I disagree. If you have the data, try throwing ML at it. It's probably less work than trying to "understand it" and building a heuristic. If you don't have the

I disagree. If you have the data, try throwing ML at it. It's probably less work than trying to "understand it" and building a heuristic. If you don't have the data, how are you going to validate your heuristic anyway?