{"paper_id":"2411.19637","title":"Ergodic optimal liquidations in DeFi","authors":["Jialun Cao","David Šiška"],"paper_date":"2024-11-29","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2411.19637v1","page_url":"https://thequant.space/flowcharts/ergodic-optimal-liquidations-in-defi/","scores":{"empirical_rigor":4,"math_complexity":8.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper formulates the liquidation problem in DeFi as an ergodic optimal control problem and derives closed‑form solutions for the optimal liquidation strategy and the long‑run average reward.","quote":"We address the liquidation problem arising from the credit risk management in decentralised finance (DeFi) by formulating it as an ergodic optimal control problem. By simplifying the model (linear temporary and permanent price impacts, simplified cash balance dynamics), we derive the closed-form solutions for the optimal liquidation strategies, which balance immediate executions with the temporary and permanent","summary":"We address the liquidation problem arising from the credit risk management in decentralised finance (DeFi) by formulating it as an ergodic optimal control problem. In decentralised derivatives exchanges, liquidation is triggered whenever the parties fail to maintain sufficient collateral for their…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Temporary price impact is linear: f(ν)=kν with k>0.","type":"market","stated":"explicit","quote":"(7) where S0 ∈R+ is the initial mid price. ii) We assume that liquidation intensities and sizes are symmetric for long and short positions i.e. that λ+ = λ−=: λ and η+ = η−=: η. iii) We will assume linear price impact functions, specifically that temporary price im-pact function is f(ν) = kν some constant k ∈R+ and that the permanent price impact is g(ν) = bν for some constant b ∈R+.","support":"full"},{"assumption":"Permanent price impact is linear: g(ν)=bν with b>0.","type":"market","stated":"explicit","quote":"(7) where S0 ∈R+ is the initial mid price. ii) We assume that liquidation intensities and sizes are symmetric for long and short positions i.e. that λ+ = λ−=: λ and η+ = η−=: η. iii) We will assume linear price impact functions, specifically that temporary price im-pact function is f(ν) = kν some constant k ∈R+ and that the permanent price impact is g(ν) = bν for some constant b ∈R+.","support":"full"},{"assumption":"Liquidation intensities and sizes are symmetric for long and short positions (λ+ = λ− = λ, η+ = η− = η).","type":"model","stated":"explicit","quote":"(7) where S0 ∈R+ is the initial mid price. ii) We assume that liquidation intensities and sizes are symmetric for long and short positions i.e. that λ+ = λ−=: λ and η+ = η−=: η. iii) We will assume linear price impact functions, specifically that temporary price im-pact function is f(ν) = kν some constant k ∈R+ and that the permanent price impact is g(ν) = bν for some constant b ∈R+.","support":"full"},{"assumption":"Parties do not update the initial margin balance, allowing a simplified cash‑balance dynamics.","type":"model","stated":"explicit","quote":"In order to obtain a closed-form solution, we make a number of simplifications which we will now discuss. i) We will assume that the parties don’t update the initial margin balance and thus replace (4) by dXν t = ˆSν t νtdt + rζ+S0dN+ t + rζ−S0dN− t , Xν 0 = x,","support":"full"},{"assumption":"Risk aversion φ>0 is the only free parameter needed for the optimal strategy.","type":"other","stated":"explicit","quote":"Note that while the optimal average earnings per unit time depend on all the model parameters, the optimal strategy only needs the risk aversion ϕ > 0 and temporary price impact k ∈R+.","support":"full"},{"assumption":"The mid‑price follows a Brownian motion with volatility σ (Gaussian assumption).","type":"statistical","stated":"explicit","quote":"The profit or loss from holding a position is determined by (Sν t )t≥0 which is the mid-price (and mark-to-market) process driven by a standard Brownian motion (Wt)t≥0.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Liquidation event timestamps and position sizes","provider":null,"kind":"crypto-onchain","access":"unknown","period":null,"universe":null,"quote":"We collect the times at which liq-uidations of distressed traders occur over [0, T], denoted by (τi)N i=1 ∈[0, T], where N represents the total number of liquidation events observed. Additionally, we record the absolute value of the size of the distressed positions at each liquidation, represented by (ζi)N i=1.","support":"full"},{"name":"Mid‑price and net order‑flow series for calibration of b","provider":null,"kind":"market-prices","access":"unknown","period":null,"universe":null,"quote":"ERGODIC OPTIMAL LIQUIDATIONS IN DEFI 10 To estimate the permanent price impact parameter b we again divide the time in-terval [0, T] into multiple subintervals P i[ti, ti+1]. Let µi be the net order-flow in the subinterval [ti, ti+1] and ∆Si = Sti+1 −Sti be the change in the midprice.","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":"Temporary impact f(ν)=kν and permanent impact g(ν)=bν are incorporated into the control problem.","quote":"(7) where S0 ∈R+ is the initial mid price. ii) We assume that liquidation intensities and sizes are symmetric for long and short positions i.e. that λ+ = λ−=: λ and η+ = η−=: η. iii) We will assume linear price impact functions, specifically that temporary price im-pact function is f(ν) = kν some constant k ∈R+ and that the permanent price impact is g(ν) = bν for some constant b ∈R+.","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Temporary impact f(ν)=kν and permanent impact g(ν)=bν are incorporated into the control problem.","card_quote":"(7) where S0 ∈R+ is the initial mid price. ii) We assume that liquidation intensities and sizes are symmetric for long and short positions i.e. that λ+ = λ−=: λ and η+ = η−=: η. iii) We will assume linear price impact functions, specifically that temporary price im-pact function is f(ν) = kν some constant k ∈R+ and that the permanent price impact is g(ν) = bν for some constant b ∈R+.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"The ergodic constant γ has the closed‑form γ = 2rληS₀ − λη²b − 2λη²p k φ and is independent of initial conditions.","kind":"theoretical","quote":"Based on the simplifications and assumptions, we derive the closed-form expression for γ, see Theorem 2.1, as γ = 2rληS0 −λη2b −2λη2p kϕ , (8) which demonstrates that the ergodic constant γ is independent of any initial conditions and is determined solely by the market parameters.","support":"full"},{"finding":"The optimal liquidation rate is ν* (q) = r φ k q, i.e., a constant fraction of the current inventory.","kind":"theoretical","quote":"The ergodic optimal Markov control is given by ν∗(q) = r ϕ k q , (9) i.e. that the optimal strategy is indeed a fraction of the current position and the coef-ficient is determined by the parameters, see Corollary 2.1.","support":"full"},{"finding":"The optimal strategy only requires the risk‑aversion parameter φ and the temporary impact coefficient k, making it robust to misspecification of other parameters.","kind":"theoretical","quote":"This means that ERGODIC OPTIMAL LIQUIDATIONS IN DEFI 5 the optimal strategy is robust to parameter misspecification; only the temporary price impact has to be estimated correctly.","support":"full"},{"finding":"Numerical simulations confirm that the proposed optimal strategy outperforms naive strategies and that the simplified model closely approximates the original market environment.","kind":"performance","quote":"Numerical simulations further highlight the effectiveness of the proposed optimal strategy and demonstrate that the simplified model closely approximates the original market environment.","support":"full"},{"finding":"All model parameters can be calibrated from observable data, with k estimated as the mean across sub‑intervals and b obtained via linear regression of price changes on net order‑flow.","kind":"other","quote":"The overall parameter k is taken as the mean value of k(i) across all subintervals, i.e. k = 1 M PM i=1 k(i). As we assume the permanent price impact is linear in the model, hence we can perform the linear regression ∆Si = bµi + εi , where the error terms εi are assumed to be iid with a normal distribution.","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":24,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"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":"2026-11-14T19:54:01Z"},"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/"}