[dmoy]

Competitive games is also a weird domain for measuring something like this. Optimal play can be both difficult to achieve, but have an incredible advantage. A 95th percentile player will lose to a 99th percentile player like 10 out of 10 times.

I mean shit, different blizzard game, but I was on a team that was somewhere around 99.9-99.99%, and if we went up against actual pro gamers (99.999%+), we'd lose 10 out of 10 times, and half the time we'd be unaware of why we lost (analogous to dan luu's "basic mistake", from the perspective of optimal play).

But on the flip side, we could literally play 2v3 and win 9.5 times out of 10 vs 97%-ile players. The advantage you get from seemingly small things in the game get amplified. For example if you just intuitively know when every ability comes off cooldown and becomes available again (both your own team and your opponents'), you can do things that seem otherwise impossible.

I don't think this translates all that well into a lot of other domains where it isn't a gamified hyper competitive / meta optimized scenario.