About the Performance of Equity Agency Trading Algorithms
The performance of equity agency trading algorithms is driven by hundreds of factors with varying degrees of interaction. These factors range from client instructions and market conditions to various algorithm settings. Often recommendations to improve performance are based on past experience and intuition and employ a lot of discretion. This approach is expensive and does not scale beyond a set of focus clients.
Machine Learning-Based Transaction Cost Analysis
Swagato formulates a methodology using machine learning to sift through troves of order execution data to identify key drivers of algorithm performance and provide actionable recommendations to clients in delivering execution alpha. When a client executes an order, the entire state of the order and the market is stored in a high-performance data repository. He applies machine learning algorithms on this extensive data store to search the parameter space and identify performance drivers ranked by their order of importance.
Using machine learning we are able to analyze and attribute the performance of algorithmic trading orders and provide clients with never before seen insights on the key drivers of execution performance beyond traditional metrics such as average daily volume, spread, and volatility. This approach provides us the ability to focus on the important performance drivers and optimize those for further enhancing algorithm performance. He finds this to be a highly scalable and efficient process versus current Transaction Cost Analysis (TCA) methods that focus on a standard set of metrics with few actionable insights for improving the client execution experience.
This presentation was held at the RavenPack Research Symposium in New York on September 10, 2019.Request Event Materials