All: I am just getting back into this again, and I have a new model with a corr with meta model of .11- the percentile is saying 3.4, but my understanding had been that a low corr with meta model was actually a good thing for mmc. Can someone clarify?
I think people have done research posts on this here in the forum, or perhaps on Discord, or the old RocketChat days.
I don’t think there’s any real correlation between CWMM and MMC. My recollection was that there’s a sweet spot around 0.7, and things actually get more dangerous for very low CWMM models. That’s just from memory, though, could be all wrong. That was on years old data too.
Nothing can be said just from CWMM.
A good way to think about this is, if the CWMM is as low as yours,
mmc almost equals corr_with_target, because the model has very little overlap with MM hence all of the contribution is differentiated.
Now, having a CWMM near 0, while having a high sharpe corr is very difficult (to put it mildly).
But one must try, for the love of the game.
Best.
1 Like
