Electronic Trading
The transition from voice trading of liquid high volume assets like equity and FX to electronic trading occurred some time ago. Now, many institutions face big challenges to move as much as possible of their business onto electronic trading platforms. Computer algorithms execute the orders and make substantive decisions without any human intervention. There is a growing need to develop better quantitative trading algorithms. OMI expertise in this area comes from many directions. In algorithmic trading, academic members are experts in the areas of algorithmic and high-frequency trading. There is extensive work on limit order books, optimal execution, and market making, where the main tools are drawn from market microstructure, stochastic optimal control, and machine learning.
Recent relevant publications can be found here and here
Data Analysis and Patterns in Data
The existence of easily accessible big data sets and the ability to extract meaningful information from them will shape the future in many research areas. From the analysis of the history of financial data, coupled with the history of the sentiment extracted from the web, one can endeavour to answer questions which will be important to all stakeholders in financial markets. Large heterogeneous data sets demand development of novel methodology to better describe and detect patterns.
The OMI aims to produce research using tools that benefit from the availability of this data, and endeavours to produce novel algorithmic innovations in machine learning which have downstream financial applications. Examples of this include deep learning, time-series forecasting, graph-based machine learning, natural language processing, reinforcement learning, Bayesian machine learning or causal inference. Expertise in answering fundamental questions in these fields allows for an improved understanding on how to perform financial forecasts, give insight on the connectivity between financial assets or allow to quantify the uncertainty of model predictions.
Recent relevant papers can be found here and here
Natural Language Processing
Market agents are exchanging a substantial amount of information in natural language format via text and audio data. Social media posts, news articles, central bank statements, analyst reports, and company filings are just a few examples of the wide array of potential sources. Advancements in modern natural language processing (NLP) methods allow for ever more nuanced and precise automated information processing and signal extraction. Particularly, the fusion of traditional tabular financial data with text data seems promising to enrich financial and economic analyses.
OMI members conduct NLP research in finance, for example, by working on improved multimodal machine learning models, investigating pivotal time-series structures for NLP models in financial contexts, and analysing news signals in central bank communication.
The OMI also hosts the annual research conference on NLP for Social Data Sciences and its members co-organize the OxNLP group.
Multi-Agent Systems in Finance
The rise of artificial intelligence and adoption of autonomous agents will shape many aspects of financial markets. There is a growing need to further understand the interactions of multiple autonomous agents and the impact they have on prices, liquidity, and the efficiency and integrity of financial markets.
The OMI is at the forefront of building the necessary theory for relevant stakeholders to understand how markets can become more efficient to reinforce competition, and to also understand various risks such as collusion and bubbles. The main tools come from game theory, stochastic approximation, optimal control, and misspecified learning.
Recent relevant papers can be found here
Decision Making under Uncertainty, Asset Allocation and Pricing
Having to act in a context of uncertainty, or “take risks” is at the centre of much of human endeavour. Risks are hard to quantify and it is not always straightforward to make a decision under uncertainty. In finance, risks stem from the randomness of a future outcome (e.g., unexpected changes in: prices, demand, supply, etc.) and from assuming that a model is a correct representation of a financial system. In both cases, deciding what is an optimal financial strategy or policy, requires a deep understanding of how key financial variables are interconnected to understand the system and to make predictions.