Personally, I think aimbot detection is a waste of time for teeworlds (except maybe for learning purposes, then it's a great idea), as botters will find a way to bypass whatever system is put in place (even if the system is closed-source).
But still, here's an idea : use the same system as spam detection (bayesian filtering). The only thing needed is a modified server, which gathers stats about players (movement speed, hook usage, accuracy, kill/death ratio, kill streaks (à la Doube/Multi kill in Unreal Tournament), and maybe others), and assigns a probability that the player is an aimbot. There would need to be a learning period when admins have to manually flag the bots/humans to detect bot and non-bot characteristics.
When there is sufficient data, the server can automatically ban players above a certain threshold. The learning database could then be shared between servers.
enable kicking @ HuH-Server
A note on that : there are many false positives. It's very annoying to get kicked from your own server because people don't know how to recognize aimbots. That's why it was disabled in the first place, but then the aimbot numbers grew.
The aimbot paranoia is way more annoying than the aimbots themselves (at least to me).