> My heuristic is to book a little over two weeks in advance, but not to book much more in advance of that in case plans change or want to change, since in expectation price changes are pretty small and maybe you decide to stay an extra day for some reason even if you are confident you won’t cancel.

N00b here - is that actually where the optimal prices emerge, rather than many months prior? I used to wonder why the optimal prices didn't emerge the day before the flight, because I'd assume airlines would rather fill a seat with lower profits than waste the space, but I guess that could cause forecasting issues if everyone waited until the last second.

Lot of last minute(day) demand is very price insensitive. Say a company needs to get engineer on certain site. Or in general business travel where someone else is paying. In such cases you can get a few times the lowest price. So not fire selling seats makes sense.

On other hand if someone like travel agency has pre-paid for seats in block, selling them at any price will lower the losses. So different actors have ways to act differently.

Optimal prices don’t appear just before the flight because they probably overbooked it

My understanding is that (at least for long haul), seats are divided up with a bunch of lower cost tickets, then medium cost, then a few high cost tickets, and sold in that order. So waiting until the last 2 weeks would be a bad idea. But also perhaps my understanding is wrong / naive?

It depends but often cheapest prices IME are;

1) as the flight is released (e.g a year in advance or more)

or

2) about 5 weeks out.