Not exactly. I have a “low computation requiered” algorithm which points at good coins. Only when the coin is “good enough” (distance / value < 25), I use my heavy computation part where I check if this good coin isn’t making me going an overall wrong path. If it does, I rule this coin out.
I invite you to check the full code there, the line I’m refering too is marked with a // (3) in my code. That helped in pushing back the hard execution limit. However, my algorith is still far from being light.