A sample size of anything over ~70 is fine for getting a somewhat decent statistic.>>175721
Tournaments in general aren't a good balance yardstick. First of all, what is good in a best of 1 isn't good in a best of 3, which in turn isn't good in a best of 5 or best of 7. So if for example you have a race heavy on gimmicks that can easily take one or two maps then you're going to see them over-populated in lower brackets but eliminated quickly once you start getting to Bo3-5 (which is what happened with the Dreamhack numbers btw)
Second, Terran has several years of statistics showing it has abysmal win rates when the players don't know who they're facing next. Terran wins lots in GSL where brackets are known weeks in advance, but rarely wins Olimoleague where they don't know who they're playing next until 20 minutes or so in advance.
And then there's the simple fact that the actual population of tournaments are very familiar with each other, which leads to several levels if mind games that are totally independent of race or balance. For example sOs generally dabs on Maru even at stages of the game that are generally considered Terran favored, simply because they play against each other so much that sOs just has a good read on Maru.
Masters to low GM is honestly a better indicator of balance than the tournament level. In fact, Masters to low GM often has a better meta than the tournament level too, most of the "amazing" new strats you see pros bust out are just refined versions of what low masters or even Diamond players do. For example the BC openers for Terran, the Maxpax build, or even the way Stats has been playing PvZ lately.