# Would an R^2 function be useful?

All,

I have been wondering if it would be possible for P123 to supply a coefficient of determination (R^2) function, either for direct use by users or perhaps for review by users for other items, like ranking system performance?

I initially asked about such a function shortly after joining P123 last year. I was interested in the function because in my reading about momentum (which I thought was in the book Quantitative Momentum, but I could not find it), one researcher found that when measuring the different types of momentum, the smoothness of the returns in the period measured, as measured by R^2, predicted higher returns from momentum in the next period. But because there was no such function, I implemented the smoothness of returns by other means called “Frog in the Pan”, which the authors discussed in Chapter 6, p. 100.

Another use for R^2 could be for measuring the goodness of fit for a factor in the bucket performance test for a ranking system. Mr. @Jrinne recently posted on another thread how a poorly fitting factor for performance does more harm than good (see What now? 12345 - #4 by Jrinne). If an R^2 were provided for each factor in a ranking system, the user could tell from the value just how well each factor and the overall ranking system fit the expected performance curve.

Anyway, I may be way off-base here but it is something that I have wondered about regarding P123’s functions.

Cary

We now offer R2 for users with an ultimate subscription. See https://www.portfolio123.com/doc/doc_detail.jsp?factor=R2.

A redo of the bucket performance is one of the projects at the top of the list. The current output is quite pathetic

I love the bucket performance test just the way it is. Improvements: better still I am sure!!! Cannot wait.

For now, not that hard to download the result into Excel and get an R^2. which is just one of the reasons I love the bucket performance test. Others are getting things like slope (which is not totally unrelated to R^2 BTW) from the downloads, maybe? But very useful for many the way it is, I think.