I would hope they are smart enough to get a representative sample of players for the whole game. It would not be that hard.
If we assume true randomness, and we assume 100k players, accept a 5% margin of error, a 95% confidence level, and a 20% response rate, then about 2,000 truly randomly selected players need to be asked/included. Of those 2000, only about 400 need to actually respond or be included. For any population size over 20,000, the numbers do not change much.
If we break the population down, either by league or by level segments, the sample can be smaller.
This all assumes true randomness in the selection.
(Yes, I am aware computers cannot be truly random, but they can be close enough that it would be acceptable.)