i dont get what that analogy has to do with tournament results
Because by the same statistical logic, you should take up smoking to avoid cardiovascular disease.
It's not an analogy, it's reductio ad absurdum, which means applying the exact same logic to circumstances to another issue which is comparable in terms of the input (tracking smoking status for a relationship with cardiovascular disease vs. checking character choice for a relationship with tournament results).
Since in both cases it is modeled in terms of raw numbers, the results are that there is a correlation between being a non-smoker and having cardiovascular disease. Using the same statistical reasoning there is a correlation between using a particular character and getting certain results.
I guess that's fine, until we look a little more closely and see that there are significantly fewer smokers then non-smokers, furthermore, a much higher percentage of smokers get the disease then non-smokers.
So, there's a statistically significant relationship with being a smoker and getting cardiovascular disease, yet using raw numbers showed the opposite, why?
Because raw numbers are irrelevant, statistics can only show the difference between two populations in terms of a factor when the populations are taken into account, in other words a ratio measure.
That means that tournament results tracks CURRENT centralization (relevant spots taken in tournaments), not power in the metagame.