Yurii Kvasiuk, Tianyi Li, Owen Colegrove, Moritz Münchmeyer
We explore the application of LLM-driven algorithm optimization to several common tasks in quantitative finance. MadEvolve, a general-purpose algorithm optimization framework inspired by DeepMind's Alpha-Evolve, was recently developed to optimize algorithms in computational cosmology. Here we demonstrate the utility of MadEvolve to optimize algorithmic trading strategies and alpha generation at the example of Bitcoin trading. On our simulation and backtesting setup, we achieve significant improvements on all tasks we considered, such as evolving feature sets for signal generation, optimizing separate components of the trading strategy, and jointly evolving the feature pipeline together with the execution strategy. Additionally, we compare our method to other agentic search approaches, specifically Claude Code, and carefully evaluate p-hacking probabilities on our simulation setup. Our findings strongly support the utility of AI-driven agentic and evolutionary algorithms for algorithmic trading and quantitative finance.
LLM agents are promising tools for empirical discovery, but their flexibility can also turn discovery into uncontrolled search. We study how to use agents under a reproducible protocol through cryptocurrency factor discovery. Our framework casts the task as sequential hypothesis search: an agent reads an append-only experiment trace, proposes falsifiable factor hypotheses, and maps them to executable recipes, while a deterministic engine enforces fixed data splits, selection gates, transaction costs, and portfolio tests. Candidate actions are restricted to a point-in-time factor DSL, making both successful and failed hypotheses auditable. A ridge-combined portfolio trained only on 2020--2022 data achieves a 44.55% annualized return and Sharpe ratio of 1.55 in the 2024--2026 pure out-of-sample period after a 5 basis point one-way trading cost.
Dynamic-weight AMMs (aka Temporal Function Market Makers, TFMMs) implement algorithmic asset allocation, analogous to index or smart beta funds, by continuously updating pools' weights. A strategy updates target weights over time, and arbitrageurs trade the pool back toward those weights. This creates a sequence of small, predictable mispricings that grow until taken, effectively executing rebalances as a series of Dutch reverse auctions. Prior theoretical and simulation work (Willetts & Harrington, 2024) predicted that this mechanism could outperform CEX-style rebalancing. We test that claim on two live pools on the QuantAMM protocol, one on Ethereum mainnet and one on Base, across two short rebalancing windows six months apart (July 2025 and January 2026). We perform block-level arbitrage analysis, and then measure long term outcomes using Loss-vs-Rebalancing (LVR) and Rebalancing-vs-Rebalancing (RVR) benchmarks. On mainnet, rebalancing becomes markedly more efficient over time (more frequent arbitrage trades with lower value extracted per trade), reaching performance comparable to or better than CEX-based models. On Base, rebalancing persists even when per-trade extraction is near (or below) zero, consistent with routing-driven execution, and achieves efficiencies that meet or exceed standard "perfect rebalancing" LVR baselines. These results demonstrate dynamic-weight AMMs as a competitive execution layer for tokenised funds, with superior performance on L2s where routing and lower data costs compress arbitrage spreads.
Three traits of decentralized finance are studied. First, the market impact function is derived for optimal-growth liquidity providers. For a standard random walk, the classic square-root impact is recovered. An extension is then derived to fit general fractional Ornstein-Uhlenbeck processes. These findings break with the linearized liquidity models used in most decentralized exchanges. Second, a Constant Product Market Maker is viewed as a multi-phase Carnot engine, where one phase matches the exchange of tokens by a liquidity taker, and another the change of pool size by a liquidity provider. Third, stablecoin de-pegging is a form of catastrophe risk. By using growth optimization, default odds are linked to the cost of catastrophe bonds. De-pegging insurance can act as a counterweight and a key marketing tool when the law forbids the payment of interest on stablecoins.
This study introduces a novel approach to walk-forward optimization by parameterizing the lengths of training and testing windows. We demonstrate that the performance of a trading strategy using the Exponential Moving Average (EMA) evaluated within a walk-forward procedure based on the Robust Sharpe Ratio is highly dependent on the chosen window size. We investigated the strategy on intraday Bitcoin data at six frequencies (1 minute to 60 minutes) using 81 combinations of walk-forward window lengths (1 day to 28 days) over a 19-month training period. The two best-performing parameter sets from the training data were applied to a 21-month out-of-sample testing period to ensure data independence. The strategy was only executed once during the testing period. To further validate the framework, strategy parameters estimated on Bitcoin were applied to Binance Coin and Ethereum. Our results suggest the robustness of our custom approach. In the training period for Bitcoin, all combinations of walk-forward windows outperformed a Buy-and-Hold strategy. During the testing period, the strategy performed similarly to Buy-and-Hold but with lower drawdown and a higher Information Ratio. Similar results were observed for Binance Coin and Ethereum. The real strength was demonstrated when a portfolio combining Buy-and-Hold with our strategies outperformed all individual strategies and Buy-and-Hold alone, achieving the highest overall performance and a 50 percent reduction in drawdown. A conservative fee of 0.1 percent per transaction was included in all calculations. A cost sensitivity analysis was performed as a sanity check, revealing that the strategy's break-even point was around 0.4 percent per transaction. This research highlights the importance of optimizing walk-forward window lengths and emphasizing the value of single-time out-of-sample testing for reliable strategy evaluation.
Kundan Mukhia, Buddha Nath Sharma, Salam Rabindrajit Luwang, Md. Nurujjaman · 7 authors
We study how the 2024 U.S. presidential election, viewed as a major political risk event, affected cryptocurrency markets by distinguishing human-driven peer-to-peer stablecoin transactions from automated algorithmic activity. Using structural break analysis, we find that human-driven Ethereum Request for Comment 20 (ERC-20) transactions shifted on November 3, two days before the election, while exchange trading volumes reacted only on Election Day. Automated smart-contract activity adjusted much later, with structural breaks appearing in January 2025. We validate these shifts using surrogate-based robustness tests. Complementary energy-spectrum analysis of Bitcoin and Ethereum identifies pronounced post-election turbulence, and a structural vector autoregression confirms a regime shift in stablecoin dynamics. Overall, human-driven stablecoin flows act as early-warning indicators of political stress, preceding both exchange behavior and algorithmic responses.
This paper examines factors that influence prices of most common five cryptocurrencies such as Bitcoin, Ethereum, Dash, Litecoin, and Monero over 2010-2018 using weekly data. The study employs ARDL technique and documents several findings. First, cryptomarket-related factors such as market beta, trading volume, and volatility appear to be significant determinant for all five cryptocurrencies both in short- and long-run. Second, attractiveness of cryptocurrencies also matters in terms of their price determination, but only in long-run. This indicates that formation (recognition) of the attractiveness of cryptocurrencies are subjected to time factor. In other words, it travels slowly within the market. Third, SP500 index seems to have weak positive long-run impact on Bitcoin, Ethereum, and Litcoin, while its sign turns to negative losing significance in short-run, except Bitcoin that generates an estimate of -0.20 at 10% significance level. Lastly, error-correction models for Bitcoin, Etherem, Dash, Litcoin, and Monero show that cointegrated series cannot drift too far apart, and converge to a long-run equilibrium at a speed of 23.68%, 12.76%, 10.20%, 22.91%, and 14.27% respectively.
This paper introduces a hybrid framework for portfolio optimization that fuses Long Short-Term Memory (LSTM) forecasting with a Proximal Policy Optimization (PPO) reinforcement learning strategy. The proposed system leverages the predictive power of deep recurrent networks to capture temporal dependencies, while the PPO agent adaptively refines portfolio allocations in continuous action spaces, allowing the system to anticipate trends while adjusting dynamically to market shifts. Using multi-asset datasets covering U.S. and Indonesian equities, U.S. Treasuries, and major cryptocurrencies from January 2018 to December 2024, the model is evaluated against several baselines, including equal-weight, index-style, and single-model variants (LSTM-only and PPO-only). The framework's performance is benchmarked against equal-weighted, index-based, and single-model approaches (LSTM-only and PPO-only) using annualized return, volatility, Sharpe ratio, and maximum drawdown metrics, each adjusted for transaction costs. The results indicate that the hybrid architecture delivers higher returns and stronger resilience under non-stationary market regimes, suggesting its promise as a robust, AI-driven framework for dynamic portfolio optimization.
Given a universe of N assets, investors often form equally weighted portfolios (EWPs) by selecting subsets of assets. EWPs are simple, robust, and competitive out-of-sample, yet the uncertainty about which subset truly performs best is largely ignored. Traditional approaches typically rely on a single selected portfolio, but this fails to consider alternative investment strategies that may perform just as well when accounting for statistical uncertainty. To address this selection uncertainty, we introduce the Selection Confidence Set (SCS) for EWPs: the set of all portfolios that, under a given loss function and at a specified confidence level, contains the unknown set of optimal portfolios under repeated sampling. The SCS quantifies selection uncertainty by identifying a range of plausible portfolios, challenging the idea of a uniquely optimal choice. Like a confidence set, its size reflects uncertainty -- growing with noisy or limited data, and shrinking as the sample size increases. Theoretically, we establish that the SCS covers the unknown optimal selection with high probability and characterize how its size grows with underlying uncertainty, corroborating these results through Monte Carlo experiments. Applications to the French 17-Industry Portfolios and Layer-1 cryptocurrencies underscore the importance of accounting for selection uncertainty when comparing equally weighted strategies.
This paper investigates an important problem of an appropriate variance-covariance matrix estimation in the Modern Portfolio Theory. We propose a novel framework for variancecovariance matrix estimation for purposes of the portfolio optimization, which is based on deep learning models. We employ the long short-term memory (LSTM) recurrent neural networks (RNN) along with two probabilistic deep learning models: DeepVAR and GPVAR to the task of one-day ahead multivariate forecasting. We then use these forecasts to optimize portfolios of stocks and cryptocurrencies. Our analysis presents results across different combinations of observation windows and rebalancing periods to compare performances of classical and deep learning variance-covariance estimation methods. The conclusions of the study are that although the strategies (portfolios) performance differed significantly between different combinations of parameters, generally the best results in terms of the information ratio and annualized returns are obtained using the LSTM-RNN models. Moreover, longer observation windows translate into better performance of the deep learning models indicating that these methods require longer windows to be able to efficiently capture the long-term dependencies of the variance-covariance matrix structure. Strategies with less frequent rebalancing typically perform better than these with the shortest rebalancing windows across all considered methods.
Bitcoin treasury companies have taken stock markets by storm amassing billions of dollars worth of tokens in hundreds of entities. The paper discusses, how leverage - whether created through corporate debt or investors using stock as loan collateral - fuels this trend. The extension of the binary-choice Kelly criterion to incorporate uncertainty in the form of the Kullback-Leibler divergence or more generally Bregman divergence is also briefly discussed.
Antonino Castelli, Paolo Giudici, Alessandro Piergallini
The rapid growth of crypto markets has opened new opportunities for investors, but at the same time exposed them to high volatility. To address the challenge of managing dynamic portfolios in such an environment, this paper presents a practical application of a multi-agent system designed to autonomously construct and evaluate crypto-asset allocations. Using data on daily frequencies of the ten most capitalized cryptocurrencies from 2020 to 2025, we compare two automated investment strategies. These are a static equal weighting strategy and a rolling-window optimization strategy, both implemented to maximize the evaluation metrics of the Modern Portfolio Theory (MPT), such as Expected Return, Sharpe and Sortino ratios, while minimizing volatility. Each step of the process is handled by dedicated agents, integrated through a collaborative architecture in Crew AI. The results show that the dynamic optimization strategy achieves significantly better performance in terms of risk-adjusted returns, both in-sample and out-of-sample. This highlights the benefits of adaptive techniques in portfolio management, particularly in volatile markets such as cryptocurrency markets. The following methodology proposed also demonstrates how multi-agent systems can provide scalable, auditable, and flexible solutions in financial automation.
Jun 26, 2025·Proceedings of Workshops at the 50th International Conference on Very Large Data Bases, {VLDB} 2024, Guangzhou, China, August 26-30, 2024
This study investigates the impact of data source diversity on the performance of cryptocurrency forecasting models by integrating various data categories, including technical indicators, on-chain metrics, sentiment and interest metrics, traditional market indices, and macroeconomic indicators. We introduce the Crypto100 index, representing the top 100 cryptocurrencies by market capitalization, and propose a novel feature reduction algorithm to identify the most impactful and resilient features from diverse data sources. Our comprehensive experiments demonstrate that data source diversity significantly enhances the predictive performance of forecasting models across different time horizons. Key findings include the paramount importance of on-chain metrics for both short-term and long-term predictions, the growing relevance of traditional market indices and macroeconomic indicators for longer-term forecasts, and substantial improvements in model accuracy when diverse data sources are utilized. These insights help demystify the short-term and long-term driving factors of the cryptocurrency market and lay the groundwork for developing more accurate and resilient forecasting models.
Ruixue Jing, Ryota Kobayashi, Luis Enrique Correa Rocha
The rapidly evolving cryptocurrency market presents unique challenges for investment due to its inherent volatility and evolving regulatory environment. Collective price movements can be exploited to construct diversified portfolios with improved risk-return profiles. This paper introduces an integrated framework that combines network analysis, price forecasting, and portfolio theory to identify stable groups of highly correlated cryptocurrencies for profitable portfolio construction. We employ the Louvain community detection algorithm together with consensus clustering to extract temporally persistent correlation clusters, and incorporate ARIMA-based price forecasts to enhance forward-looking cluster formation. Using 5 years of daily closing prices, we evaluate portfolio performance across multiple strategies and holding horizons, assessing both profitability and downside risk with return-based and tail-risk metrics. Our empirical results show that predictive consensus-clustering portfolios maintain consistently positive and stable performance up to a 14-day horizon, exhibit favourable gain-loss asymmetry, and achieve tighter tail-risk control. These findings demonstrate that stable interdependencies in cryptocurrency markets can be leveraged to construct profitable and risk-aware portfolios across short-term holding horizons.
Although modern blockchains almost universally produce blocks at fixed intervals, existing models still lack an analytical formula for the loss-versus-rebalancing (LVR) incurred by Automated Market Makers (AMMs) liquidity providers in this setting. Leveraging tools from random walk theory, we derive the following closed-form approximation for the per block per unit of liquidity expected LVR under constant block time: \[ \overline{\mathrm{ARB}}= \frac{\,σ_b^{2}} {\,2+\sqrt{2π}\,γ/(|ζ(1/2)|\,σ_b)\,}+O\!\bigl(e^{-\mathrm{const}\tfracγ{σ_b}}\bigr)\;\approx\; \frac{σ_b^{2}}{\,2 + 1.7164\,γ/σ_b}, \] where $σ_b$ is the intra-block asset volatility, $γ$ the AMM spread and $ζ$ the Riemann Zeta function. Our large Monte Carlo simulations show that this formula is in fact quasi-exact across practical parameter ranges. Extending our analysis to arbitrary block-time distributions as well, we demonstrate both that--under every admissible inter-block law--the probability that a block carries an arbitrage trade converges to a universal limit, and that only constant block spacing attains the asymptotically minimal LVR. This shows that constant block intervals provide the best possible protection against arbitrage for liquidity providers.
Tarun Chitra, Theo Diamandis, Nathan Sheng, Luke Sterle · 5 authors
Decentralized perpetuals protocols have collectively reached billions of dollars of daily trading volume, yet are still not serious competitors on the basis of trading volume with centralized venues such as Binance. One of the main reasons for this is the high cost of capital for market makers and sophisticated traders in decentralized settings. Recently, numerous decentralized finance protocols have been used to improve borrowing costs for perpetual futures traders. We formalize this class of mechanisms utilized by protocols such as Jupiter, Hyperliquid, and GMX, which we term~\emph{Perpetual Demand Lending Pools} (PDLPs). We then formalize a general target weight mechanism that generalizes what GMX and Jupiter are using in practice. We explicitly describe pool arbitrage and expected payoffs for arbitrageurs and liquidity providers within these mechanisms. Using this framework, we show that under general conditions, PDLPs are easy to delta hedge, partially explaining the proliferation of live hedged PDLP strategies. Our results suggest directions to improve capital efficiency in PDLPs via dynamic parametrization.
Ayush Jha, Abootaleb Shirvani, Ali Jaffri, Svetlozar T. Rachev · 5 authors
This study presents the Adaptive Minimum-Variance Portfolio (AMVP) framework and the Adaptive Minimum-Risk Rate (AMRR) metric, innovative tools designed to optimize portfolios dynamically in volatile and nonstationary financial markets. Unlike traditional minimum-variance approaches, the AMVP framework incorporates real-time adaptability through advanced econometric models, including ARFIMA-FIGARCH processes and non-Gaussian innovations. Empirical applications on cryptocurrency and equity markets demonstrate the proposed framework's superior performance in risk reduction and portfolio stability, particularly during periods of structural market breaks and heightened volatility. The findings highlight the practical implications of using the AMVP and AMRR methodologies to address modern investment challenges, offering actionable insights for portfolio managers navigating uncertain and rapidly changing market conditions.
This study investigates three central questions in portfolio optimization. First, whether time-varying moment estimators outperform conventional sample estimators in practical portfolio construction. Second, whether incorporating a turnover penalty into the optimization objective can improve out-of-sample performance. Third, what type of optimal portfolio strategies can consistently outperform the naive 1/N benchmark. Using empirical evidence from the cryptocurrencies market, this paper provides comprehensive answers to these questions. In the process, several additional findings are uncovered, offering further insights into the dynamics of portfolio construction in highly volatile asset classes.
Bitcoin, widely recognized as the first cryptocurrency, has shown increasing integration with traditional financial markets, particularly major U.S. equity indices, amid accelerating institutional adoption. This study examines how Bitcoin exchange-traded funds and corporate Bitcoin holdings affect correlations with the Nasdaq 100 and the S&P 500, using rolling-window correlation, static correlation coefficients, and an event-study framework on daily data from 2018 to 2025.Correlation levels intensified following key institutional milestones, with peaks reaching 0.87 in 2024, and they vary across market regimes. These trends suggest that Bitcoin has transitioned from an alternative asset toward a more integrated financial instrument, carrying implications for portfolio diversification, risk management, and systemic stability. Future research should further investigate regulatory and macroeconomic factors shaping these evolving relationships.
This paper presents a Multi Agent Bitcoin Trading system that utilizes Large Language Models (LLMs) for alpha generation and portfolio management in the cryptocurrencies market. Unlike equities, cryptocurrencies exhibit extreme volatility and are heavily influenced by rapidly shifting market sentiments and regulatory announcements, making them difficult to model using static regression models or neural networks trained solely on historical data. The proposed framework overcomes this by structuring LLMs into specialised agents for technical analysis, sentiment evaluation, decision-making, and performance reflection. The agents improve over time via a novel verbal feedback mechanism where a Reflect agent provides daily and weekly natural-language critiques of trading decisions. These textual evaluations are then injected into future prompts of the agents, allowing them to adjust allocation logic without weight updates or finetuning. Back-testing on Bitcoin price data from July 2024 to April 2025 shows consistent outperformance across market regimes: the Quantitative agent delivered over 30\% higher returns in bullish phases and 15\% overall gains versus buy-and-hold, while the sentiment-driven agent turned sideways markets from a small loss into a gain of over 100\%. Adding weekly feedback further improved total performance by 31\% and reduced bearish losses by 10\%. The results demonstrate that verbal feedback represents a new, scalable, and low-cost approach of tuning LLMs for financial goals.
The article investigates the usage of Informer architecture for building automated trading strategies for high frequency Bitcoin data. Three strategies using Informer model with different loss functions: Root Mean Squared Error (RMSE), Generalized Mean Absolute Directional Loss (GMADL) and Quantile loss, are proposed and evaluated against the Buy and Hold benchmark and two benchmark strategies based on technical indicators. The evaluation is conducted using data of various frequencies: 5 minute, 15 minute, and 30 minute intervals, over the 6 different periods. Although the Informer-based model with Quantile loss did not outperform the benchmark, two other models achieved better results. The performance of the model using RMSE loss worsens when used with higher frequency data while the model that uses novel GMADL loss function is benefiting from higher frequency data and when trained on 5 minute interval it beat all the other strategies on most of the testing periods. The primary contribution of this study is the application and assessment of the RMSE, GMADL, and Quantile loss functions with the Informer model to forecast future returns, subsequently using these forecasts to develop automated trading strategies. The research provides evidence that employing an Informer model trained with the GMADL loss function can result in superior trading outcomes compared to the buy-and-hold approach.
A simple model-free and distribution-free statistic, the functional relationship between the number of "effective" degrees of freedom and portfolio size, or N*(N), is used to discriminate between two alternative models for the correlation of daily cryptocurrency returns within a retail universe of defined by the list of tradable assets available to account holders at the Robinhood brokerage. The average pairwise correlation between daily cryptocurrency returns is found to be high (of order 60%) and the data collected supports description of the cross-section of returns by a simple isotropic correlation model distinct from a decomposition into a linear factor model with additive noise with high confidence. This description appears to be relatively stable through time.
We propose a new way of building portfolios of cryptocurrencies that provide good diversification properties to investors. First, we seek to filter these digital assets by creating some clusters based on their path signature. The goal is to identify similar patterns in the behavior of these highly volatile assets. Once such clusters have been built, we propose "optimal" portfolios by comparing the performances of such portfolios to a universe of unfiltered digital assets. Our intuition is that clustering based on path signatures will make it easier to capture the main trends and features of a group of cryptocurrencies, and allow parsimonious portfolios that reduce excessive transaction fees. Empirically, our assumptions seem to be satisfied.
In this research paper, we investigate into a paper named "A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem" [arXiv:1706.10059]. It is a portfolio management problem which is solved by deep learning techniques. The original paper proposes a financial-model-free reinforcement learning framework, which consists of the Ensemble of Identical Independent Evaluators (EIIE) topology, a Portfolio-Vector Memory (PVM), an Online Stochastic Batch Learning (OSBL) scheme, and a fully exploiting and explicit reward function. Three different instants are used to realize this framework, namely a Convolutional Neural Network (CNN), a basic Recurrent Neural Network (RNN), and a Long Short-Term Memory (LSTM). The performance is then examined by comparing to a number of recently reviewed or published portfolio-selection strategies. We have successfully replicated their implementations and evaluations. Besides, we further apply this framework in the stock market, instead of the cryptocurrency market that the original paper uses. The experiment in the cryptocurrency market is consistent with the original paper, which achieve superior returns. But it doesn't perform as well when applied in the stock market.