Once something has passed WF analyisis, Correlation, Montecarlo and incubation it can be live tested but small. This is to check for execution errors ie. slippage, bugs, out of hours spreads vs back test results. If the results are a direct match to live then I deem it safe enough to try out.
So far I grade them based on Return/DD ratio over 1 year and factor in Frequency (THIS is very important) to find out if they are A-grade or B-grade in regards to productivity and the equity curve you go on to get there. Risking 1 or 2 R twice a month is not the same as risking it everyday on your most vicious algorithm!
Passing OOS by incubating for 3months, then live with 1or 2 contracts for 1-3months seems to be the only data that really matters.
If all that is fine then it really comes down to how it might fit into my portfolio which is really about correlation when you think about it. Most importantly, can I live with the equity curve? Most of this you won't really know at the start until you live test 30-50 for 6 months or so.
I tend to build bots and find edges in things I wasn't good at like holding powerful breakouts and trend systems. The reason these are so useful is not only can you have a system take and execute to play out it's edge over time but they can train you see these things setting up. Then you already know your odds and learn to do things your brain previously didn't let you!
(20min delay)