I quite like this idea.
But using stops and targets based on market activity, although sound, will muddy the stats results, no? If he wants to go long for example at 4840 which he thinks is support, but the coin tells him to short, then he might want to play it very tight, and place the stop right at 4842, just 2 points higher, cause he "knows" its gonna stop out. But if he does in fact get a long from the coin "signal", then he might stick with his standard 5 point stop, hence the stop will be at 4835, 5 points below the long, because he will believe in the long, even if support is broken. Using what the market tells you first assumes you can read the market properly, but then it also assumes that the market will follow what you think each time you make this prediction. There is an inherent randomness in all of this, and this is what I assume you're getting at with this experiment.
I would imagine what would be better is to use the same stop and target for each trade, and if it so happens that the coin tells you to go in the opposite direction of what you want, just let it play out. Hopefully you're better at guessing direction better than the coin, but this certainly won't be 80% of the time I don't imagine. Sometimes, maybe even 30-40% of the time, I bet going in the opposite direction of what you think will be the winning trade.
Of course I have no idea where you are going with this, and so perhaps what you're trying to illustrate is different than what I have in mind, in which case, forget what I just said.