Problems and questions in hedge fund return generation

Questions that are worth considering. Driven by the presentation "Problems that keep me awake at night" by Giuseppe Paleologo at the Fordham Quantvision 2024.

Each of the these are key "big questions" that will impact any model or research. 

1. How do you better incorporate flow into modeling, "Your flow is my alpha"? Flow creates price pressure which should lead to dislocations that will mean revert. How do you measure this flow and how do you find flow that is meaningful? 

2. Still have a problem with crowding. Accept that it occurs but how can it be exploited? If you are first, you want crowding, but when does it become too much?

3. What is the best way to combine signals. Use separate models and then net signals? Combine signals into a single model as different factors in a regression? How do you control costs with multiple signals?

4. How do you more effectively deal with factor or strategy rankings? Most of long/short factors based on buying and selling portfolios based on rankings, yet the distance between ranks can be highly variable. How do you rank multiple factors? 

5. How do you form effective feedback loops from trading, or how do you learn from mistakes. This is critical with decision-making and is only more difficult when you have a "wicked" versus "kind" learning environment.

6. We know that returns are not normal, but how do you effectively deal with fat-tailed world beyond Winsorizing? Fat-tailed distributions are often caused by mixed distributions, so how do you deal with changing distributions? All the real action is in the tails.