But what if expected returns (vis-a-vis Kelly) can be based on multiple performance drivers?
I don’t disagree with the KISS philosophy, but I am challenging the aversion to generalizing.
Primus, yes that is what I meant, if you use Performance Drivers (Value, Momentum, etc.) that might be a good approach. Still then you Need to know if These Performance Drivers produce a stable Distribution, which they do not, they cluster (like volatility) with different market Regimes.
Which I tried with Portfolio weight (which is not kelly betting, yes) and the results where not convincing…
So then you need to time Different market Regimes (possible), but it gets complicated and if your assumptions just of by a bit….
e.g. back to the roots and bet Always a percentage of your port 
I just do not believe in stable enough distributions so you can use kelly…
Regards
Andreas
Edit: I am not sure what Andreas means by “stable Distribution.” So my post probably was not responsive to his idea or his concerns about Kelly betting.
-Jim
I totally agree.
Like all “academic” frameworks (e.g., mvo under MPT), Kelly can (and is prone to) be abused.
And because the relationships between variables and returns are non-stationary, I am attracted to the idea of a Bayesian ranking system. But I imagine that varying weights could miss out on the big of style drift reversions, which I feel make up a significant portion of many Ports’ expected returns.
Yes, therfore I try to implement a ranking System that combines factors that “Always work” as a good as possible.
The combination of nano caps, value, Momentum and Quality for example gives based on my Research the best average possible equity curve possible in all market Regimes.
And I found out this year that I have to exclude financialls because they really drag on Performance in certain market Regimes and because I concentrate on nano and Micro caps they had a too big junk of my port because there are “too much” regional Banks (they where great in 2016, but average in 2017 and not so good in 2018).
So where I might by variable in the future is on the industries I exclude in the future but let the betting size stable.