arXiv subjects

Andrey Urusov

Publications and source records attributed to Andrey Urusov.

4 recordsLinked to original sources

arXiv2026

Liquidity provision in CLMMs: evidence from transactions data

The emergence of Concentrated Liquidity Market Makers (CLMMs) has made liquidity provision on decentralized exchanges an active and risk-sensitive task. However, the standalone profitability of liquidity provision remains unclear for liquidity providers (LPs) who neither hedge their inventory risk nor receive off-pool profits. This paper studies the actual outcomes of LP activity using historical transaction-level data from WETH/USD liquidity pools on the Base chain across the Uniswap, Aerodrome, PancakeSwap and SushiSwap protocols. We propose a methodology for reconstructing LP PnL dynamics from on-chain events and introduce an original metric that captures both the terminal state of LP capital and its path over time. Based on this framework, we estimate the share of successful LPs, classify their behavior and develop a taxonomy of 15 position types as structural components of PnL. We further identify a distinct class of multi-LPs operating across several pools and show that the dominant profitable position configurations are concentrated around the current pool price. The results show that only about one out of six LPs avoids losses in the selected market segment, raising an open question about the true economic motives of LP participation. Evidence also suggests that successful LPs often close positions before the full range is traversed, making observed behavior closer to profit-target-based strategies.

q-fin.TR↗

arXiv2026

Dynamic Liquidity Provision in Decentralized Markets: Strategy Optimization and Performance Evaluation in Concentrated Liquidity AMMs

Concentrated Liquidity Market Makers (CLMMs) represent a fundamental innovation in market microstructure, transforming liquidity provision from passive portfolio allocation to active risk management. This evolution creates significant challenges for performance evaluation and strategy optimization, particularly due to the absence of comprehensive historical liquidity data. We address these challenges through a novel methodological framework that reconstructs historical liquidity states from swap transaction data, enabling rigorous backtesting of dynamic liquidity provision strategies. Our parametric reconstruction method achieves high accuracy (approximation errors averaging around 2\%) without relying on historical liquidity snapshots, addressing a critical data gap in decentralized finance research. We apply this framework to evaluate tau-reset strategies--dynamic liquidity reallocation approaches that respond to market movements--across multiple Uniswap v3 pools. Using machine learning to optimize strategy parameters based on market conditions, we identify consistent outperformance (13--23\% higher fees) compared to uniform allocation benchmarks. Our analysis reveals important insights into the risk-return tradeoffs in automated market making, including the critical role of impermanent loss as a dominant risk factor and the effectiveness of asymmetric strategy modifications for capital preservation. These findings contribute to the broader understanding of market microstructure in decentralized exchanges, providing both methodological innovations for performance evaluation and practical insights for liquidity providers navigating this evolving financial landscape.

q-fin.MF↗

arXiv2024

Backtesting Framework for Concentrated Liquidity Market Makers on Uniswap V3 Decentralized Exchange

Decentralized finance (DeFi) has revolutionized the financial landscape, with protocols like Uniswap offering innovative automated market-making mechanisms. This article explores the development of a backtesting framework specifically tailored for concentrated liquidity market makers (CLMM). The focus is on leveraging the liquidity distribution approximated using a parametric model, to estimate the rewards within liquidity pools. The article details the design, implementation, and insights derived from this novel approach to backtesting within the context of Uniswap V3. The developed backtester was successfully utilized to assess reward levels across several pools using historical data from 2023 (pools Uniswap v3 for pairs of altcoins, stablecoins and USDC/ETH with different fee levels). Moreover, the error in modeling the level of rewards for the period under review for each pool was less than 1\%. This demonstrated the effectiveness of the backtester in quantifying liquidity pool rewards and its potential in estimating LP's revenues as part of the pool rewards, as focus of our next research. The backtester serves as a tool to simulate trading strategies and liquidity provision scenarios, providing a quantitative assessment of potential returns for liquidity providers (LP). By incorporating statistical tools to mirror CLMM pool liquidity dynamics, this framework can be further leveraged for strategy enhancement and risk evaluation for LPs operating within decentralized exchanges.

q-fin.MF↗

arXiv2023

A Spiking Binary Neuron -- Detector of Causal Links

Causal relationship recognition is a fundamental operation in neural networks aimed at learning behavior, action planning, and inferring external world dynamics. This operation is particularly crucial for reinforcement learning (RL). In the context of spiking neural networks (SNNs), events are represented as spikes emitted by network neurons or input nodes. Detecting causal relationships within these events is essential for effective RL implementation. This research paper presents a novel approach to realize causal relationship recognition using a simple spiking binary neuron. The proposed method leverages specially designed synaptic plasticity rules, which are both straightforward and efficient. Notably, our approach accounts for the temporal aspects of detected causal links and accommodates the representation of spiking signals as single spikes or tight spike sequences (bursts), as observed in biological brains. Furthermore, this study places a strong emphasis on the hardware-friendliness of the proposed models, ensuring their efficient implementation on modern and future neuroprocessors. Being compared with precise machine learning techniques, such as decision tree algorithms and convolutional neural networks, our neuron demonstrates satisfactory accuracy despite its simplicity. In conclusion, we introduce a multi-neuron structure capable of operating in more complex environments with enhanced accuracy, making it a promising candidate for the advancement of RL applications in SNNs.

cs.NE↗