Preview of our v1.1 of AI Factor: intelligent features, grid search, categorical & macro

[danparquette]

@pitmaster

This factor would be useful but it does not work in ranking systems:

Eval(IndCount > 20, FRank("GMgn%Gr%TTM", #Industry), FRank("GMgn%Gr%TTM", #subsector))

You can write it like this:
Eval(FCount("1", #Industry)>=20, FRank("GMgn%Gr%TTM", #Industry), FRank("GMgn%Gr%TTM", #Subsector))

[pitmaster]

This fallback improves my ranking systems especially for small European universes + intuitively it makes sense + it should lower turnover. Thanks @danparquette & @AlgoMan !

I'm figuring out what should be the threshold. Currently I use 3% of total stocks in universe.

[AlgoMan]

The use of fallback is even more important in ML, normalizing with a few number of stocks does not work well, almost a guarantee that stocks will fallout of the deviation threshold.