{"id":"f42823b3-a019-4512-b4e0-888238b9b7da","arxiv_id":"2501.12576","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A block-size choice is shown to restore near-optimal social welfare in blockchain order books where miners myopically chase fees, with exact optimality for homogeneous-quantity (NFT) trading.","lead":"This paper models blockchain order books as a game among buyers, sellers, and fee-seeking miners, and shows that selfish miners can cause arbitrarily large welfare losses. It proposes choosing the block size so that the welfare loss becomes bounded, and for NFT-style homogeneous quantities it claims the loss disappears entirely.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3's zero-fee equilibrium profiles contradict the paper's own rule that miners reject zero-fee transactions, so the claimed PoA=1 (and PoA≤b/b) is not established for the regimes where equations (92)-(100) set fees to zero.","rationale":"The paper's headline results are equilibrium statements: Theorems 3 and 4 compute social welfare at the Stage I fee-setting equilibria. The reader's weakest assumption correctly identifies that the equilibrium construction is load-bearing, and the specific flaw in Theorem 3 Case 3 is real and internal to the model. The same zero-fee issue also appears in Case 2 subcases (b) and (c), where one side of every pair is assigned zero fees; under the paper's own rule that miners reject zero-fee transactions, the assumed matching cannot occur. This is a correctness risk in the proof of the central claim, not a disagreement with external consensus. It does not automatically mandate rejection, because a repair may exist: for example, setting positive epsilon fees on all matched pairs might restore the equilibrium and the welfare bound. But the paper does not supply such a repair, and the implementation section only demonstrates that the gas limit can be changed, not that the constructed fee equilibria are valid. Therefore the reader's CONDITIONAL verdict remains appropriate: the central claim is not fully supported as written, but the issue is plausibly fixable. A concrete numerical test would settle whether the natural epsilon repair works.","tokens_in":50090,"tokens_out":7016,"duration_ms":77627,"concrete_test":"Instantiate Theorem 3 Case 3 with K=N=2, j=1, R1=0.7, R2=0.2, C1=0.3, C2=0.8, so R2<C1 and C2>R1. Equations (99)-(100) then prescribe all fees equal to zero. Compute the miner's optimal selection in (16) under this profile: since every fee is zero and miners reject zero-fee transactions, no pair is included, giving sw=0 while swopt=R1-C1=0.4. Then test the natural repair: set f_buy,1=f_sell,1=ε and f_buy,2=f_sell,2=0, and verify whether this is a valid Stage I equilibrium under the stated tie-breaking and zero-fee rejection rules. If the repair is an equilibrium, Theorem 3 can be fixed; if not, a different construction is required before the PoA claims can stand.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that setting A=Ath yields PoA≤b/b, with PoA=1 for homogeneous quantities. The proof evaluates this claim at fee profiles constructed in Theorem 3. Several of those profiles set fees to zero on at least one side: Case 2(b) sets f_sell,n=0 for all sellers (eq. 93), Case 2(c) sets f_buy,k=0 for all buyers (eq. 95), and Case 3 sets both sides to zero when R_{j+1}<C1 and C_{j+1}>R1 (eqs. 99-100). Section II-A3 states that miners reject zero-fee transactions. Under that rule, the proposed profiles are not Nash equilibria of the Stage I game: in Case 3 no transaction has positive fee, so the miner's Stage II problem (16) selects nothing, and welfare is zero rather than the claimed social optimum. In Cases 2(b) and 2(c), one side being zero still prevents the assumed matching if zero-fee transactions are individually rejected. Since these cases cover parameter regimes of the mechanism, the central claim is unsupported as written until valid fee profiles are supplied or the zero-fee rejection assumption is explicitly relaxed.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes blockchain-based order books as a two-stage game in which buyers and sellers set transaction fees and a selfish miner matches transactions. It claims the first analytical quantification of the efficiency loss caused by selfish miners, proving an unbounded price of anarchy (PoA), and then proposes an adjustable block size mechanism. The main theoretical results are Theorem 3, which states that setting block size A = Ath gives PoA ≤ b/b in general and PoA = 1 for homogeneous quantities, and Theorem 4, which extends the bound to an incomplete-information limit with N → ∞. The paper also reports an Ethereum testbed implementation and experiments on NFT and Bitcoin order-book datasets, claiming social optimum for homogeneous-quantity trading and up to 3.7× welfare improvement over a benchmark.","tokens_in":50295,"tokens_out":11107,"duration_ms":124617,"significance":"The problem is relevant and the proposed mechanism is attractively simple: it only adjusts a protocol parameter and does not require changing the matching protocol. If the main theorems were fully correct, the paper would be a solid contribution to mechanism design for decentralized exchanges. The Ethereum implementation and the use of real datasets are genuine strengths, as is the attempt to handle mixed-strategy equilibria. However, the proof gaps discussed below directly affect the central claims, so the paper is not yet at the standard required for acceptance.","major_comments":[{"comment":"Theorem 3's proof constructs Stage I equilibria in which all fees on at least one side are zero: Case 2(b) sets f_sell,n = 0 for all sellers (eq. (93)), Case 2(c) sets f_buy,k = 0 for all buyers (eq. (95)), and Case 3 sets all fees to zero when R_{j+1} < C1 and C_{j+1} > R1 (eqs. (99)-(100)). This contradicts the model assumption in Section II-A3 that miners reject zero-fee transactions. In the all-zero regime of Case 3, the Stage II problem (16) selects no transaction at all, so the resulting social welfare is 0, not the claimed swopt in eq. (101). The same problem invalidates the assumed matching in Cases 2(b) and 2(c), because zero-fee buyers or sellers cannot be included in a miner's selection. Since these regimes are explicitly covered by Theorem 3, the claims PoA = 1 and PoA ≤ b/b are unsupported as written. The authors need to supply valid strictly-positive fee equilibria with a complete no-deviation check, or explicitly relax the zero-fee rejection assumption and recompute the welfare consequences.","section":"Section VII-D-1 (Theorem 3 proof), eqs. (92)-(100)"},{"comment":"The proof of Theorem 4 relies on asymptotic approximations for which the required regularity conditions are not stated in the theorem. Equation (125) defines ρ = K/N and the subsequent order-statistics approximations (e.g., eqs. (130)-(131), (178)-(179), (183)-(186), (196)-(199)) assume that ρ is a fixed positive constant and that the relevant quantiles are interior points where the densities of R and C are positive and continuous. The theorem statement only says 'N → ∞' and does not state that K scales with N, nor does it state conditions such as C(η) + N^{-ψ} < 1 and C(η) - N^{-ψ} > 0 for all sufficiently large N. Without these assumptions, the normal approximations, the variance formulas, and the limits in eqs. (44)-(45) are not justified. The theorem should either include explicit distributional and scaling assumptions or prove the bounds by a different method that is robust to the unstated cases.","section":"Section VII-D-2 (Theorem 4 proof), eqs. (125)-(151)"},{"comment":"Theorem 4 inherits the zero-fee equilibrium problem from Theorem 3 because its welfare computation evaluates the social welfare at the complete-information fee equilibria of Propositions 1 and 2 for the realized buyer and seller types. If those equilibria are not valid in the zero-fee regimes identified above, then the social-welfare lower bounds in the proof of Theorem 4 (e.g., eqs. (204)-(218)) are not grounded. Thus Theorem 4 cannot be accepted until the equilibrium issue raised in the first major comment is resolved.","section":"Sections IV-B and VII-D-2 (Theorem 4)"}],"minor_comments":[{"comment":"In the high-block-size example, the text says the social optimum is to match R1 with C2, but the displayed inequalities imply that the optimum match is R1 with C1; the ratio should be (R1 - C1)/(4ϵ) rather than (R1 - C2)/(4ϵ). In addition, the limit in eq. (88) should be taken as d approaches (R1+R2-C1-C2)/2 from below, not from above. These are typos, but they should be corrected because they obscure the construction.","section":"Section VII-C, proof of Theorem 2"},{"comment":"The notation b/b requires the lower bound b on trading quantities to be strictly positive. The model only states bk, qn ∈ [b,b] and does not explicitly assume 0 < b ≤ b; without this assumption, the bound PoA ≤ b/b is not well-defined. Please add the strict positivity assumption explicitly.","section":"Section II-A-2, Theorems 3 and 4"},{"comment":"There is an inconsistency in the upper support of the mixed-strategy distributions: eq. (35c) and eq. (36c) use (⌈ min{...}/A⌉ - 1)d, while Lemma 2.4 and the derivation around eqs. (79)-(80) use ⌈ min{...}/A⌉ d. This discrepancy affects the explicit form of the CDF and should be reconciled.","section":"Equations (35), (36) and Lemma 2.4"},{"comment":"The proof of Theorem 1 is only a short sketch and does not fully justify why the argmax in (24) must be a top-i fee set in the presence of constraint (5). Since the later analysis verifies candidate equilibria under this selection rule, I do not treat this as a blocking issue, but the proof should be expanded.","section":"Proof of Theorem 1"}],"recommendation":"major_revision","confidential_remarks":"The paper has a plausible and practically motivated mechanism, and the Ethereum testbed implementation is a real strength. The main obstacle is the zero-fee equilibrium construction in Theorem 3, which is load-bearing for both Theorem 3 and Theorem 4. I would advise the editor that the manuscript can be reconsidered after the authors supply valid fee equilibria or explicitly change the zero-fee rejection assumption, and after Theorem 4 is restated with the missing scaling and regularity assumptions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The core idea is genuinely good: set the block size to the number of profitable buyer-seller pairs, and a simple protocol parameter fixes the welfare disaster caused by myopic miners. The paper is the first analytical treatment of selfish miners in blockchain order books, and the infinite-PoA diagnosis is nicely constructed. The game-theoretic setup with PSNE and MSNE, plus the incomplete-information extension, is substantial. The experiments with real NFT and Bitcoin data are a useful sanity check, and the reported 3.7x improvement over the benchmark is plausible if the mechanism works.\n\nThe soft spot is load-bearing. Theorem 3's proof constructs fee profiles with zero fees on one or both sides in several cases (equations 93, 95, 99-100). But Section II-A3 explicitly states that miners reject zero-fee transactions. Under the paper's own rule, those profiles are not Nash equilibria: in Case 3 with all-zero fees, no trade occurs and welfare is zero, not the claimed social optimum. The PoA=1 and PoA≤b/b results are therefore not established as written. This is fixable—one can likely supply positive fees or relax the zero-fee assumption—but it cannot be ignored.\n\nOther issues are smaller. Theorem 4 uses asymptotic normal approximations for order statistics without stating regularity conditions on the CDFs; the proof has some algebraic leaps around the variance terms. The implementation section demonstrates adjusting the gas limit on Ethereum, which shows feasibility of block-size changes, but it does not test the matching equilibrium itself. The citation pattern looks fair; self-citations are not a problem here. There is no circularity: the PoA bounds come from equilibrium analysis, not from fitting parameters to the welfare results.\n\nWho should read this: anyone working on mechanism design for DEXs, NFT markets, or blockchain transaction-fee design. The paper deserves a careful referee despite the flaw, because the central idea is important and the error appears localized and repairable. My recommendation: send it to peer review, but require a corrected Theorem 3 with valid fee profiles (or an explicit relaxation of the zero-fee rejection rule) before acceptance.","headline":"Real mechanism-design idea and substantial analysis, but the PoA theorem relies on zero-fee equilibria that contradict the paper's own zero-fee rejection rule.","tokens_in":50874,"tokens_out":1792,"would_cite":true,"duration_ms":21649,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91A10","91B26","91A80"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper argues that a simple choice of block size — the largest index with buyer utility at least seller cost — bounds the price of anarchy by the quantity ratio and makes homogeneous-quantity trading (NFTs) socially optimal.","keywords":["blockchain order book","price of anarchy","selfish miners","mechanism design","adjustable block size","mixed-strategy Nash equilibrium","social welfare","NFT trading"],"falsifier":"A concrete check: in the Case 3 regime $R_{j+1} < C_1$, equations (99)-(100) set all fees to zero; since the model assumes miners reject zero-fee transactions, the resulting outcome has no trades, so the profile is not a Nash equilibrium unless a positive-fee equilibrium is constructed. Simulating this two-buyer two-seller regime under $A = A_{\\mathrm{th}}$ would settle whether the PoA bound holds there.","tokens_in":49829,"feed_emoji":"⛓","tokens_out":4901,"duration_ms":49218,"temperature":0.7,"pith_summary":"Blockchain order books let miners choose which buyer-seller pairs to match, and miners maximize fees, not social welfare. This paper tries to prove that this can drive social welfare arbitrarily far from optimal — price of anarchy infinite — and that a one-parameter fix restores performance: set the block size to the number of mutually profitable matches. If the analysis is right, a system designer who knows only the distribution of utilities, costs, and quantities can choose a block size so that the worst equilibrium is within a bounded factor of the optimum, and exactly optimal when every trade involves the same quantity, as in NFT markets. The paper also reports an Ethereum testbed implementation and dataset experiments supporting these bounds.","feed_headline":"Right block size makes NFT order books socially optimal","feed_subtitle":"A simple block-size rule bounds selfish-miner welfare loss and hits the optimum for NFT-style trades.","key_machinery":"The load-bearing object is the threshold block size $A_{\\mathrm{th}}$, the number of buyer-seller pairs for which the buyer's utility is at least the seller's cost. It is paired with a two-stage game: in Stage I buyers and sellers choose fees (pure Nash equilibrium when $A \\geq A_{\\mathrm{th}}$, mixed-strategy equilibrium when $A < A_{\\mathrm{th}}$), and in Stage II miners select the top-fee feasible matching under block size $A$. The PoA bounds follow from evaluating social welfare at these fee equilibria and comparing with the welfare-maximizing matching, using the fact that the threshold block size admits only profitable pairs into the first block.","core_discovery":"Under the proposed adjustable block size (ABS) mechanism, the system designer sets $A = A_{\\mathrm{th}}$, where $A_{\\mathrm{th}}$ is the largest $i$ with $R_i \\geq C_i$ (or $\\min\\{K,N\\}$ if all pairs are profitable). Theorem 3 states that for complete information the price of anarchy is at most $\\bar{b}/\\underline{b}$, where $\\underline{b}$ and $\\bar{b}$ are the lower and upper bounds on traded quantities, and exactly $1$ when $\\underline{b} = \\bar{b}$ (homogeneous quantities). Theorem 4 extends this to incomplete information: with i.i.d. values, using $A^* = \\lfloor N(C(\\eta) + N^{-\\psi}) \\rfloor$ gives the same bounds as $N \\to \\infty$. The social optimum is restored for homogeneous-quantity trading because the threshold block size excludes exactly the unprofitable pairs that a fee-maximizing miner would otherwise include to collect extra fees.","pith_inferences":["The paper's threshold rule assumes the fee-setting equilibrium in every regime; the zero-fee case of Theorem 3 ($R_{j+1} < C_1$) is the regime most likely to need a tie-breaking or minimum-fee rule in practice.","An implicit consequence is that the price-of-anarchy bound degrades linearly with quantity dispersion $\\bar{b}/\\underline{b}$, so the mechanism is most attractive in markets where order sizes are similar; for high dispersion, quantity-aware matching rules would be a natural next step.","The asymptotic result suggests a testable prediction: in large NFT-like markets, the observed social welfare ratio should approach 1 as the number of sellers grows, and the benchmark gap should persist.","The mechanism's robustness to non-selfish miners is only tested empirically at 20% cooperation; a natural extension is to characterize the PoA as a function of the fraction of protocol-following miners."],"forward_implications":["If Theorem 3 is correct, any blockchain order book that sets block size to $A_{\\mathrm{th}}$ guarantees at least a $\\underline{b}/\\bar{b}$ fraction of optimal social welfare even with fully selfish miners.","For homogeneous-quantity markets such as NFT trading, the mechanism is exactly socially optimal, not merely approximately so.","If Theorem 4 is correct, the designer needs no individual bid/ask data — only the distributions — and the same guarantees hold asymptotically.","The mechanism changes only the block size (gas limit in Ethereum), so it can be deployed without altering consensus, matching, or fee rules."],"supporting_citations":[{"why":"Supplies the idea that block size is a controllable mechanism parameter in Bitcoin-like protocols.","marker":"[15]"},{"why":"Basis for the assumption that miners reject zero-fee transactions, which is load-bearing in the fee equilibria.","marker":"[16]"},{"why":"Empirical basis for myopic miner behavior in Bitcoin.","marker":"[25]"},{"why":"Further support for selfish transaction ordering by miners.","marker":"[26]"},{"why":"Provides the mining-power distribution used to justify myopia and the probability weights in miner payoffs.","marker":"[27]"},{"why":"NFT dataset used for homogeneous-quantity experiments.","marker":"[37]"},{"why":"Bitcoin order book dataset used for heterogeneous-quantity experiments.","marker":"[38]"},{"why":"Benchmark BBOB matching algorithm that the mechanism is compared against.","marker":"[40]"},{"why":"Ethereum client modified to implement the gas-limit change in the testbed.","marker":"[41]"}],"fun_headline_variants":["Block-size rule neutralizes selfish miners, hits social optimum for NFTs","One block-size knob restores fairness: PoA of 1 for NFT order books","Selfish miners lose: adjustable block size yields zero welfare gap","Blockchain order books: block size tweak boosts welfare 3.7x"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire price-of-anarchy bound rests on the Stage I fee-setting equilibria in Propositions 1 and 2 being valid for every parameter regime; in the regime of Theorem 3 where $R_{j+1} < C_1$, the proposed all-zero fees contradict the assumption that miners reject zero-fee transactions, so no valid trade occurs unless another equilibrium is supplied.","fun_headline_variants_meta":{"raw":{"variants":["Block-size rule neutralizes selfish miners, hits social optimum for NFTs","One block-size knob restores fairness: PoA of 1 for NFT order books","Selfish miners lose: adjustable block size yields zero welfare gap","Blockchain order books: block size tweak boosts welfare 3.7x"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000436,"raw_usage":{"total_tokens":2275,"prompt_tokens":1059,"completion_tokens":1216,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":675,"completion_tokens_details":{"reasoning_tokens":1135}},"tokens_in":675,"tokens_out":1216,"duration_ms":12565,"temperature":1.0,"reasoning_tokens":1135,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:04:21.366748+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete check: in the Case 3 regime $R_{j+1} < C_1$, equations (99)-(100) set all fees to zero; since the model assumes miners reject zero-fee transactions, the resulting outcome has no trades, so the profile is not a Nash equilibrium unless a positive-fee equilibrium is constructed. Simulating this two-buyer two-seller regime under $A = A_{\\mathrm{th}}$ would settle whether the PoA bound holds there.","supporting_citations":[{"cited_title":"Bitcoin xt,","cited_arxiv_id":null,"evidence_quote":"Supplies the idea that block size is a controllable mechanism parameter in Bitcoin-like protocols."},{"cited_title":"Ethereum fee,","cited_arxiv_id":null,"evidence_quote":"Basis for the assumption that miners reject zero-fee transactions, which is load-bearing in the fee equilibria."},{"cited_title":"Bitcoin protocol,","cited_arxiv_id":null,"evidence_quote":"Empirical basis for myopic miner behavior in Bitcoin."},{"cited_title":"Selfish & opaque transaction ordering in the bitcoin blockchain: the case for chain neutrality,","cited_arxiv_id":null,"evidence_quote":"Further support for selfish transaction ordering by miners."},{"cited_title":"Ethereum staking distribution,","cited_arxiv_id":null,"evidence_quote":"Provides the mining-power distribution used to justify myopia and the probability weights in miner payoffs."},{"cited_title":"Nft trading dataset,","cited_arxiv_id":null,"evidence_quote":"NFT dataset used for homogeneous-quantity experiments."},{"cited_title":"Bitcoin order book dataset,","cited_arxiv_id":null,"evidence_quote":"Bitcoin order book dataset used for heterogeneous-quantity experiments."},{"cited_title":"Blockchain order book protocol,","cited_arxiv_id":null,"evidence_quote":"Benchmark BBOB matching algorithm that the mechanism is compared against."},{"cited_title":"Ethereum clients,","cited_arxiv_id":null,"evidence_quote":"Ethereum client modified to implement the gas-limit change in the testbed."}],"review_version":1}