Pith. sign in

REVIEW 4 major objections 5 minor 10 references

Measuring DEX Efficiency and The Effect of an Enhanced Routing Method on Both DEX Efficiency and Stakeholders' Benefits

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read Maximizing the new metric STAP removes all DEX arbitrage, the paper proves.

desk verdict A plausible efficiency metric and a useful routing simulation, but the paper's central no-arbitrage theorem has a units error that breaks the abstract's headline claim. read the letter →

arxiv 2508.03217 v1 pith:N4VWIPTL submitted 2025-08-05 q-fin.TR

classification q-fin.TR MSC 90C25
keywords DEXefficiencystandardizedtotalarbitrageprofitconvexoptimizationtokenroutingUniswapV2line-graph-baseddetectionliquidityproviders
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 sets out a systematic measure of decentralized-exchange efficiency: standardized total arbitrage profit (STAP), defined as the maximum arbitrage profit available in a token graph divided by the total value locked in its pools. Its central claim is that the trade order maximizing STAP is itself arbitrage-clearing: after the trade is executed and the retained fees are returned to the pools, no cyclic arbitrage inside the DEX and no price-gap arbitrage against centralized exchanges survives. From this, the authors derive a practical criterion: in an efficient DEX with zero STAP, the monetized value of received target tokens cannot exceed the monetized value of the source tokens, regardless of the routing algorithm. Applied to an 11-token Uniswap V2 graph, the metric shows efficiency falling from June to November 2024, and simulations indicate that a more profitable line-graph routing method improves DEX efficiency and trader welfare, while DFS routing transfers value to liquidity providers. If the proof holds, STAP is both an efficiency benchmark and a tripwire for arbitrage.

What carries the argument

The load-bearing object is the standardized total arbitrage profit, STAP = TAP/TVL, where TAP is the maximum over token prices times net token outflows across all pools, computed as a convex program with CPMM constraints of the form (reserve plus fee-adjusted input minus output) product at least the constant product. The argument turns on the two-phase CPMM trade: phase one executes the swap along the constant-product curve, and phase two adds the retained fee to reserves. The proofs use the Lagrangian of the convex program to show exchange rates align with CEX prices at the optimum, and slope comparisons around the post-fee point show no trade in either direction is profitable after fee reintegration. The line-graph-based routing method supplies the more profitable routing path used in the simulations.

What would settle it

Compute the STAP-maximizing order on a small token graph with exhaustive search over discretized trade sizes and compare with the reported solver output; a case with two distinct optima, or one where executing the reported trade and re-adding fees still leaves a profitable cyclic or DEX-CEX arbitrage path, would falsify the central claim.

Watch

Extended reading notes

Core claim

The central discovery is that total arbitrage profit in a DEX token graph, normalized as STAP = TAP/TVL and maximized subject to CPMM pool constraints, is not just a metric: the maximizing trade itself removes arbitrage. The paper proves that at the optimum, no cyclic arbitrage or DEX-CEX arbitrage remains, and that adding the retained transaction fees back into the pools does not create new arbitrage. The corollary is an efficiency test: when STAP is zero, every trade must deliver target tokens worth no more than the source tokens, and any violation signals arbitrage. The authors then treat STAP as a measurable indicator, observing declining Uniswap V2 efficiency in their sample, and as a tool for comparing routing algorithms.

Load-bearing premise

The argument assumes the STAP maximization problem is convex with a single global optimum that the numerical solver can always reach, but the paper states this without proof and equation (2) actually maximizes TAP, not the ratio STAP.

Editorial extensions

If this is right

  • If STAP is zero, any routing algorithm that returns target tokens worth more than the source tokens is proof that arbitrage exists, so STAP can serve as an arbitrage-detection standard.
  • Executing a STAP-maximizing trade and returning fees to pools leaves a DEX fully arbitraged, making such a trade order a candidate strategy that removes all detected profit.
  • A routing algorithm that finds more profitable paths reduces STAP over time, meaning better routing itself makes the DEX more efficient.
  • Poor routing transfers value from traders to liquidity providers, visible as rising TVL and falling trader proceeds, so trader and LP benefits are in tension.
  • DEX efficiency as measured by STAP is not static: it declined on Uniswap V2 over the sampled period.

Reading between the lines

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

  • Because the corollary gives a per-trade necessary condition for efficiency, STAP could be monitored continuously on-chain: any swap whose monetized output exceeds its input would flag a pool or path for immediate arbitrage scanning, without solving the full optimization.
  • The proof's reliance on fee reintegration suggests the result depends on fees being returned to the same pool; under alternative fee policies, such as a protocol fee taken out of the pool, the no-arbitrage property may fail.
  • The TVL result implies protocol designers face a distributional choice: routing that maximizes trader surplus may lower LP fee income, so 'better routing' depends on whose welfare is weighted.
  • If the convexity assumption fails for larger graphs, STAP's numerical values would need a global-optimization guarantee; testing on graphs beyond 18 pools would show whether the metric remains reliable.
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 / 5 minor

Summary. The paper proposes Standardized Total Arbitrage Profit (STAP), defined as the ratio of total arbitrage profit (TAP) to total value locked (TVL), as a systematic measure of DEX efficiency. It claims that solving the TAP-maximization problem and then reintegrating the retained transaction fees eliminates both DEX-internal cyclic arbitrage and DEX–CEX arbitrage (Theorems 1–4), and infers a corollary that in an efficient DEX no trade can deliver target-token value in excess of source-token value. The empirical part measures STAP on an 11-token, 18-pool Uniswap V2 graph from June 21 to November 8, 2024, finding a decline in efficiency, and simulates 1000 trades to compare the line-graph-based routing method of [1] with a DFS algorithm, concluding that the former improves DEX efficiency and trader welfare while the latter increases TVL.

Significance. If the theoretical claim can be established, STAP would be a principled, optimization-based DEX efficiency metric and the fee-reintegration result would be a useful structural property connecting routing quality to trader and liquidity-provider welfare. The paper makes a genuine empirical effort using real Uniswap V2 pool data and formulates the arbitrage-measurement problem as an optimization problem, which are strengths. However, as written the central proof contains a units/comparison error, and the empirical sections lack statistical support and robustness checks. The results are therefore conditional rather than established.

major comments (4)
  1. [Section 3, Theorem 3 (Eqs. (7)-(8))] The proof of the fee-reintegration claim is not valid as written. First, Eq. (7) sets the constant product after fee reintegration to (a1+λΔa*)b1; from the CPMM relation the correct value is (a1+Δa*)(b1−Δb*) = a1b1 + λΔa*(b1−Δb*), so the equations do not describe the post-fee state. Second, implicit differentiation of Eq. (7) gives dΔa/dΔb|+_E = (a1+Δa*)/(γ(b1−Δb*)) and dΔa/dΔb|−_E = γ(a1+Δa*)/(b1−Δb*), the reciprocals of the expressions printed in Eq. (8). Third, dΔa/dΔb has units of A tokens per B token, whereas PA/PB has units of B tokens per A token, so the inequalities in Eq. (8) compare quantities of incompatible dimensional meaning. The correct argument would compare dΔb/dΔa with PA/PB for the sell-A direction and dΔa/dΔb with PB/PA for the sell-B direction. Because Theorem 4 and Corollary 5 rely on Theorem 3, the paper’s central no-arbitrage claim is not supported as written.
  2. [Section 3, Eq. (2) and Theorem 1] The text asserts without proof that the optimization problem is convex and that its optimal point is unique. Convexity of the feasible set defined by product-form constraints is not demonstrated, and uniqueness is not implied by convexity when the objective is linear (linear programs commonly have non-unique optima). The numerical STAP values in Section 5.1 depend on cvxpy returning a global optimum, but no optimality certificate, solver tolerance, or sensitivity analysis is reported. The authors should either prove the required convexity/optimality properties or soften the claims and verify global optimality numerically.
  3. [Section 5.2 and Figs. 6-9] The routing comparison rests on a single sequence of 1000 randomly generated trades with no stated seed, no repetitions, no confidence intervals, and no statistical tests. In Fig. 6 the LG and DFS standardized-TAP curves separate only near the end of the sequence, yet the text concludes that the line-graph method improves DEX efficiency and trader benefits. The captions of Figs. 6–10 say 'BFS' while the body text says 'DFS', and the precise routing algorithms (linear DFS versus the line-graph method of [1]) are not fully specified. These issues make the empirical conclusions about routing algorithms unsupported.
  4. [Section 4 and Section 5.1] The token graph is selected using arbitrary thresholds: pools are kept only if TVL exceeds ten thousand dollars, nodes with degree less than two are iteratively removed, and pools with the lowest TVL are removed one by one 'until there are only around twenty tokens left,' yet the final graph has 11 tokens and 18 pools. No sensitivity analysis is provided for these choices, so the observed decline in DEX efficiency from June 21 to November 8, 2024, may be an artifact of the particular graph rather than a robust property of Uniswap V2.
minor comments (5)
  1. [Section 3, Theorem 2] The statement that 'tokens’ relative prices in the DEX are equal to those in CEXs in M' is imprecise: with γ<1 there is a bid–ask spread, and the no-arbitrage condition is that the marginal rates in both directions are worse than the corresponding CEX rates, not that the two prices coincide.
  2. [Section 5.2] The simulation description says only 'execute the trade and update the corresponding pools’ reserves'; it should state explicitly whether the retained transaction fee is added to the pool after each trade, since the pool-update rule matters for the TVL and TAP results.
  3. [Section 4] The text says the graph is reduced 'until there are only around twenty tokens left' and mentions keeping about twenty pools, but the actual graph has 11 tokens and 18 pools; these numbers should be reconciled.
  4. [Section 5.1, Fig. 5] Figure 5 plots TAP and standardized TAP on two different y-axes without a legend indicating which curve corresponds to which axis; this makes the trend claims hard to verify.
  5. [Throughout] The acronyms DFS and BFS are used inconsistently: the text says DFS but several figure captions say BFS. Also, Theorems 1 and 2 should be framed as direct consequences of global optimality rather than as separate substantive theorems, to avoid the appearance of circularity with the definition of TAP.

Circularity Check

2 steps flagged · score 6.0 of 10

Definitional core: no-arbitrage at the TAP optimum and the STAP=0 value bound are true by construction; the fee-reintegration theorem is independent (but Eq. (8) appears to contain a reciprocal/units error).

  1. self definitional [Section 3, Eq. (2) and Theorems 1–2]
    "So, the total summation T AP= PA P i(∆ai out − ∆ai in) + PB P i(∆bi out − ∆bi in) + · · ·measures the total arbitrage profit that exists in the corresponding token graph, with T APbeing an abbreviation of the phrase ‘total arbitrage profit’. ... Assuming there is a cyclic arbitrage opportunity, which is A → B → C → A ... the objective function will increase by P · (oup − inp), where P is the corresponding token’s price. This result contradicts to the premise that the objective function is already maximal."

    TAP is, by the paper's own definition, the total arbitrage profit in the token graph, and the optimized objective in Eq. (2) is exactly TAP (not STAP). An 'arbitrage opportunity' here is any feasible trade with positive monetized net profit; such a trade mechanically increases the TAP objective. Hence the theorem 'at the TAP-optimal point there are no arbitrage opportunities' is an immediate consequence of the definition and of global optimality, not a separate result about DEX/CEX price relations. Theorem 2's DEX–CEX no-arbitrage claim is the same tautology for single-pool trades. Consequently the statement that STAP=0 corresponds to full efficiency is built into the construction of STAP rather than derived from it.

  2. self definitional [Section 3, Corollary 5]
    "Corollary 5. In an efficient DEX where no cyclic arbitrage opportunities and no arbitrage opportunities between DEX and CEX, the monetized value of the received target token is less than or at most equal to the monetized value of the source token in token trading, whatever routing methods traders use."

    The corollary is the contrapositive of the definition of arbitrage profit: if the monetized target value exceeds the source value, the trade itself is an arbitrage, so TAP>0 and STAP>0. No routing property or DEX structure is used; the 'regardless of routing' clause is automatic because the statement follows from the definition for any single trade. Thus the corollary reduces to the way TAP/STAP is defined, not to the preceding theorems.

full rationale

The paper's central no-arbitrage claims split into two parts. The first part — that executing the trade order that maximizes the TAP/STAP objective leaves no arbitrage — is definitional rather than derived: TAP is defined as total arbitrage profit, so any remaining arbitrage would, by definition, be a positive-profit feasible trade and would increase the objective. Theorems 1 and 2, and the accompanying claim that STAP=0 signals full efficiency, unpack this definition and global optimality rather than deriving a new market property. Corollary 5 is likewise a restatement of the definition of arbitrage profit. These are genuine self-definitional reductions, and they affect the abstract's headline claim. The second part — Theorem 3's fee-reintegration step — is not circular: it attempts to show, by an independent calculation, that adding the retained fee to the pool creates no new arbitrage. That step has independent content, although as written Eq. (8) compares dΔa/dΔb (units A/B) with PA/PB (units B/A), and implicit differentiation of Eq. (7) yields the reciprocal of the printed fractions; this is a correctness risk rather than a circularity and is not scored here. The empirical STAP measurements and the LG-vs-DFS routing simulation are self-contained numerical exercises and do not reintroduce the theorems as fitted inputs. Overall circularity is partial: the no-arbitrage conclusion after the TAP-maximizing order reduces by construction, while the fee-reintegration theorem and the empirical measurements are independent content.

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

The central theoretical argument rests on the assumption that the optimization problem defining STAP is convex and globally solvable, and that external CEX prices are correct. The empirical results depend on several hand-chosen graph selection thresholds that are not rooted in a formal sampling scheme.

free parameters (4)
  • Pool TVL threshold = $10,000
    Chosen by hand to filter pools; affects token graph composition and subsequent STAP calculations.
  • Token graph size target = ~20 tokens (11 tokens, 18 pools)
    Chosen because cvxpy cannot handle larger graphs; selection of which pools to remove is arbitrary and may bias efficiency measurements.
  • Simulation trade size = $1000 per trade (Ms)
    Chosen for convenience in simulation; affects absolute values but not the relative comparison.
  • Number of simulated trades = 1000
    Arbitrary; no sensitivity analysis or multiple seeds reported.
assumptions (4)
  • domain assumption CPMM invariant with fee (a1+γΔa)(b1-Δb)=a1b1
    Restricts the model to Uniswap V2 style constant-product pools; other AMMs excluded.
  • domain assumption CEX prices from CoinGecko represent true fundamental prices
    Arbitrage profits and STAP are measured against these prices; if prices are stale or wrong, STAP is miscalibrated.
  • ad hoc to paper The STAP/TAP optimization problem is convex with a unique global optimum
    The paper asserts convexity and uniqueness for the ratio objective, but equation (2) maximizes unstandardized TAP; the convexity of TAP/TVL is not proven.
  • domain assumption Retained transaction fees are added back to pools in a second phase
    This matches Uniswap V2 mechanics but is central to Theorems 3 and 4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Measuring DEX Efficiency and The Effect of an Enhanced Routing Method on Both DEX Efficiency and Stakeholders' Benefits." pith.science (2026). https://pith.science/paper/N4VWIPTL

@misc{pith2026250803217,
  author       = {Pith},
  title        = {Pith review of: Measuring DEX Efficiency and The Effect of an Enhanced Routing Method on Both DEX Efficiency and Stakeholders' Benefits},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N4VWIPTL}},
  note         = {Machine review of arXiv:2508.03217}
}
read the original abstract

The efficiency of decentralized exchanges (DEXs) and the influence of token routing algorithms on market performance and stakeholder outcomes remain underexplored. This paper introduces the concept of Standardized Total Arbitrage Profit (STAP), computed via convex optimization, as a systematic measure of DEX efficiency. We prove that executing the trade order maximizing STAP and reintegrating the resulting transaction fees eliminates all arbitrage opportunities-both cyclic arbitrage within DEXs and between DEXs and centralized exchanges (CEXs). In a fully efficient DEX (i.e., STAP = 0), the monetary value of target tokens received must not exceed that of the source tokens, regardless of the routing algorithm. Any violation indicates arbitrage potential, making STAP a reliable metric for arbitrage detection. Using a token graph comprising 11 tokens and 18 liquidity pools based on Uniswap V2 data, we observe a decline in DEX efficiency between June 21 and November 8, 2024. Simulations comparing two routing algorithms-Yu Zhang et al.'s line-graph-based method and the depth-first search (DFS) algorithm-show that employing more profitable routing improves DEX efficiency and trader returns over time. Moreover, while total value locked (TVL) remains stable with the line-graph method, it increases under the DFS algorithm, indicating greater aggregate benefits for liquidity providers.

Figures

Figures reproduced from arXiv: 2508.03217 by the authors.

Figure 1
Figure 1. Token trading curve and two phases during token [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The trading interface of Sushiswap. At around [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. TVL of liquidity pools in the token graph. The [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: Token graph. Each node denotes a token, and each [PITH_FULL_IMAGE:figures/full_fig_p006_3.png]
Figure 7
Figure 7. Figure 7: TAP with the linear DFS and line-graph-based [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 6
Figure 6. Figure 6: Standardized TAP with the linear DFS and line [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 9
Figure 9. Figure 9: Accumulation of all monetized value of the target [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: TVL of all liquidity pools by using the BFS and [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 9 canonical work pages

  1. [1]

    A Line Graph-Based Framework for Identifying Optimal Routing Paths in Decentralized Exchanges

    Y . Zhang, Y . Li, and C. Tessone, “A line graph-based framework for identifying optimal routing paths in decentralized exchanges,” arXiv preprint arXiv:2504.15809, 2025

  2. [2]

    Bitcoin: A peer-to-peer electronic cash system,

    S. Nakamoto, “Bitcoin: A peer-to-peer electronic cash system,” Bitcoin.–URL: https://bitcoin.org/bitcoin.pdf , 2008

  3. [3]

    Transaction flows and holding time scaling laws of bitcoin,

    D. Sornette and Y . Zhang, “Transaction flows and holding time scaling laws of bitcoin,” Physica A: Statistical Mechanics and its Applications, vol. 658, p. 130294, 2025

  4. [4]

    Sok: Decentralized finance (defi)–fundamentals, taxonomy and risks,

    K. Gogol, C. Killer, M. Schlosser, T. Bocek, B. Stiller, and C. Tessone, “Sok: Decentralized finance (defi)–fundamentals, taxonomy and risks,” arXiv preprint arXiv:2404.11281 , 2024

  5. [5]

    Cyclic arbitrage in decentralized exchanges,

    Y . Wang, Y . Chen, H. Wu, L. Zhou, S. Deng, and R. Wattenhofer, “Cyclic arbitrage in decentralized exchanges,” in Companion Proceed- ings of the Web Conference 2022 , 2022, pp. 12–19

  6. [6]

    A large scale study of the ethereum arbitrage ecosystem,

    R. McLaughlin, C. Kruegel, and G. Vigna, “A large scale study of the ethereum arbitrage ecosystem,” in 32nd USENIX Security Symposium (USENIX Security 23) , 2023, pp. 3295–3312

  7. [7]

    Global order routing on exchange networks,

    V . Danos, H. E. Khalloufi, and J. Prat, “Global order routing on exchange networks,” in Financial Cryptography and Data Security. FC 2021 International Workshops: CoDecFin, DeFi, VOTING, and WTSC, Virtual Event, March 5, 2021, Revised Selected Papers 25 . Springer, 2021, pp. 207–226

  8. [8]

    An improved algorithm to identify more arbitrage opportunities on decentralized exchanges,

    Y . Zhang, T. Yan, J. Lin, B. Kraner, and C. J. Tessone, “An improved algorithm to identify more arbitrage opportunities on decentralized exchanges,” in 2024 IEEE International Conference on Blockchain and Cryptocurrency (ICBC). IEEE, 2024, pp. 1–7

Show all 10 references
  1. [9]

    An empir- ical study of market inefficiencies in uniswap and sushiswap,

    J. A. Berg, R. Fritsch, L. Heimbach, and R. Wattenhofer, “An empir- ical study of market inefficiencies in uniswap and sushiswap,” arXiv preprint arXiv:2203.07774, 2022. Appendix

  2. [10]

    Here, we propose the matrix analysis to measure the DEX efficiency

    Calculating Tokens’ Price in DEX Using Matrix Analysis and Measuring the DEX Efficiency by the Prices’ Difference against CEX One problem with the convex optimization method is its high computational complexity and its inaccuracy when the problem scale is large. Here, we propo...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.