Omega Point Blog

Evaluating Portfolios Using AI

Omer Cedar

Evaluating Portfolios Using AI

April 23, 2018

Is this a good or bad portfolio?


This is a question we're asked often. Academics will tell you that positive exposure to momentum and value is generally good, but positive exposure to volatility is generally bad. So, what's the answer?

As a starting point, let me reformulate the original question to: “How is this portfolio positioned in the current market environment?”

To answer the reformulated question, we've trained a deep learning-based system that defines the current market environment based on the interplay of the factors in the market across a universe of 8000+ securities. For example, in higher volatility securities we typically see that investors have a stronger preference to quality, while this relationship is weaker in low volatility securities. As markets shift this relationship can change, so our AI-training balances a multi-year history with information from the more recent environment.

Screen Shot 2018-04-30 at 11.43.35 AM

Every day, the system provides a single multifactor score (“market aware score”) for each security in the market based on its combined input factor exposures. Portfolios are scored based on the weighted sum of their market aware score in the securities.


This market aware score can be interpreted as an excess daily beta to the current market environment. For example, a score of +1.0 suggests that assuming current market environment (as defined by the factors) continues, the portfolio would outperform the market by 1% per day. In practice, we've seen the score as a forward one-month leading indicator of a portfolio's relative performance to the market.

On the snapshot date of the portfolio (Dec 26, 2016), its combined score was -0.53, meaning that the portfolio was poorly positioned in the current market environment (post-Trump rally). The same portfolio scored as high as +0.75 during the factor meltdown at the end of 2015 and early 2016, when investors showed a preference for quality and lower beta securities.

How do our customers use this score in practice?


We offer several approaches for customers to integrate this score:

  1. Through our insights engine (shown above) — portfolio managers can test various approaches to rebalance the portfolio using the market aware score and have it automatically run on a go-forward basis.
  2. Standalone strategy - allocators looking to use this score as an independent source of alpha can work with us to develop an index that can be leveraged in a multifactor mandate.
  3. Alerts & reporting - risk groups can customize alerts to identify changes in portfolio composition or the market environment that may create uncertain conditions for the portfolio.

Feel free to contact us to learn how we can help with your use case.