An electronic market-making algorithm.
Doctoral thesis, UCL (University College London).
In finance, a market-maker quotes both a buy and a sell price in a financial instrument or commodity, with the intention of making a profit on the bid/ask spread. The function of market-making in financial instruments is at present almost exclusively the domain of humans. Computational finance (systems used lo make trading, hedging and investment decisions) has made significant progress in recent years but the focus has been on pricing and hedging complex financial instruments. The proportion of trades that are executed electronically has also increased remarkably in all asset classes, yet market-making has remained relatively labor intensive, lacking theoretical research to underpin practical innovation. In this thesis, we investigate a novel and more intuitive theoretical approach which aims to model the market-making process. The fundamental principle of market-making is inventory management: we propose a unified approach, applicable to all asset classes, that manages the risk of a market-making book automatically. This study reveals several interesting properties which are generally intuitive to most professional market-makers but are hard to precisely encode into an automated strategy. The primary contribution of the methodology is that it enables us to grade different market-making policies and then to search for the optimal one. This can be performed not only for a single asset but also for a portfolio of correlated assets. Furthermore, in the case of a portfolio of assets, the framework enables the market marker to specify each asset's price and market-making dynamics. We then apply the theoretical framework to build an automated market-making algorithm for the European government bonds book at Deutsche Bank. A completely automated market-making book requires numerous components, from price discovery to trade capture and risk analysis, in addition to the inventory management engine: we describe in detail how each component is implemented and integrated within the market-making system. Previous academic studies in this field have generally fallen short of any real application, because they do not have access to real trading environments. The performance of the proposed algorithm is assessed in terms of risk-management and profits or losses on a trading book especially set up for this purpose on the Global Rates Trading desk. The initial three months' performance was mediocre with regards to profits, although we used this period to set the main parameters of the system and further develop weak areas. Thereafter we have had positive results in terms of both risk-management and profits. The contributions to science of this thesis are: (i) a mathematical framework to classify different market-making strategies within a portfolio of assets; (ii) a methodology to search for the optimal strategy for specific order arrival and asset price dynamics; (iii) the real-world application and test of the proposed algorithm.
|Title:||An electronic market-making algorithm|
|Additional information:||Authorisation for digitisation not received|
Archive Staff Only