>>528746638
As >>528748769 says its half tests after and half tests before for me - if I write them at all.
Depends on what you're doing, if the game is systems heavy having a robust suite of tests allows you to be 100% sure that everything works even if you fiddle with it. For example I'm making a game where a lot of units navigate and avoid each other, so I made some utility functions for computing free paths and such. These are separate from the engine and thus easily testable, just a function with input and output. Once I had some basic working prototype I moved to tests. I started with the cases I knew about and I as I was writing the tests I noticed various edge cases for which I added tests and if they were failing I fixed the code. This was way easier that booting the game and trying to get to each edge case at runtime. On the other hand it rarely makes sense to write tests for highly abstract game mechanics or such.
Also this is a good point >>528749014