization of Ethereum’s builder market. arXiv preprint arXiv:2405.01329. 10 Appendix A: Additional Derivations

    work page arXiv 2024

  35. [35]

    Affiliation Diagnostics Figure A1 provides the visual counterpart to the affiliation estimates used in Section IV B. Each panel compares the largest and second-largest log extracted values within the same block and MEV type; the positive slope is strongest for liquidations and naked arbitrage, but is present across all categories. FIG. A1: Affiliation dia...

    work page

  36. [36]

    Searcher Concentration Figure A2 shows that extracted MEV is highly concentrated at the searcher level across all types, which supports the right-tail interpretation of the revenue-decomposition results. 11 FIG. A2: Searcher revenue concentration by MEV type. The Lorenz curves and top-kshares show that extracted MEV is highly concentrated across searchers...

    work page

  37. [37]

    This supports the use of type-specificn(τ) in the honest-disclosure benchmark

    Allocation-Effect Diagnostic Figure A3 reports the Board-style allocation-effect check: revenue and bribe shares do not move uniformly with effective bidder count. This supports the use of type-specificn(τ) in the honest-disclosure benchmark. FIG. A3: Board-style allocation-effect diagnostic. The panels compare revenue and bribe shares across effective-bi...

    work page

  38. [38]

    Cross-Builder Bribe Distributions Table A1 reports summary statistics for the eight largest builders by total extracted value. Mean bribe percentages are homogeneous across the top five builders, while Titan’s bribe standard deviation is roughly 3×beaverbuild’s, 12 consistent with the integrated searcher-builder extension of the model. Builder Count Total...

    work page