That thread gives clarifications:
2) Statistics now count every Pokemon in the team, not every Pokemon used.
2) Calculate the cut-off point C = S x (1 - (0.5)^(1 / T)) / 6.
And the definition used is
Number of pokemon that appear among T teams at least more often than not in Standard
So in short, let's say you take 1000 samples of 20 random teams, a pokemon would be OU if it's at least in 501 of those samples. That's the current definition. But it doesn't represent a 1/20 ratio at all. Because pokemon used in 1/29 teams are OU by that definition.
That also contradicts what he says later:
No, the list of Pokemon with T=20 would signify that each Pokemon in the list will be in at least one in 20 teams.
Because that's actually not the same thing.
From the first definition, T = 1 would mean 50% usage, from the second definition it would mean 100% usage.
I don't understand why using the first definition instead of the simple "A pokemon is OU if it is at least in one of out 20 teams" (which is the second definition). But at least the 3.41% number doesn't appear arbitrary anymore.
PS: Thanks for the link. It gives all the clarification about the current OU formula and you can disregard all what I said before between PO's and Shoddy's statistics differences, because that's outdated.