With hindsight, I realise that most of the improvements we measured were really just statistical noise. We never had a control plane, so if one by accident flew better one test, then we would tweak something and then we would naturally see regression to the mean, but we would record that as a tweak that made things worse.
To little me it was a lesson in deliberate design. To grown-up me I guess it's a lesson in how easy it is to delude yourself without proper experiment design and statistics.
Anyway, then after seeing my interest my father brought me an early Windows program that contained animated instructions on some paper plane designs. One of them was vastly superior to anything I had ever seen. But with the years, I forgot how to fold it. I had been looking for it for over 20 years when a few months ago, I wanted to show my son (of two years) a good paper plane and guess what? I think I've found the design! But it took a blurry YouTube video and some trial and error. I ought to write it down some day.