January 2023 CDT Workshop

4:00 Lorenzo Lucchese (Imperial CDT Student)

The Short-Term Predictability of Returns in Order Book Markets: A Deep Learning Perspective

In this talk, we will explore the order book-driven predictability of high-frequency returns by leveraging deep learning techniques. First, we will introduce a new and robust representation of the order book, the volume representation. Next, we will conduct an extensive empirical experiment to address various questions regarding predictability. We will investigate if and how far ahead there is predictability, the importance of a robust data representation, the advantages of multi-horizon modelling, and the presence of universal trading patterns. We will use model confidence sets, which provide a formalized statistical inference framework particularly well suited to answer these questions. Our findings show that the class of deep learning models considered in our work can consistently identify predictability in mid-price returns. The performance of the models is strongly dependent on the choice of order book representation, and in this respect, the volume representation appears to have multiple practical advantages.

 

4:35 Aldair Petronilia (Oxford CDT Student)

Systemic Risk Models with Smoothed Endogenous Contagion

Structural models for default contagion typically take the approach of modelling entities via absorbing diffusions on the positive half line. Default occurs when an entity hits the origin and this causes contagion via a jump toward the origin from the other entities. In the large particle limit, this model leads to a McKean–Vlasov equation with positive feedback which can have jumps in the limiting equation if the contagion is strong enough. We propose a new approach to constructing solutions to these limiting McKean–Vlasov equations using an approximation via smoothed contagion. Smoothing the contagion in time removes the jumps in the system and we show that these smoothed approximations converge to the limiting McKean–Vlasov problem. This provides a solution approach to problems of this type which allows more easily for general drift and diffusion coefficients. We also show a rate of convergence up to the first jump time under some regularity conditions on the contagion.

 

5:05 Coffee break

 

5:30 Sam Webster (Rates Risk Management at Millennium)

The Dynamics of Interest Rate Markets

 

 

6:30 Reception