Abstract: This tutorial will center around on analyzing investment performance. This is a surprisingly contentious topic in the investment industry, and for good reason. A lot more discretionary decisions go into this process than one would expect. Yet if you do not know how to analyze your performance, you cannot improve your research or your algorithms, and nor can you benchmark yourself appropriately.
We will start from the top down. First we will review aggregate measures of performance for your portfolio and how these are calculated, Sharpe and Sortino being the most common ones. Not only are many people, including in the industry, unable to give their exact definitions, but the formulas for calculating them leave room for personal decisions in their assumptions.
We will then dive into a multitude of ways of breaking down the performance of your investment algorithms at both the portfolio and the security level:
performance in security selection
performance in sector allocation
performance on the long vs the short side
performance in managing your net exposure
correlation to factors
appropriate aggregate portfolio risk metrics
Throughout the talk we will aim to improve our understanding of pure alpha and how you can measure for it.
Given the importance of backtests in systematic finance, I will then review how to build a robust backtest machine and common pitfalls. In this section we will also look at backtests together and try to analyze them.
At the coding level, I will be using pandas in a jupyter notebook. We will use programming purely to support our conceptual discussion. I will explain why a manual approach works best here and will touch upon packages developed for quantitative finance. But the focus will remain throughout on the analysis and metrics, not on programming prowess.
Based on my interactions with both the professional systematic investing industry, and personal investors, I do believe that this would provide for a helpful discussion for the audience as part of the conference’s focus on quant finance this year.
I intend to make a jupyter notebook with all the concepts, explanations, and code, available to the audience of course.
Bio: Alexandru works as the COO of Empirical Capital, a quantitative investment firm based in London, and also teaches a module on data analytics for healthcare at Pierre et Marie Curie University in Paris.
COO at Empirical Capital Analytics and Lecturer at Pierre et Marie Curie University