Tested my new wifi autochannel algorithm on 2.4 GHz in a noisy office. Most of the routers in the room were on channels 1, 6, or 11, but there were at least a couple of outliers on every single channel (this is New York after all).
Anyway, my algorithm picked channel 2. Conventional wisdom is that this is exactly wrong, because it is a partial overlap with channel 1. Sadly, I double checked the math, and it checks out. It may have actually made the right decision for a very complex situation, assuming my algorithm makes sense at all.
I'm pretty sure this is how it will feel when the machines take over.
Meanwhile, I guess I should just do a speedtest on every single channel and do this objectively.