I have bordered on your approach once. Were a few bots doing my head in so I tried to outperform them for 6 weeks. For 3 weeks I was a genius, up a few extra K's then for the next three I cut meaningful trades short or scratched the trade. Logged everything and it was a round trip for me. Lesson was, well two actually was to not mess with statistical samples. Second lesson was if that's what I'm thinking to do I shouldn't be running them in the first place.
Only exception I found was that ftse Hammer I have that when it fails a second time it marks a great V and I've done well trading against it. Now I have launched Operation Conviction where I have a suite of my favourite trades and add size manually, trade against their weaknesses and generally use them to tell me what my bias is and where to be looking. Bit like you have I guess.
Seems to go: oversold awaiting confirmation/impulse candle. Here I am allowed to scalp or take 5 min TF trades but not allowed to leave the screen. I soon tire of that and when the breakouts or trends fire I look to add more at same or better prices. Once trends are de-risked I need to go find something else to do for a week or two before getting interested again. If on-trend but low vol and markets meandering I'm allowed to trade overnight levels. Then look for topping structures and trade the process all over again. Rinse repeat.
It's brilliant really. kept me out of many many opinions I might otherwise have. Yet to combine a long short into one though, I see lots of traders looking to do that. For me I just check correlation and run them as pieces so I can adjust size if needed. Like LEGO, really fun LEGO.
(20min delay)