[rz2k]

I'm taking a Coursera class on discrete optimization where the grading feedback depends on your implementation of canonical problems. Optimizing your problems with better heuristics get you a better score.

Simply matching up like tiles in 2048 gets you further than only randomly moving. Then trying to put the largest sums in sorted order along one edge gets you further, but very rarely results in winning. Then by developing strategies about when to move the mass of tiles toward a single tile and when to move a single tile toward a mass, or setting up paths that are longer than the 4 tile length edge, I can win about half the time (or maybe a quarter).

Complexity that arises from simple rules is may he easier to program than the complex behavior directly, but I think you're on to something about those games being popular with people who program.

(Also, since I'm suspicious about the placement of stories about games going viral, I kind of enjoy the conceit of a narrative that the virality of these games is less organic. Is it really because they are derivative, or because trusted sources are saying they are knockoffs rather than, "You should try this!")