arXiv subjects

Fabrizio Lillo

Publications and source records attributed to Fabrizio Lillo.

At least 37 records · Page 2Linked to original sources

arXiv2025

Tackling estimation risk in Kelly investing using options

The Kelly criterion provides a general framework for optimizing the growth rate of an investment portfolio over time by maximizing the expected logarithmic utility of wealth. However, the optimality condition of the Kelly criterion is highly sensitive to accurate estimates of the probabilities and investment payoffs. Estimation risk can lead to greatly suboptimal portfolios. In a simple binomial model, we show that the introduction of a European option in the Kelly framework can be used to construct a class of growth optimal portfolios that are robust to estimation risk.

q-fin.MF↗

arXiv2025

Deviations from Tradition: Stylized Facts in the Era of DeFi

Decentralized Exchanges (DEXs) are now a significant component of the financial world where billions of dollars are traded daily. Differently from traditional markets, which are typically based on Limit Order Books, DEXs typically work as Automated Market Makers, and, since the implementation of Uniswap v3, feature concentrated liquidity. By investigating the twenty-four most active pools in Uniswap v3 during 2023 and 2024, we empirically study how this structural change in the organization of the markets modifies the well-studied stylized facts of prices, liquidity, and order flow observed in traditional markets. We find a series of new statistical regularities in the distributions and cross-autocorrelation functions of these variables that we are able to associate either with the market structure (e.g., the execution of orders in blocks) or with the intense activity of Maximal Extractable Value searchers, such as Just-in-Time liquidity providers and sandwich attackers.

q-fin.TR↗

arXiv2025

A high-frequency approach to Realized Risk Measures

We propose a new approach, termed Realized Risk Measures (RRM), to estimate Value-at-Risk (VaR) and Expected Shortfall (ES) using high-frequency financial data. It extends the Realized Quantile (RQ) approach proposed by Dimitriadis and Halbleib by lifting the assumption of return self-similarity, which displays some limitations in describing empirical data. More specifically, as the RQ, the RRM method transforms intra-day returns in intrinsic time using a subordinator process, in order to capture the inhomogeneity of trading activity and/or volatility clustering. Then, microstructural effects resulting in non-zero autocorrelation are filtered out using a suitable moving average process. Finally, a fat-tailed distribution is fitted on the cleaned intra-day returns. The return distribution at low frequency (daily) is then extrapolated via either a characteristic function approach or Monte Carlo simulations. VaR and ES are estimated as the quantile and the tail mean of the distribution, respectively. The proposed approach is benchmarked against the RQ through several experiments. Extensive numerical simulations and an empirical study on 18 US stocks show the outperformance of our method, both in terms of the in-sample estimated risk measures and in the out-of-sample risk forecasting

q-fin.RM↗

arXiv2025

Bayesian Autoregressive Online Change-Point Detection with Time-Varying Parameters

Change points in real-world systems mark significant regime shifts in system dynamics, possibly triggered by exogenous or endogenous factors. These points define regimes for the time evolution of the system and are crucial for understanding transitions in financial, economic, social, environmental, and technological contexts. Building upon the Bayesian approach introduced in \cite{c:07}, we devise a new method for online change point detection in the mean of a univariate time series, which is well suited for real-time applications and is able to handle the general temporal patterns displayed by data in many empirical contexts. We first describe time series as an autoregressive process of an arbitrary order. Second, the variance and correlation of the data are allowed to vary within each regime driven by a scoring rule that updates the value of the parameters for a better fit of the observations. Finally, a change point is detected in a probabilistic framework via the posterior distribution of the current regime length. By modeling temporal dependencies and time-varying parameters, the proposed approach enhances both the estimate accuracy and the forecasting power. Empirical validations using various datasets demonstrate the method's effectiveness in capturing memory and dynamic patterns, offering deeper insights into the non-stationary dynamics of real-world systems.

stat.ML↗

arXiv2025

Modeling metaorder impact with a Non-Markovian Zero Intelligence model

Devising models of the limit order book that realistically reproduce the market response to exogenous trades is extremely challenging and fundamental in order to test trading strategies. We propose a novel explainable model for small tick assets, the Non-Markovian Zero Intelligence, which is a variant of the well-known Zero Intelligence model. The main modification is that the probability of limit orders' signs (buy/sell) is not constant but is a function of the exponentially weighted mid-price return, representing the past price dynamics, and can be interpreted as the reaction of traders with reservation prices to the price trend. With numerical simulations and analytical arguments, we show that the model predicts a concave price path during a metaorder execution and to a price reversion after the execution ends, as empirically observed. We analyze in-depth the mechanism at the root of the arising concavity, the components which constitute the price impact in our model, and the dependence of the results on the two main parameters, namely the time scale and the strength of the reaction of traders to the price trend.

q-fin.TR↗

arXiv2024

Score-Driven Exponential Random Graphs: A New Class of Time-Varying Parameter Models for Dynamical Networks

Motivated by the increasing abundance of data describing real-world networks that exhibit dynamical features, we propose an extension of the Exponential Random Graph Models (ERGMs) that accommodates the time variation of its parameters. Inspired by the fast-growing literature on Dynamic Conditional Score models, each parameter evolves according to an updating rule driven by the score of the ERGM distribution. We demonstrate the flexibility of score-driven ERGMs (SD-ERGMs) as data-generating processes and filters and show the advantages of the dynamic version over the static one. We discuss two applications to temporal networks from financial and political systems. First, we consider the prediction of future links in the Italian interbank credit network. Second, we show that the SD-ERGM allows discriminating between static or time-varying parameters when used to model the U.S. Congress co-voting network dynamics.

stat.AP↗

arXiv2024

Modelling shock propagation and resilience in financial temporal networks

Modelling how a shock propagates in a temporal network and how the system relaxes back to equilibrium is challenging but important in many applications, such as financial systemic risk. Most studies so far have focused on shocks hitting a link of the network, while often it is the node and its propensity to be connected that are affected by a shock. Using as starting point the configuration model, a specific Exponential Random Graph model, we propose a vector autoregressive (VAR) framework to analytically compute the Impulse Response Function (IRF) of a network metric conditional to a shock on a node. Unlike the standard VAR, the model is a nonlinear function of the shock size and the IRF depends on the state of the network at the shock time. We propose a novel econometric estimation method that combines the Maximum Likelihood Estimation and Kalman filter to estimate the dynamics of the latent parameters and compute the IRF, and we apply the proposed methodology to the dynamical network describing the electronic Market of Interbank Deposit (e-MID).

econ.GN↗

arXiv2024

Dimensionality reduction techniques to support insider trading detection

Identification of market abuse is an extremely complicated activity that requires the analysis of large and complex datasets. We propose an unsupervised machine learning method for contextual anomaly detection, which allows to support market surveillance aimed at identifying potential insider trading activities. This method lies in the reconstruction-based paradigm and employs principal component analysis and autoencoders as dimensionality reduction techniques. The only input of this method is the trading position of each investor active on the asset for which we have a price sensitive event (PSE). After determining reconstruction errors related to the trading profiles, several conditions are imposed in order to identify investors whose behavior could be suspicious of insider trading related to the PSE. As a case study, we apply our method to investor resolved data of Italian stocks around takeover bids.

q-fin.ST↗

arXiv2024

Online Learning of Order Flow and Market Impact with Bayesian Change-Point Detection Methods

Financial order flow exhibits a remarkable level of persistence, wherein buy (sell) trades are often followed by subsequent buy (sell) trades over extended periods. This persistence can be attributed to the division and gradual execution of large orders. Consequently, distinct order flow regimes might emerge, which can be identified through suitable time series models applied to market data. In this paper, we propose the use of Bayesian online change-point detection (BOCPD) methods to identify regime shifts in real-time and enable online predictions of order flow and market impact. To enhance the effectiveness of our approach, we have developed a novel BOCPD method using a score-driven approach. This method accommodates temporal correlations and time-varying parameters within each regime. Through empirical application to NASDAQ data, we have found that: (i) Our newly proposed model demonstrates superior out-of-sample predictive performance compared to existing models that assume i.i.d. behavior within each regime; (ii) When examining the residuals, our model demonstrates good specification in terms of both distributional assumptions and temporal correlations; (iii) Within a given regime, the price dynamics exhibit a concave relationship with respect to time and volume, mirroring the characteristics of actual large orders; (iv) By incorporating regime information, our model produces more accurate online predictions of order flow and market impact compared to models that do not consider regimes.

q-fin.TR↗

arXiv2024

Reinforcement Learning for Optimal Execution when Liquidity is Time-Varying

Optimal execution is an important problem faced by any trader. Most solutions are based on the assumption of constant market impact, while liquidity is known to be dynamic. Moreover, models with time-varying liquidity typically assume that it is observable, despite the fact that, in reality, it is latent and hard to measure in real time. In this paper we show that the use of Double Deep Q-learning, a form of Reinforcement Learning based on neural networks, is able to learn optimal trading policies when liquidity is time-varying. Specifically, we consider an Almgren-Chriss framework with temporary and permanent impact parameters following several deterministic and stochastic dynamics. Using extensive numerical experiments, we show that the trained algorithm learns the optimal policy when the analytical solution is available, and overcomes benchmarks and approximated solutions when the solution is not available.

q-fin.TR↗

arXiv2023

Transient impact from the Nash equilibrium of a permanent market impact game

A large body of empirical literature has shown that market impact of financial prices is transient. However, from a theoretical standpoint, the origin of this temporary nature is still unclear. We show that an implied transient impact arises from the Nash equilibrium between a directional trader and one arbitrageur in a market impact game with fixed and permanent impact. The implied impact is the one that can be empirically inferred from the directional trader's trading profile and price reaction to order flow. Specifically, we propose two approaches to derive the functional form of the decay kernel of the Transient Impact Model, one of the most popular empirical models for transient impact, from the behaviour of the directional trader at the Nash equilibrium. The first is based on the relationship between past order flow and future price change, while in the second we solve an inverse optimal execution problem. We show that in the first approach the implied kernel is unique, while in the second case infinite solutions exist and a linear kernel can always be inferred.

q-fin.TR↗

arXiv2022

A machine learning approach to support decision in insider trading detection

Identifying market abuse activity from data on investors' trading activity is very challenging both for the data volume and for the low signal to noise ratio. Here we propose two complementary unsupervised machine learning methods to support market surveillance aimed at identifying potential insider trading activities. The first one uses clustering to identify, in the vicinity of a price sensitive event such as a takeover bid, discontinuities in the trading activity of an investor with respect to his/her own past trading history and on the present trading activity of his/her peers. The second unsupervised approach aims at identifying (small) groups of investors that act coherently around price sensitive events, pointing to potential insider rings, i.e. a group of synchronised traders displaying strong directional trading in rewarding position in a period before the price sensitive event. As a case study, we apply our methods to investor resolved data of Italian stocks around takeover bids.

q-fin.ST↗

arXiv2022

Network-wide assessment of ATM mechanisms using an agent-based model

This paper presents results from the SESAR ER3 Domino project. Three mechanisms are assessed at the ECAC-wide level: 4D trajectory adjustments (a combination of actively waiting for connecting passengers and dynamic cost indexing), flight prioritisation (enabling ATFM slot swapping at arrival regulations), and flight arrival coordination (where flights are sequenced in extended arrival managers based on an advanced cost-driven optimisation). Classical and new metrics, designed to capture network effects, are used to analyse the results of a micro-level agent-based model. A scenario with congestion at three hubs is used to assess the 4D trajectory adjustment and the flight prioritisation mechanisms. Two different scopes for the extended arrival manager are modelled to analyse the impact of the flight arrival coordination mechanism. Results show that the 4D trajectory adjustments mechanism succeeds in reducing costs and delays for connecting passengers. A trade-off between the interests of the airlines in reducing costs and those of non-connecting passengers emerges, although passengers benefit overall from the mechanism. Flight prioritisation is found to have no significant effects at the network level, as it is applied to a small number of flights. Advanced flight arrival coordination, as implemented, increases delays and costs in the system. The arrival manager optimises the arrival sequence of all flights within its scope but does not consider flight uncertainties, thus leading to sub-optimal actions.

physics.soc-ph↗

arXiv2022

From Zero-Intelligence to Queue-Reactive: Limit Order Book modeling for high-frequency volatility estimation and optimal execution

The estimation of the volatility with high-frequency data is plagued by the presence of microstructure noise, which leads to biased measures. Alternative estimators have been developed and tested either on specific structures of the noise or by the speed of convergence to their asymptotic distributions. Gatheral and Oomen (2010) proposed to use the Zero-Intelligence model of the limit order book to test the finite-sample performance of several estimators of the integrated variance. Building on this approach, in this paper we introduce three main innovations: (i) we use as data-generating process the Queue-Reactive model of the limit order book (Huang et al. (2015)), which - compared to the Zero-Intelligence model - generates more realistic microstructure dynamics, as shown here by using an Hausman test; (ii) we consider not only estimators of the integrated volatility but also of the spot volatility; (iii) we show the relevance of the estimator in the prediction of the variance of the cost of a simulated VWAP execution. Overall we find that, for the integrated volatility, the pre-averaging estimator optimizes the estimation bias, while the unified and the alternation estimator lead to optimal mean squared error values. Instead, in the case of the spot volatility, the Fourier estimator yields the optimal accuracy, both in terms of bias and mean squared error. The latter estimator leads also to the optimal prediction of the cost variance of a VWAP execution.

q-fin.TR↗

arXiv2022

How Covid mobility restrictions modified the population of investors in Italian stock markets

This paper investigates how Covid mobility restrictions impacted the population of investors of the Italian stock market. The analysis tracks the trading activity of individual investors in Italian stocks in the period January 2019-September 2021, investigating how their composition and the trading activity changed around the Covid-19 lockdown period (March 9 - May 19, 2020) and more generally in the period of the pandemic. The results pinpoint that the lockdown restriction was accompanied by a surge in interest toward stock market, as testified by the trading volume by households. Given the generically falling prices during the lockdown, the households, which are typically contrarian, were net buyers, even if less than expected from their trading activity in 2019. This can be explained by the arrival, during the lockdown, of a group of about 185k new investors (i.e. which had never traded since January 2019) which were on average ten year younger and with a larger fraction of males than the pre-lockdown investors. By looking at the gross P&L, there is clear evidence that these new investors were more skilled in trading. There are thus indications that the lockdown, and more generally the Covid pandemic, created a sort of regime change in the population of financial investors.

q-fin.TR↗

arXiv2022

Score Driven Generalized Fitness Model for Sparse and Weighted Temporal Networks

While the vast majority of the literature on models for temporal networks focuses on binary graphs, often one can associate a weight to each link. In such cases the data are better described by a weighted, or valued, network. An important well known fact is that real world weighted networks are typically sparse. We propose a novel time varying parameter model for sparse and weighted temporal networks as a combination of the fitness model, appropriately extended, and the score driven framework. We consider a zero augmented generalized linear model to handle the weights and an observation driven approach to describe time varying parameters. The result is a flexible approach where the probability of a link to exist is independent from its expected weight. This represents a crucial difference with alternative specifications proposed in the recent literature, with relevant implications for the flexibility of the model. Our approach also accommodates for the dependence of the network dynamics on external variables. We present a link forecasting analysis to data describing the overnight exposures in the Euro interbank market and investigate whether the influence of EONIA rates on the interbank network dynamics has changed over time.

stat.AP↗

arXiv2021

Instabilities in Multi-Asset and Multi-Agent Market Impact Games

We consider the general problem of a set of agents trading a portfolio of assets in the presence of transient price impact and additional quadratic transaction costs and we study, with analytical and numerical methods, the resulting Nash equilibria. Extending significantly the framework of Schied & Zhang (2019) and Luo & Schied (2020), who considered the single asset case, we prove the existence and uniqueness of the corresponding Nash equilibria for the related mean-variance optimization problem. We then focus our attention on the conditions on the model parameters making the trading profile of the agents at equilibrium, and as a consequence the price trajectory, wildly oscillating and the market unstable. While Schied & Zhang (2019) and Luo & Schied (2020) highlighted the importance of the value of transaction cost in determining the transition between a stable and an unstable phase, we show that also the scaling of market impact with the number of agents J and the number of assets M determines the asymptotic stability (in J and M ) of markets.

q-fin.TR↗

arXiv2021

Information dynamics of price and liquidity around the 2017 Bitcoin markets crash

We study the information dynamics between the largest Bitcoin exchange markets during the bubble in 2017-2018. By analysing high-frequency market-microstructure observables with different information theoretic measures for dynamical systems, we find temporal changes in information sharing across markets. In particular, we study the time-varying components of predictability, memory, and synchronous coupling, measured by transfer entropy, active information storage, and multi-information. By comparing these empirical findings with several models we argue that some results could relate to intra-market and inter-market regime shifts, and changes in direction of information flow between different market observables.

q-fin.ST↗