{"paper_id":"2406.12417","title":"Fees in AMMs: A quantitative study","authors":["Abe Alexander","Lars Fritz"],"paper_date":"2024-06-18","paper_type":"methodological","paper_url":"https://arxiv.org/pdf/2406.12417v2","page_url":"https://thequant.space/flowcharts/fees-in-amms-a-quantitative-study/","scores":{"empirical_rigor":6,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper models arbitrage dynamics for a concrete AMM pool and studies how fee choice affects AMM revenue, proposing dynamical asymmetric fees to mitigate losses from toxic flow.","quote":"In this paper we will investigate the mechanics of said arbitrage and try to understand how AMMs can maximize the revenue creation or in other words minimize the losses. To that end, we model the dynamics of arbitrage activity for a concrete implementation of a pool and study its sensitivity to the choice of fee aiming to maximize the revenue for the AMM.","summary":"In the ever evolving landscape of decentralized finance automated market makers (AMMs) play a key role: they provide a market place for trading assets in a decentralized manner. For so-called bluechip pairs, arbitrage activity provides a major part of the revenue generation of AMMs but also a major…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Price discovery occurs on the CEX and the AMM price is slaved to it.","type":"market","stated":"explicit","quote":"We assume that price discovery happens on the CEX and the price dynamics of the AMM is ’slaved’ to it.","support":"full"},{"assumption":"Arbitrage can always be performed in an optimal fashion.","type":"execution","stated":"explicit","quote":"Furthermore, we assume that arbitrage can always be performed in an optimal fashion.","support":"full"},{"assumption":"The AMM is modeled as a constant function market maker (CFMM).","type":"model","stated":"explicit","quote":"3.1 The AMM For simplicity, we model the AMM as a constant function market maker.","support":"full"},{"assumption":"The CEX is assumed to have infinite liquidity and zero price impact.","type":"market","stated":"explicit","quote":"It is important here to reiterate that we assume that the CEX is of infinite liquidity and there is no price impact. While in practice the CEX is not of infinite liquidity, it is reasonable to assume that it is of significantly higher liquidity justifying the subsequent approach.","support":"full"},{"assumption":"Transaction costs other than the AMM fee and flash‑loan fee are negligible (TXN = 0).","type":"execution","stated":"explicit","quote":"We are assuming for now that transaction costs are negligible (TXN = 0), meaning we have a condition α 1 −f 1 + piA∆FL(1 −f) −(1 + ffl) > 0 .","support":"full"},{"assumption":"Random walk steps are symmetric with equal up/down probability unless otherwise stated.","type":"statistical","stated":"explicit","quote":"Symmetric random walks, i.e. pup = pdown = 1/2, exhibit interesting prop-erties some of which we review here: The average price after n steps is ⟨p⟩= p0 and the variance is ⟨(pn −p0)2⟩= ⟨p2 n⟩−p2 0 = p2 0σ2n. We distinguish two case, one in which the probability of going up and down is balanced, called ’symmetric walk’, and another in which there is a preferential direction, referred to as ’asymmetric walk’.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Synthetic CEX price simulations via Brownian motion","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"To this end, we make the following experiment: we fix the relative volatility of the CEX price per time step at 0.001. Consequently, we evolve the CEX price for 1000 time steps and arbitrage the AMM.","support":"full"},{"name":"Random walk simulations (σ = 0.02, 10,000 runs)","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Symmetric walk: pup = pdown In this case, we made a simulation in the following way: in all cases we use σ = 0.02 (which is a rather big value) and run 10000 simulations.","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Swap fee f charged by the AMM and flash‑loan fee ffl are included; transaction cost TXN is set to zero in the main analysis.","quote":"For that, ARB has to pay the feeffl meaning in total it costs ∆FLffl. Additionally, the AMM charges a fee f for swaps, meaning the actual ∆B entering the pool will be ∆B = ∆FL(1 −f) .","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Swap fee f charged by the AMM and flash‑loan fee ffl are included; transaction cost TXN is set to zero in the main analysis.","card_quote":"For that, ARB has to pay the feeffl meaning in total it costs ∆FLffl. Additionally, the AMM charges a fee f for swaps, meaning the actual ∆B entering the pool will be ∆B = ∆FL(1 −f) .","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"For sideways (no‑drift) price motion the AMM revenue from fees does not depend on the fee level in the long‑time limit.","kind":"theoretical","quote":"We find that for sideways motion without preferential direction the revenue generation does not depend on the fee setting in the long time limit.","support":"full"},{"finding":"When price drift is present, revenue depends strongly on fee directionality, suggesting asymmetric fees can capture more arbitrage gains.","kind":"theoretical","quote":"However, in the case with drift there is a strong dependence that is directional.","support":"full"},{"finding":"Fees roughly three times the per‑block volatility yield the highest fee return in simulations of symmetric price walks.","kind":"performance","quote":"The main observation is that fees compared small to the per block volatility lead to suboptimal fee generation whereas fees on the order three times the per block volatility seem to make a good return. though, meaning the fees per arbitrage attempt will be moderate. The main observation is that fees compared small to the per block volatility lead to suboptimal fee generation whereas fees on the order three times the","support":"full"},{"finding":"For larger static fees the probability of a successful arbitrage event scales approximately as 1/f.","kind":"statistical","quote":"The main finding is that for larger fees, the probability of having an arbitrage event follows ∝1/f.","support":"full"},{"finding":"In a drift scenario the revenue for the majority token grows linearly with the fee level, while the minority token’s revenue saturates or decays.","kind":"theoretical","quote":"Since the revenue per arbitrage still scales linear in f, overall we find a linear in f behavior, explaining the numerical findings/ Summary of this subsection: In a situation with constant volatility and preferential movement/drift, the revenue depends linearly on the fees for the majority token. We find that the probability of having an arbitrage event becomes constant while the revenue per trade grows linearly","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":20,"citations":{"cited_by_in_corpus":2,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":1,"unclassified":1},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"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":"2026-11-14T19:56:51Z"},"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/"}