Tuesday, May 7, 2024

Multiple models versus an ensemble


"A man with two watches is never sure (what time it is)." 

- Segal's Law 

There is the view that having competing models is a bad thing. Investors should form one model that incorporates all their thinking. If you have more than one model you will never know or believe what is the true reality.  Yet, we also know that models will fail. We will make mistakes. Most models will explain only a small portion of the variation in asset returns.  

Yes, having a single model is theoretically pure, but I will usually choose to form an ensemble. An average prediction from an ensemble will usually do better than a single model. The reality is that prediction is about getting it right and if there is a means that is not perfect but gets us closer to a better answer should always be preferred.

No comments: