Pith. sign in

REVIEW 4 major objections 6 minor 8 references

Does multi-block MEV exist? Analysis of 2 years of MEV Data

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read Multi-block MEV is rarer than random chance in two years of Ethereum block data, the paper finds.

desk verdict Useful dataset and honest analysis, but the main negative claim about multi-block MEV rests on a circular Monte Carlo baseline and should stay conditional. read the letter →

arxiv 2501.12827 v1 pith:BLBDGFBQ submitted 2025-01-22 cs.GT

classification cs.GT
keywords multi-blockMEVEthereumMEV-Boostproposer-builderseparationMonteCarlosimulationblocksequenceanalysisbasefeevolatility
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper examines two years of Ethereum proposer-builder data after the Merge to test whether multi-block maximal extractable value (MEV) — controlling multiple consecutive slots to capture extra value — is actually happening. Using a Monte Carlo simulation that randomly assigns builders to slots based on their observed daily market shares, it finds that multi-slot sequences occur significantly less often than chance would predict, with the longest observed run at 25 slots. Average MEV-Boost payments do rise with sequence length, from about 0.05 ETH for single slots to 0.08 ETH for nine-slot runs, but the per-slot increase is slight. The paper concludes that deliberate multi-block MEV strategies are not being deployed at scale under the current proposer-builder separation mechanism, which matters because proposed protocol changes could remove the risk that currently keeps such strategies in check.

What carries the argument

The load-bearing instrument is the Monte Carlo simulation that generates a null distribution of multi-slot sequences. Builders are randomly assigned to each slot according to their observed daily market share, repeated 100 times, and the average frequency of consecutive sequences of every length is compared with the empirically observed distribution. The gap between those two distributions is what supports the claim that multi-block sequences are not being deliberately pursued.

What would settle it

Recompute the expected sequence-length distribution with a null model that shuffles the order of slots within each day while preserving each builder's total daily count; if observed sequences of length two or more exceed the 95th percentile of that shuffled distribution, the finding of fewer-than-expected sequences would be overturned. Alternatively, identify a concrete pair of adjacent slots where the second slot's winning bid is larger in a way that is only rational if the builder had prepared the first slot, and show the pattern repeats.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the distribution of consecutive builder wins in Ethereum block production is not consistent with a market where agents deliberately accumulate multiple slots. Comparing the observed frequency of sequences of length 1 through 25 against a Monte Carlo baseline built from daily market shares shows fewer long sequences than expected, and this holds for nearly every top builder individually. The paper further finds that average MEV-Boost payments increase roughly linearly with sequence length, but that the payment per slot position within longer sequences rises only mildly, which the author reads as a sign that builders are not bidding aggressively for the later slots that would be needed to capture prepared MEV opportunities. A weak positive autocorrelation in payments and the absence of builder specialization by base-fee volatility round out a picture of a competitive just-in-time auction rather than systematic multi-block control.

Load-bearing premise

The comparison assumes that each slot's builder is drawn independently from the builder's observed daily market share, and if those market shares are themselves influenced by multi-block strategies, the baseline is not independent of the phenomenon being measured.

Editorial extensions

If this is right

  • If confirmed, the result implies that the current MEV-Boost auction, operating as a just-in-time auction, has not been producing systematic multi-block control in the first two years after the Merge.
  • Protocol designers can treat the empirical absence of multi-block MEV as a baseline for evaluating proposed attester-proposer separation and pre-confirmation designs, which could remove the risk that currently deters such strategies.
  • The rising average payment with sequence length, together with only a slight per-slot increase, suggests that any extra value in longer sequences is modest and not enough to elicit aggressive bidding for later slots.
  • The weak short-range autocorrelation of payments means that MEV predictability from public data is limited to one to three slots ahead, so reliable forecasting of MEV spikes is not feasible.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The daily-market-share null is conservative in a specific way: if builders occasionally coordinate or happen to win streaks, those wins inflate the market share used to generate the null, making the null produce longer sequences; so the observed shortfall could actually understate how far the market is from deliberate multi-block control.
  • A stricter test would condition on builder identity and slot timing, for instance by comparing the number of times builder X follows builder X against a null that respects the exact per-builder daily totals; such a test would separate the chicken-and-egg problem cleanly.
  • The same data could be used to backtest an early-warning detector: if the per-slot payment premium for slots at the end of a sequence ever becomes large and stable, that would be the signature of prepared multi-block MEV appearing under a new mechanism design.
  • The author's own suggested cause — that the risk of losing a prepared opportunity to a competing builder deters multi-block strategies — implies that mechanisms with commitments (e.g., pre-confirmations) may need explicit anti-collusion and anti-sequencing protections even though none have appeared historically.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper analyzes Ethereum proposer-builder and MEV-Boost payment data from September 2022 to May 2024 to detect multi-block MEV. The main empirical findings are: (1) observed consecutive same-builder sequences are fewer than those generated by a Monte Carlo simulation based on daily market shares; (2) average MEV-Boost payments increase with sequence length; (3) per-slot payments also rise slightly within longer sequences; (4) there is weak positive autocorrelation in MEV-Boost payments; and (5) there is no builder specialization according to base-fee volatility. The paper concludes that there is currently no indication of systematic deliberate multi-slot MEV strategies.

Significance. If the frequency claim is correct, the paper provides a valuable negative result for the MEV literature, directly addressing a question that has been discussed in the context of PBS redesign and preconfirmations. The manuscript is transparently exploratory, ships its analysis as a Jupyter notebook/Google Colab, and explicitly acknowledges key limitations, which are assets for reproducibility. The central conclusion, however, depends on a Monte Carlo null whose construction is entangled with the phenomenon under test, and the reported sensitivity of the main result to the market-share aggregation window means the headline claim is not yet robust. With the null model repaired, the study could become a useful reference point for future work on multi-block MEV.

major comments (4)
  1. [Section 3, Finding 1 / Figure 1] The central claim that 'significantly fewer multi-slot sequences occur than expected' rests on a Monte Carlo null that draws builders independently from observed daily market shares (Section 2). As the paper itself acknowledges in Section 4, this is a 'chicken and egg-problem': if multi-block MEV is active, it inflates the very market shares used to parameterize the null, so the expected sequence counts already embed part of the phenomenon. This circularity makes the observed deficit difficult to interpret as evidence of absence of deliberate strategies. Moreover, the paper reports that using monthly or total market shares flips the result to 'more longer sequences than expected'. The choice of the daily window is therefore load-bearing, and the conclusion is not robust to this choice. Please provide a null that is independent of the phenomenon being tested, or provide a quantitative sensitivity analysis with a principled justification for the daily window. At minimum, report Monte Carlo confidence intervals so that 'significantly fewer' is a testable statement.
  2. [Section 2 / Figures 3 and 4] The payment findings (Finding 2 and Finding 3) are based on MEV-Boost payments as a proxy for realized MEV, which the paper acknowledges is imperfect. The sample sizes for long sequences are small (the gray numbers in Figure 3), and no confidence intervals or statistical tests are provided for the increasing trend in average payments or the per-slot position trend. Because the discussion in Finding 3 uses the only slight increase in payments to argue against systematic multi-block strategies, the lack of uncertainty quantification is load-bearing for that argument. Please add bootstrap or regression-based confidence intervals, and examine whether the trend holds after controlling for builder identity and time period.
  3. [Section 3, Finding 1] The statement 'significantly fewer' is not supported by a formal comparison against the null distribution. The Monte Carlo is run 100 times, but the paper reports only the average expected frequency. The observed frequencies in Figure 1 and the per-builder deviations in Figure 2 should be plotted against the range or confidence interval of the null distribution, not just its mean. Without this, the reader cannot tell whether the observed deficit is statistically meaningful or within the noise of the simulation.
  4. [Section 4 / Methodology] The limitations list incomplete data, duplicates, and excluded missed slots, but the paper does not discuss the likely direction or magnitude of bias these omissions introduce into the sequence-length distribution. For example, missed slots break what would otherwise be a longer sequence by one builder, and excluding them may artificially reduce the observed frequency of long sequences. Please add a brief quantitative or qualitative bias analysis so that the main frequency claim can be assessed against these data-quality issues.
minor comments (6)
  1. [Abstract / General] The abstract says '2 years of MEV Data', but the study period is September 2022 to May 2024, which is about 20 months; please adjust the wording to match the actual data window.
  2. [Figure 1] The note 'slots>25 have been summarized in slot 25 for brevity' is confusing because the text states the longest observed sequence is 25; clarify whether the figure aggregates Monte Carlo sequences longer than 25 into a single bar at 25.
  3. [Section 2] The sentence 'The ascending MEV-Boost first-price auction by its nature of being public essentially functions like a second price+1 wei auction' is unclear; consider rephrasing to explain the relationship between the public first-price auction and the second-price outcome.
  4. [Section 2, Methodology] The paper mentions that the analysis was also run on monthly and overall market shares, but the results are only described in text and are not shown; please include a supplementary table or appendix with these results, since they are central to the sensitivity concern.
  5. [Section 3, Finding 5] The Cramér's V values of 0.0664 and 0.0772 are presented as 'low' without a reference benchmark; state what values would be considered meaningful for this type of categorical association, or provide a comparison against a random baseline.
  6. [References] Some references (e.g., [3], [4], [5]) are cited with arXiv identifiers but without the full author list in the reference entry; consider adding complete bibliographic details for consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the Monte Carlo baseline is an independence null, not a fitted prediction, and the chicken-and-egg limitation is an endogeneity threat rather than a derivation that reduces to its inputs.

full rationale

The paper's central finding compares an observed distribution of consecutive-slot sequences with a Monte Carlo null distribution generated from observed daily market shares. This is not circular by construction: the market-share marginals are the input, and the sequence-length joint distribution is the output, so the comparison is a goodness-of-fit test of independence given those marginals. The paper explicitly acknowledges the 'chicken and egg-problem' in Section 4, but that is a genuine endogeneity threat to the null—if multi-block MEV inflates the very market shares used to build the baseline, the null may be misspecified under the alternative—not an instance of the conclusion being equivalent to the input. No parameter is fitted to the sequence counts and then renamed a prediction; the Monte Carlo is a stated null model, and the sensitivity to daily versus monthly market shares is reported rather than hidden. The other findings (payment versus sequence length, per-slot payment trend, autocorrelation, and volatility-builder correlation) are direct empirical correlations and do not depend on the Monte Carlo baseline. There are no load-bearing self-citations: reference [3] is prior work by other authors and is used only as a comparison point. Accordingly, no circular step is exhibited and the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central frequency claim depends on observed daily market shares as inputs to the Monte Carlo null; those shares are not externally fixed constants, so they function as fitted inputs. The volatility analyses depend on hand-chosen thresholds. The MEV-Boost payment proxy and base-fee proxy are domain assumptions acknowledged in the text. No new entities are introduced.

free parameters (3)
  • Naive high-volatility percentile cutoff = 10th and 90th percentiles of per-slot base fee changes
    Hand-chosen threshold defines high-volatility periods for the builder specialization analysis; changing this cutoff could change the Cramér's V result in Finding 5.
  • Garman-Klass high-volatility cutoff = top 20% of epoch-level Garman-Klass values
    Hand-chosen threshold for the second volatility classification; affects Finding 5 and the conclusion about builder specialization.
  • Market-share aggregation window = daily (1 day)
    The Monte Carlo expected sequence frequencies depend on this window. The authors note that monthly and overall shares shift the results, and they choose daily as a balance between precision and isolated effects, making this a load-bearing modeling choice.
assumptions (4)
  • domain assumption MEV-Boost payments are a usable proxy for captured MEV per block.
    Section 2 justifies this via competitive auctions and low builder margins, but it is an approximation and excludes non-MEV-Boost blocks and private deals; acknowledged in Section 4.
  • domain assumption Base fee volatility is a meaningful proxy for high- and low-MEV environments.
    Section 2 and Finding 5 state that base fee changes are driven by empty or full blocks, which are partially results of MEV opportunities; this link is asserted rather than demonstrated.
  • ad hoc to paper Builder slots are independent within a day conditional on daily market shares in the Monte Carlo baseline.
    Section 2 randomly assigns builders to slots using daily market shares; this ignores within-day dynamics and is the source of the acknowledged chicken-and-egg problem.
  • domain assumption The cleaned public MEV-Boost dataset is sufficient to estimate multi-block sequence frequencies and payment trends.
    Section 4 acknowledges roughly 10% non-MEV-Boost blocks, duplicate entries, missed slot exclusions, and private off-chain agreements are outside the dataset, so the analysis relies on this assumption for its descriptive claims.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Does multi-block MEV exist? Analysis of 2 years of MEV Data." pith.science (2026). https://pith.science/paper/BLBDGFBQ

@misc{pith2026250112827,
  author       = {Pith},
  title        = {Pith review of: Does multi-block MEV exist? Analysis of 2 years of MEV Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BLBDGFBQ}},
  note         = {Machine review of arXiv:2501.12827}
}
read the original abstract

This study analyzes proposer-builder data and MEV-Boost payment data following the Ethereum merge in September 2022 to identify patterns of multi-block MEV. Our findings reveal fewer multi-slot sequences of builders than predicted by a random Monte Carlo simulation, with the longest observed sequence spanning 25 slots. Additionally, we observe that average MEV-Boost payments increase with the length of consecutive sequences, from approximately 0.05 ETH for single slots to 0.08 ETH for nine consecutive slots. Within longer sequences, payments per slot show a slight increase, suggesting that builders bid higher for longer sequences or the first slot after a longer sequence. A weak positive autocorrelation is found between subsequent MEV-Boost payments, challenging the hypothesis of alternating periods of low and high MEV. Finally, our comparison of builders during periods of low and high base fee volatility reveals minimal correlation, indicating the absence of builder specialization based on base fee volatility.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

8 extracted references · 4 canonical work pages

  1. [1]

    Clockwork finance: Automated analysis of economic security in smart contracts,

    K. Babel, P. Daian, M. Kelkar, and A. Juels, “Clockwork finance: Automated analysis of economic security in smart contracts,” in 2023 IEEE Symposium on Security and Privacy (SP) . IEEE, 2023, pp. 2499–2516

  2. [2]

    Twap oracle attacks: Easier done than said?

    T. Mackinga, T. Nadahalli, and R. Wattenhofer, “Twap oracle attacks: Easier done than said?” in 2022 IEEE International Conference on Blockchain and Cryptocurrency (ICBC) . IEEE, 2022, pp. 1–8

  3. [3]

    Multi-block MEV

    J. R. Jensen, V. von Wachter, and O. Ross, “Multi-block mev,”arXiv preprint arXiv:2303.04430 , 2023

  4. [4]

    Decentralization of ethereum’s builder market,

    S. Yang, K. Nayak, and F. Zhang, “Decentralization of ethereum’s builder market,”arXiv preprint arXiv:2405.01329, 2024

  5. [5]

    Who wins ethereum block building auctions and why?

    B. Öz, D. Sui, T. Thiery, and F. Matthes, “Who wins ethereum block building auctions and why?”arXiv preprint arXiv:2407.13931, 2024

  6. [6]

    The centralizing effects of private order flow on proposer-builder separation,

    T. Gupta, M. M. Pai, and M. Resnick, “The centralizing effects of private order flow on proposer-builder separation,” arXiv preprint arXiv:2305.19150 , 2023

  7. [7]

    Non-atomic arbitrage in decentralized finance,

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

  8. [8]

    The Garman-Klass volatility estimator revisited

    I. Meilijson, “The garman-klass volatility estimator revisited,”arXiv preprint arXiv:0807.3492 , 2008

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.