An analysis of the airfare prediction app Hopper

I've used Hopper for a while, including booking tickets through the app when the price was advantageous.

A memorable failure was when I was trying to book travel to Charleston, SC for the eclipse last year. Ticket prices were very high, and Hopper kept telling me to hold out, they would soon drop. They never did, of course--and I didn't really expect them to.

It was a good illustration of the shortcomings of machine learning. If a situation is not in the training data, the system is clueless. Hopper had no idea there was a rare external factor distorting the market, and apparently there was no way to tell it.

Cheap fare hacks are a game of walls and ladders. If a hack is really good, it spreads wide and fast, and airlines block it.