Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Estimating the Sharpe ratio using a Kalman filter

I've gone through all of the lectures and notebooks on Quantopian regarding Kalman filters as well as whatever other material I could find online. While it seems rather straightforward now to "plug in" a linear equation into a Kalman filter to smooth out and estimate the inputs, I'm not quite sure where to start when trying to estimate something like the Sharpe ratio. I was thinking of estimating the mean and standard deviation separately followed by computing the Sharpe ratio, but then I realized the filter should have all the information necessary to compute the ACTUAL Sharpe ratio.

I read this paper on how to create the so-called "fundamental" Sharpe ratio using a Kalman filter, but it's rather complex and maybe even overkill.

My question is how should I go about trying to estimate the Sharpe ratio using a Kalman filter? Any information you might have or resources pointing me in the right direction would go a long way!

Edit: I also want to point out that this was referenced in "Kalman Filters" lecture: "For instance, if we have already computed the moving Sharpe ratio, we can smooth it using a Kalman filter."

2 responses

Note that since the SR is a ratio, if you take the log, you'll end up with difference between two terms. So, if the Kalman filter operates like K(a-b) = K(a) - K(b), then maybe the trick of taking the log would help. A half-baked idea at this point, but I thought I'd share it, regardless.

That's an idea worth exploring for sure. Thank Grant!