Yeah, no, that's just about lousy definitions. The premise suggests that it's easy to become 95%-ile in anything, say, deadlifting among men who deadlift and not just entire population of the planet Earth. Before I can suggest author to try, he hastily corrects himself, that it doesn't mean "95%-ile among people who practice regularly", so "men who deadlift" is invalid domain after all. Rather, it should be... "peopl…
Gaming ladders with automated ELO systems are also not run like tournament ELO systems in more controlled environments like competitive chess. Games like Overwatch, WoW, Starcraft, Hearthstone, etc. will not dynamically update when the player is inactive in order to not discourage players. You have to lose games to lose your ELO category rank, although obviously you can go up and down the leaderboard without playing games. There are some gaming rank systems like CSGO that have some rank decay for inactivity but it is just at a flat rate.
So, "95th percentile" in these systems is not actually indicative of that unless it is based on the precise rank on the leaderboard. In many of these seasonal games rank inflation happens towards the end of the season because of this lack of passive decay. It means that at the beginning of the season, the top rank is very competitive, but as people become less active as the season goes on, it becomes easier to attain rank.
In a lot of domains, if you are 95th percentile, you are extremely competitive. You may be noncompetitive with the 99%ile cohort and the bottom of the 99%ile may be noncompetitive with the 0.1%ile, but to argue that that's not "very good" is just goofy perfectionism. The fact that the author is using goofy noncontrolled systems as an example to support his weakly-defined subjective argument only undermines his own case.