I created it. It's derived from a sigmoidal curve function. I dabbled in them when creating my own neural networks long ago.
Defense in the game will be soft-capped around 200 and ~90% damage reduction. I didn't want to hard cap damage reduction like The Elder Scrolls. If a player finds a way to squeeze out more defense then it will continually be effective but never reach 100%. The 0.2 controls how steep it climbs, making it more effective early on and when it drops off.
I don't know more common alternatives. I just used what I know. I doubt it's a good function, I tend to hack stuff together. It had the shape I wanted in Desmos.