[always_good]

Well, do you think it's worthwhile to implement your own collision map system, vector math, tile editor, etc. every time you start a new game?

It sure is tempting, but not necessarily healthy to the odds that you will ship a game. Though it's also not obvious how much a game engine will help you (like if you'll run into issues with your specific mechanics) until you credentialize in it, so it's always a bit of a crap shoot because games are so open-ended.

For example, I once migrated away from PhaserJS because I couldn't figure out how to add networking and server-side simulation without having more of the physics code in my git diffs.