{"paper_id":"2506.02869","title":"Optimal Dynamic Fees in Automated Market Makers","authors":["Leonardo Baggiani","Martin Herdegen","Leandro Sánchez-Betancourt"],"paper_date":"2025-06-03","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2506.02869v2","page_url":"https://thequant.space/flowcharts/optimal-dynamic-fees-in-automated-market-makers/","scores":{"empirical_rigor":6.5,"math_complexity":8.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper determines the optimal dynamic fees in a constant function market maker and derives approximate closed‑form solutions for the control problem.","quote":"In this work, we determine the optimal dynamic fees in a constant function market maker.","summary":"Automated Market Makers (AMMs) are emerging as a popular decentralised trading platform. In this work, we determine the optimal dynamic fees in a constant function market maker. We find approximate closed-form solutions to the control problem and study the optimal fee structure. We find that there…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The quantity of asset Y in the pool takes values in a finite grid and the corresponding quantity of asset X is given by a decreasing convex level function.","type":"model","stated":"explicit","quote":"We assume that quantity of asset Y in the pool takes values in a finite grid given by {y−N := y, . . . , y0, . . . , yN := y}, where y and y can be interpreted as reserve constraints satisfying 0 < y < y0 < y < p2.","support":"full"},{"assumption":"The traded amounts Δ⁺(y) and Δ⁻(y) are constant (δ⁺, δ⁻) to enable a quadratic expansion of the exponential terms.","type":"model","stated":"explicit","quote":"(5.1) We assume that ∆−(y) and ∆+(y) are constant equal to δ−and δ+, respectively. Remark 5.1.","support":"full"},{"assumption":"For the first approximation the external oracle price is constant (St = S0, i.e., σ = 0).","type":"market","stated":"explicit","quote":"4 Constant oracle price To compute the first approximate solution, we take St = S0, which is equivalent to assuming that σ = 0.","support":"full"},{"assumption":"The exponential terms in the HJB equation are approximated by a linear‑quadratic expansion.","type":"statistical","stated":"explicit","quote":"Second, we approximate the exponential terms in the Hamilton–Jacobi–Bellman (HJB) equation using a linear-quadratic expansion.","support":"full"},{"assumption":"The trading function f(x,y) is strictly increasing, twice differentiable, and convex in its level function φ(y) to exclude round‑trip arbitrage.","type":"model","stated":"explicit","quote":"Let p2 > 0 denote the depth of the pool and let f : R+ × R+ →R+ be the trading function, which is strictly increasing and twice differentiable in both arguments. The latter is then automatically differentiable and decreasing, and we assume in addition that it is convex in order to exclude roundtrip arbitrage, see Cartea et al. (2024b).","support":"full"}]},"datasets":{"status":"none_found","items":[],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"none"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R4a","candidates":{"front_matter":"n/a","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Dynamic fees are modeled as functions linear in inventory and sensitive to external price changes; they serve as the revenue component in the control problem.","card_quote":"Our results also show that dynamic fees that are linear in inventory and are sensitive to changes in the external price are a good approximation of the optimal fee structure and thus constitute suitable candidates when designing fees for AMMs. Second, we approximate the exponential terms in the Hamilton–Jacobi–Bellman (HJB) equation using a linear-quadratic expansion.","meaning":"the paper reports no trading or execution result; the cost the extractor found is a model parameter"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"Two distinct fee regimes are optimal: higher fees to deter arbitrageurs and lower fees to increase volatility and attract noise traders.","kind":"theoretical","quote":"We find that there are two distinct fee regimes: one in which the AMM imposes higher fees to deter arbitrageurs, and another where fees are lowered to increase volatility and attract noise traders. In both approximations, we obtain the same insight: the venue finds it optimal to adopts two distinct fee regimes; one in which the AMM imposes higher fees to deter arbitrageurs, and another where fees are lowered to","support":"full"},{"finding":"Dynamic fees that are linear in inventory and track external price changes closely approximate the optimal fee structure.","kind":"theoretical","quote":"Our results also show that dynamic fees that are linear in inventory and are sensitive to changes in the external price are a good approximation of the optimal fee structure and thus constitute suitable candidates when designing fees for AMMs. Our results also show that dynamic fees that are linear in inventory and that track changes in the external price are a good approximation of the optimal fee structure and","support":"full"},{"finding":"In simulations with a constant external price, the optimal dynamic fees outperform constant fee strategies by a significant margin.","kind":"performance","quote":"In the case of a constant external price, we find that the optimal fees outperform constant fee strategies by a significant margin.","support":"full"},{"finding":"A linear approximation of the optimal policies yields revenue outcomes nearly identical to the fully optimal fees.","kind":"performance","quote":"Furthermore, a linear approximation of the optimal policies proves to be sufficiently accurate, yielding nearly identical revenue outcomes.","support":"full"},{"finding":"Both the first and second approximation methods produce similar optimal fee policies and both outperform constant fees.","kind":"performance","quote":"53.02 53.45 53.45 1.01 Constant 35.18 35.18 35.19 0.68 52.29 52.28 52.29 0.99 k = 1 Optimal SA 71.49 35.92 35.93 0.69 106.42 53.46 53.47 1.00 Optimal FA 71.51 35.47 35.44 0.68 106.38 53.46 53.51 1.01 Constant 71.23 35.60 35.61 0.69 105.90 52.91 52.91 1.00 As before, we find that either approximation yields a performance that is almost indistinguishable from the other, and both outperform constant fees substantially.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":16,"citations":{"cited_by_in_corpus":7,"by_independent_authors":4,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":4,"unclassified":3},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":true,"code_url":"https://github.com/leonardobaggiani/amm-fees","note":null},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":2,"extracted_at":"2026-11-14T19:49:59Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}