I agree that there are multiple valid reasons to use a benchmark—one of which is to estimate opportunity cost, which varies depending on where someone would invest if not actively managing their portfolio. That’s a personal decision, and it seems reasonable that users should be able to choose a benchmark that reflects their situation.
I already use a benchmark I’m satisfied with, but I don’t mind if P123 adds others that users find helpful. There’s no harm in offering more options—unless the maintenance cost is too high, which is P123’s call.
Yuval has a point about this.
This is especially important when designing ML models. The ability to isolate stock-picking skill—independent of market drift—is central to that process. So a benchmark based on the equal-weighted universe is ideal.
But is it just me, or is it relatively difficult to use the universe itself as a benchmark within P123’s AI/ML module?
Or any equal-weighted benchmark for that matter? Other than the SP500 equal weight index I do not see any other equal weight benchmarks. An equal weight benchmark for the Russell 2000 would be better for most of us, I think.
Maybe there’s an easy way to create an aggregated series and use that to generate an excess-returns target in the AI/ML module—if so, it would be great to have a walkthrough. I suspect the capability exists and has just been overlooked. As it stands, it’s often easier to download the data and calculate excess returns relative to the universe externally.
It would be unfortunate if that ends up being the main reason users avoid the AI/ML module in favor of offline work.
I voted for this. I understand that P123 already offers a large number of benchmarks and can’t keep adding new ones indefinitely—especially if there’s development or maintenance overhead. But I’d vote to add this benchmark and deprecate the 1–5 year corporate bond benchmark, which I can’t imagine ever using—either in a stock portfolio or even in an ETF strategy.
I would also vote for an equal-weight Easy to Trade Universe as a benchmark or just a clear way to make the universe a benchmark for models and the AI/ML.