all i have is empirical testing suggesting 99% of the time for 99% of strategies it is unwise to bet the farm no matter how "reliable" the signal. so whatever you do, do it small and take your time. count on the improbable. count on getting stung by the fat tail.
have no clue how to formalize beyond that. not even sure how to go about it, because accurately characterizing risk is just the one-minus of accurately characterizing the odds of success. and we all know how tricky that can be.
Market Data 
Markets 
AI
