Pith. sign in

REVIEW 2 major objections 4 minor 30 references

A Truckload of Satoshis: Detecting and Measuring One-Way Arbitrage in the Wild

T0 review · 2 major / 4 minor · reviewed 2026-07-13 · grok-4.5

Pith's one-line read One-way arbitrage is real on big crypto exchanges: 402 million sequences on Binance alone, but average net profit is pennies.

desk verdict First large-scale recovery of executed OWA from anonymized CEX trades; absolute counts are threshold-dependent, but the existence and maturation trends look solid. read the letter →

arxiv 2607.09491 v1 pith:LR7AMITM submitted 2026-07-10 cs.CE

classification cs.CE
keywords one-wayarbitragecentralizedexchangesspottradedatacryptocurrencymarketsalgorithmictradingdetectionpricediscrepanciesBinanceKraken
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

Centralized crypto exchanges publish anonymized spot trades with no trader IDs, so no one knew whether algorithmic strategies such as one-way arbitrage (converting A to B via a cheaper intermediate coin I) actually occur. This paper supplies a practical detection method that matches pairs of trades by nearly identical intermediate quantities and very short time gaps, then keeps only the parameter ranges that are profitable in aggregate. Applied to five-plus years of Binance data and nine years of Kraken data, the method recovers hundreds of millions of likely sequences that account for roughly one percent of Binance volume and a tenth of a percent of Kraken volume. Aggregate net profits after the lowest published fees reach tens of millions of dollars, yet the typical sequence earns well under a dollar; over time the trades have become both faster and less profitable. The result shows that pricing discrepancies are routine on CEXes and that professional, high-speed bots already harvest them at scale.

What carries the argument

Quantity-and-timing matching refined by aggregate profitability: candidate pairs must exchange nearly the same amount of the intermediate coin (within 0.1 basis points) inside a short latency window (200–400 ms on Binance, 1–2 s on Kraken), and only the latency bins that remain net-positive after fees are retained.

What would settle it

If an exchange later released even a small sample of trades tagged by trader identifier, one could count how many of the paper’s matched pairs actually share the same identifier; a low match rate would falsify the detection method.

Watch

Extended reading notes

Core claim

One-way arbitrage is an actively used trading strategy on major centralized exchanges: the authors recover 402 million likely sequences on Binance (0.94 percent of total volume, roughly $31 million net profit after lowest published fees) and almost two million on Kraken (0.13 percent volume, $975 thousand), with individual sequences averaging less than one dollar net and with clear trends toward lower latency and smaller price gaps.

Load-bearing premise

The method treats any pair of trades that match the chosen quantity and latency windows and that are profitable in bulk as belonging to the same rational one-way-arbitrage trader; there is no ground-truth identity data to check that claim.

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

2 major / 4 minor

Summary. The paper proposes a quantity-and-timing matching methodology to detect one-way arbitrage (OWA) sequences in anonymized CEX spot trade data. After coarse filtering on coin types, quantity difference, latency, and order type, the authors refine thresholds on a development set by retaining parameter regions where volume and hypothetical gains concentrate and profit dominance is positive (Section 4.4, Figs. 3–5). Applying the resulting cut-offs (Δq ∈ [−0.1, 0.1] bps; latency 200/400 ms on Binance and 1–2 s on Kraken) yields 402 M sequences on Binance (0.94 % of volume, estimated $31.2 M net profit after lowest published fees) and ~2 M on Kraken (0.13 % volume, $975 k). Per-sequence net profits average well under $1; longitudinally, latencies fall and per-sequence gross returns compress. The authors interpret the activity as professional, competitive OWA that helps rebalance order books.

Significance. If the inference is accepted, the work supplies the first large-scale measurement of executed (rather than merely theoretical) OWA on major CEXes, together with two longitudinal data sets that the authors intend to release. The scale (hundreds of millions of sequences, nearly $190 B volume on Binance) and the clear trends toward lower latency and smaller price discrepancies are of genuine interest to market-microstructure and crypto-finance audiences. The methodology itself—matching on quantities and timestamps when trader identifiers are absent—is a reusable contribution for other anonymized order-flow studies. The explicit prioritization of precision over recall and the transparent reporting of fee-scenario ranges are strengths that make the measurement usable even if absolute counts remain somewhat threshold-dependent.

major comments (2)
  1. Sections 4.4–4.5 and Figs. 3–5: the quantity-difference window and latency cut-offs are chosen by inspecting the same volume/gains/losses CDFs and profit-dominance plots that later become the headline results. Because the retained region is defined as the region where aggregate gains exceed losses, the absolute counts (402 M sequences), volume share (0.94 %), and net-profit totals ($31.2 M) are not independent of the detection rule. The paper correctly notes the absence of ground truth and the precision-over-recall stance, yet still presents the tuned numbers as the primary measurement of OWA prevalence. At minimum the authors should (i) report sensitivity of the headline figures to modest changes in the cut-offs, (ii) show that the longitudinal trends (speed-up, return compression) survive alternative thresholds, and (iii) clearly separate the development-set selection step from the fin
  2. Section 5.1 and the fee assumptions: net-profit estimates rest on the lowest published fee tiers (2.4 bps taker / 1.2 bps maker on Binance; 8 bps / 0 bps on Kraken) and on the modeling choice that OWA must cover two fees while the direct-conversion reference price incurs none. The paper itself shows that under this “double-fee” scenario taker/taker cumulative profits on Binance stagnate or decline for long periods while volume remains high—an outcome the authors attribute to possible private fee arrangements. Because the absolute dollar figures are sensitive to these unobservable fee levels, the $31.2 M / $975 k totals should be presented more explicitly as a lower-bound range under stated assumptions rather than as point estimates of realized profit.
minor comments (4)
  1. Section 4.3: the initial Δq intervals ([−10, 1] bps Binance, [−40, 1] bps Kraken) are motivated by published fee schedules, yet the final analysis discards all in-band-fee candidates. A short quantitative note on how much volume is thereby excluded would help readers gauge the precision–recall trade-off.
  2. Figures 7 and 12: the quarterly box-plots of gross returns would be clearer if the y-axis scales were identical across maker/taker and across exchanges, facilitating direct visual comparison of the claimed compression.
  3. Section 3.1: the mark-to-market procedure for Binance (BTC VWAP then Kraken BTC/USD) is reasonable but should note any residual basis risk when the two exchanges temporarily diverge.
  4. Table 2 and the surrounding text: the final latency thresholds (200/400 ms Binance, 1–2 s Kraken) are rounded upward from the profit-dominance zero-crossings; stating the exact zero-crossing values alongside the rounded cut-offs would improve reproducibility.

Circularity Check

3 steps flagged · score 5.0 of 10

Detection thresholds for OWA sequences were chosen by maximizing the same aggregate profitability metric later reported as the main result, so absolute counts and profits partly follow by construction.

  1. fitted input called prediction [Section 4.4 (Refining Matching Criteria) and 4.4.1–4.4.2]
    "Our hypothesis is that we can separate likely OWA activity from unrelated background noise by looking for trading parameters co-occurring with a higher concentration of volume and profits, since we assume that rational traders conduct OWA under parameters that make it profitable in aggregate. We explore (1) Δq ... and (2) Δt ... We plot these two parameters of OWA candidates in our development set against the corresponding volume and profitability to determine ranges with a higher incidence of volume and profits. ... Since we have no certainty about which two operations constitute OWA by the s"

    The quantity-difference window that defines the final OWA data set is chosen precisely because that window maximises the concentration of volume and of positive gross returns on the development set. The subsequent claim that the retained sequences are profitable OWA therefore re-uses the same profitability signal that was used to select them; the absolute counts and dollar profits are not an independent measurement of an externally defined population.

  2. fitted input called prediction [Section 4.4.2 (Latency) and Table 2 / Section 4.5]
    "To select a latency threshold, ultimately we would like to know whether hypothetical gains exceed losses in specific latency bins ... we introduce a new profit dominance metric ... Figure 5 shows a very clear picture for Binance: All 5 ms latency bins up to 180 ms for taker/taker (and 325 ms for maker/taker) are positive, whereas nearly all remaining bins are negative. ... we discard OWA candidates with higher latencies on Binance. To avoid overfitting, we round these thresholds up to 200 ms and 400 ms, respectively. ... Table 2 summarizes the selected thresholds ... This results in more than"

    Latency cut-offs are set at the points where the profit-dominance statistic (signed gross return normalised by absolute return) changes from positive to negative. The sequences that survive these cut-offs are then reported as the main empirical result (402 M sequences, $31.2 M net profit). Because the retention rule is defined by aggregate profitability, the claim that the retained set is profitable OWA is true by construction of the filter; the absolute scale is therefore not an independent prediction.

1 more flagged steps
  1. fitted input called prediction [Section 5 (Analysis) opening and 5.1 (Profits)]
    "Our data sets show that OWA is an actively used trading strategy. The volume of the detected OWA sequences reaches nearly $190 B ... This corresponds to approximately 0.94 % and 0.13 % of the total traded volume ... We estimate gross profits of $73.3 M on Binance and below $2.0 M on Kraken ... Accounting for fees leads to a sharp reduction in profits. Total profits drop ... to $31.2 M when two fees are paid on Binance, and from $2 M to $975 k on Kraken."

    The headline prevalence and profit figures are computed exclusively on the profitability-tuned subset produced in Section 4.5. Because that subset was retained precisely for exhibiting positive aggregate gains, the reported totals restate the selection criterion rather than independently measuring an externally defined OWA population. Longitudinal trends are less affected, but the absolute numbers inherit the circular dependence.

full rationale

The paper's central measurement (402 M sequences, 0.94 % of volume, $31.2 M net profit on Binance; analogous figures on Kraken) rests on a two-stage heuristic that first generates candidates by coarse quantity/latency/coin-type filters and then retains only those parameter regions where volume and gains concentrate and profit dominance is positive. Sections 4.4–4.5 explicitly state that the final Δq ∈ [−0.1, 0.1] bps and the latency cut-offs (200/400 ms Binance, 1–2 s Kraken) were selected by inspecting CDFs of volume/gains/losses and profit-dominance plots on a development set and keeping the region that is profitable in aggregate. Because the same profitability signal is both the selection criterion and the headline result, the absolute scale of the reported OWA activity is not independent of the detection rule. The paper is transparent about the absence of ground truth and its precision-over-recall priority, and the longitudinal trends (speed-up, compression of price discrepancies) are less sensitive to the exact cut-offs; nevertheless the absolute prevalence and profit numbers reduce, by construction, to the profitability-tuned thresholds. No self-citation load-bearing chain or uniqueness theorem is involved; the circularity is purely of the fitted-input-called-prediction variety. Score 5 reflects partial circularity confined to the absolute scale claims while leaving the existence of the trading pattern and its temporal evolution with independent content.

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

The central measurements rest on a trader model (profit motive, avoidance of non-anchor coins, lowest-tier fees, high-speed competition) plus a set of numerical matching thresholds that were fitted to the profitability surface of the candidate set itself. No new physical or mathematical entities are postulated; OWA is taken from Deardorff (1979).

free parameters (3)
  • quantity-difference window Δq = [-0.1, 0.1] bps
    Set to [−0.1, 0.1] bps after inspecting CDFs of volume/gains/losses on a development sample (Section 4.4.1); the window directly determines which pairs enter the final profit calculation.
  • latency cut-offs Δt = 200–400 ms Binance; 1–2 s Kraken
    Chosen as 200 ms / 400 ms (Binance) and 1 s / 2 s (Kraken) by inspecting profit-dominance plots so that retained bins remain net-positive (Section 4.4.2).
  • lowest published fee tiers = Binance 1.2/2.4 bps; Kraken 0/8 bps
    Assumed for all OWA traders (2.4 bps taker / 1.2 bps maker on Binance; 8 bps / 0 bps on Kraken) without per-trader verification; net-profit figures are highly sensitive to this choice.
assumptions (3)
  • domain assumption OWA traders are rational, profit-driven, avoid holding non-anchor coins, operate at high volume (lowest fees), and compete on speed.
    Stated as the trader model in Section 4.1; every subsequent filter (coin types, quantity match, latency, order type) is derived from it.
  • ad hoc to paper Pairs of trades satisfying the quantity, latency and profitability filters are executed by the same trader.
    Core identification assumption of Section 4; no ground-truth identifiers exist to test it.
  • domain assumption Reference price for direct conversion is the 2 s VWAP immediately preceding the first leg.
    Defined in Section 4.4; used for every gross-return calculation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Truckload of Satoshis: Detecting and Measuring One-Way Arbitrage in the Wild." pith.science (2026). https://pith.science/paper/LR7AMITM

@misc{pith2026260709491,
  author       = {Pith},
  title        = {Pith review of: A Truckload of Satoshis: Detecting and Measuring One-Way Arbitrage in the Wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LR7AMITM}},
  note         = {Machine review of arXiv:2607.09491}
}
abstract

Centralized cryptocurrency exchanges (CEXes) enable fast off-chain conversions between hundreds of coins. It is an open question which algorithmic trading patterns occur on these platforms. A major challenge to measuring CEXes is that their public trade data does not contain addresses or trader identifiers allowing linkage. We propose a novel methodology to infer one-way arbitrage (OWA) trading in anonymized spot trade data from CEXes. We identify 402 M likely OWA sequences in 5 years of trading on Binance (and almost 2 M during 9 years on Kraken), accounting for 0.94 % and 0.13 % of the total traded volume, respectively. While we estimate total profits of $31.2 M on Binance and $975 k on Kraken, profits from individual OWA sequences are less than $1 on average after accounting for trading fees. We also observe that OWA has become faster over time, while the profitability of individual sequences has decreased. Our findings highlight that pricing discrepancies regularly occur in CEXes, and raise questions for future work to identify the precise circumstances that enable profitable OWA.

Figures

Figures reproduced from arXiv: 2607.09491 by the authors.

Figure 1
Figure 1. One-way arbitrage (OWA) vs. direct conver [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Monthly trade volume in USD (blue bars, left y axis) and active currency pairs (black line, right y axis). [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. CDFs of the accumulated volume and hypothetical gains and losses of OWA candidate sequences on [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: CDFs of the accumulated volume and hypothetical gains and losses of OWA candidate sequences on [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Profit dominance of OWA candidate sequences with [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Longitudinal estimated OWA profits (cumulative), with the “no fee” scenario as the upper bound, and the [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Quarterly distribution of relative gross returns per OWA sequence on Binance. The discrepancy between [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Heatmaps of monthly OWA volume by latency between the first and second operation of the sequence, [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Longitudinal OWA operations on Binance and Kraken grouped by maker/taker configurations. [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Longitudinal OWA volume on Binance and Kraken grouped by maker/taker configurations. [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: Longitudinal daily (black line) and cumulative (red dotted line) gross profits over time on Binance and [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Quarterly distribution of price discrepancy per OWA sequence on Kraken (excluding fees). Contrary to [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Heatmaps of monthly OWA operations by latency between the first and second operation of the sequence. [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Heatmaps of monthly OWA volume by latency between the first and second operation of the sequence. [PITH_FULL_IMAGE:figures/full_fig_p019_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 3 linked inside Pith

  1. [1]

    Rosita P Chang, Sang-Hyop Lee, Sean F Reid, and S Ghon Rhee

  2. [2]

    One-way arbitrage-based interest parity: An application of the Fletcher-Taylor approach in short-date markets.Tinbergen Institute Discussion Papers2002-115 (2002), 2

  3. [3]

    2025.Binance exchange

    CoinMarketCap. 2025.Binance exchange. Retrieved 2025-04-02 from https://coinmarketcap.com/exchanges/binance/

  4. [4]

    2025.Kraken exchange

    CoinMarketCap. 2025.Kraken exchange. Retrieved 2025-04-02 from https://coinmarketcap.com/exchanges/kraken/

  5. [5]

    2025.Top cryptocurrency spot exchanges

    CoinMarketCap. 2025.Top cryptocurrency spot exchanges. Retrieved 2025-08-25 from https://coinmarketcap.com/rankings/exchanges/

  6. [6]

    Tommy Crépellière, Matthias Pelster, and Stefan Zeisberger. 2023. Arbitrage in the market for cryptocurrencies.Journal of Financial Markets64 (2023), 100817

  7. [7]

    Tomasz Czapliński and Elena Nazmutdinova. 2019. Using FIAT curren- cies to arbitrage on cryptocurrency exchanges.Journal of International Studies (2071-8330)12, 1 (2019)

  8. [8]

    Philip Daian, Steven Goldfeder, Tyler Kell, Yunqi Li, Xueyuan Zhao, Iddo Bentov, Lorenz Breidenbach, and Ari Juels. 2020. Flash boys 2.0: Frontrunning in decentralized exchanges, miner extractable value, and consensus instability. InSymposium on Security and Privacy (S&P)

Show all 30 references
  1. [9]

    Alan V Deardorff. 1979. One-way arbitrage and its implications for the foreign exchange markets.Journal of Political Economy87, 2 (1979), 351–364

  2. [10]

    Paz Grimberg, Tobias Lauinger, and Damon McCoy. 2020. Empirical analysis of indirect internal conversions in cryptocurrency exchanges. arXiv preprint arXiv:2002.12274(2020)

  3. [11]

    Denis Gromb and Dimitri Vayanos. 2010. Limits of arbitrage.Annu. Rev. Financ. Econ.2, 1 (2010), 251–275

  4. [12]

    Lioba Heimbach, Vabuk Pahari, and Eric Schertenleib. 2024. Non- atomic arbitrage in decentralized finance. InSymposium on Security and Privacy (S&P)

  5. [13]

    George Kappos, Haaroon Yousaf, Mary Maller, and Sarah Meiklejohn

  6. [14]

    InUSENIX Security Symposium

    An empirical analysis of anonymity in Zcash. InUSENIX Security Symposium

  7. [15]

    Amrit Kumar, Clément Fischer, Shruti Tople, and Prateek Saxena

  8. [16]

    InEuropean Symposium on Research in Computer Security (ESORICS)

    A traceability analysis of Monero’s blockchain. InEuropean Symposium on Research in Computer Security (ESORICS)

  9. [17]

    Igor Makarov and Antoinette Schoar. 2020. Trading and arbitrage in cryptocurrency markets.Journal of Financial Economics135, 2 (2020), 293–319

  10. [18]

    Robert McLaughlin, Christopher Kruegel, and Giovanni Vigna. 2023. A large scale study of the Ethereum arbitrage ecosystem. InUSENIX Security Symposium

  11. [19]

    Sarah Meiklejohn, Marjori Pomarole, Grant Jordan, Kirill Levchenko, Damon McCoy, Geoffrey M Voelker, and Stefan Savage. 2013. A fistful of Bitcoins: Characterizing payments among men with no names. In Internet Measurement Conference (IMC)

  12. [20]

    Andrew Morin and Tyler Moore. 2023. How cryptocurrency exchange interruptions create arbitrage opportunities. InEuropean Symposium on Security and Privacy Workshops (EuroS&PW)

  13. [21]

    Malte Möser, Kyle Soska, Ethan Heilman, Kevin Lee, Henry Heffan, Shashvat Srivastava, Kyle Hogan, Jason Hennessey, Andrew Miller, Arvind Narayanan, et al. 2017. An empirical analysis of traceability in the Monero blockchain.arXiv preprint arXiv:1704.04299(2017)

  14. [22]

    Matthias Muck, Thomas Schmidl, and Julian Wolf. 2025. Wish or reality? On the exploitability of triangular arbitrage in cryptocurrency markets.Finance Research Letters73 (2025), 106508

  15. [23]

    Zheng Nan, Taisei Kaizoji, and T Kaizoji. 2019. Bitcoin-based triangular arbitrage with the Euro/US dollar as a foreign futures hedge: Modeling with a bivariate GARCH model.Quantitative Finance and Economics 3, 2 (2019), 347–365

  16. [24]

    Gina Pieters and Sofia Vivanco. 2017. Financial regulations and price inconsistencies across Bitcoin markets.Information Economics and Policy39 (2017), 1–14

  17. [25]

    Kaihua Qin, Liyi Zhou, and Arthur Gervais. 2022. Quantifying blockchain extractable value: How dark is the forest?. InSymposium on Security and Privacy (S&P)

  18. [26]

    Jeffrey Quesnelle. 2017. On the linkability of Zcash transactions.arXiv preprint arXiv:1712.01210(2017)

  19. [27]

    2025.CEX vs DEX: Report from 2024 and forecasts for

    OAK Research. 2025.CEX vs DEX: Report from 2024 and forecasts for

  20. [28]

    Retrieved 2025-08-25 from https://oakresearch.io/en/reports/ sectors/cex-vs-dex-report-2024-forecasts-2025

  21. [29]

    Ye Wang, Yan Chen, Haotian Wu, Liyi Zhou, Shuiguang Deng, and Roger Wattenhofer. 2022. Cyclic arbitrage in decentralized exchanges. 14 A Truckload of Satoshis: Detecting and Measuring One-Way Arbitrage in the Wild InWeb Conference (WWW)

  22. [30]

    price imbalance

    Haaroon Yousaf, George Kappos, and Sarah Meiklejohn. 2019. Trac- ing transactions across cryptocurrency ledgers. InUSENIX Security Symposium. 15 Eugenio Nerio Nemmi, Tobias Lauinger, Paz Grimberg, Massimo La Morgia, Damon McCoy, and Alessandro Mei A Appendix A.1 Ethics This pa...

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.