I have personally used a previous version. It’s archived and was never registered because it had some ugly spaghetti code that made it increasingly difficult to develop. So I rewrote it from scratch to make it more maintainable, testable, and robust.
I had 3 long-only portfolios (myself, my mum and my partner) over 2-3 years starting in 2022. They all beat the market by quite a margin, especially in 2022 where they made money when the S&P lost about 20%. We had to take out the money out earlier this year for a house deposit.
My goal is to make it useful for anyone regardless of expertise level. I’ve been mostly focusing on adding features, and only now that it’s very feature rich am i adding convenience functionality (mainly aliases).
I’m about to release a new version which adds pipelines and the ability to perform cross validations and hyper-parameter tuning on them. That’s to say everything from asset selection to priors, to clustering, to optimisations.
I’ll then move on to working on enabling time-dependent constraints. So it’s possible to provide either a vector or a function which will vary the huge number of constraints depending on the walk-forward fold being computed.
I also want to have vine-copula based synthetic returns generation for a synthetic data prior. And cross-sectional factor priors.