That scenario you hypothesize is a cracker. Exactly the sort of scenario planning that needs to be done as a discretionary trader.
Regards the pattern recognition failure or fulfillment, in my eyes now the only way to trade any pattern is through a system. Unfortunately for me that means having a full data set by taking every trade that my programme detects and then enters. This is the only true way to know precise failure rates. To the point I can see I'm about to lose money and I have to let it happen because I know I can't reasonably outperform my systems and will never intervene with live data (exception happens if I made a coding/sizing error and pick it up in real time). In return, I get to sleep but wake up to my new account good or bad balance haha. I have coffee first, then a walk if I need to walk it off.
To show you what I mean, here is a live bot that trades a doji as the trigger on the NDQ. Who would of thought, just can't know until it's measured. Now as a ex scalper/swing, I see a double top on the Nasduck and my system is overdue for a DD.........but I can't touch it, if anything I have to add because of it's performance. FYI this system has current WR of 49%, DD of 16% and annualised performance around 110%.
It pays to measure everything. Risk, correlation, time, entries, exits, patterns..everything. Then if one has spare time, find something else to measure.
(20min delay)