Torpid wrote:Stats on win rate and popularity massively over simplify the issue of balance. They're not useful in any non-negligible way.
I don't think I'm understanding you correctly. It is your assertion that there is either no, or negligible correlation between whether a race wins and whether or not it is balanced? If this is the case, then I understand why you asked for a definition of whether or not a race is overpowered or not, because if the ability of one race to beat or win against another race has no relevance to its overpoweredness, then I've got nothing. I was assuming from the plain language of the terms that a race's overpoweredness was directly related to its ability to overpower another race. If on the other hand you're simply saying that the particular stats currently collected do not accurately represent the community we wish to balance for, then see my response infra regarding stat collection.
Races with more players likely have more developed metas and perform better at super high level but worse at low-level where the low-levels players all flock to that race for whatever reason external to balance.
Then you have the issue of some heroes just being easier to play a near-optimum skill level but not that great at such a level. Whereas other heroes will really slaughter everything at a near-optimum skill level but are just far harder to use at such a skill level. If we aim for top-down balance this second problem basically makes the win ratio/popularity statistics pretty useless on their own.
Great, maybe we should collect statistics from the levels that we wish to balance for, like I suggested in the following post:
Eclipse wrote:the statistics are...Useful...Because: they accurately reflect the nature of the heroes, races, and maps, in the manner most relevant to the Elite community, whether based only on tournament matches in order to minimize people playing FC for the first time skewing the results; or the entire community since the latest patch to maximize sample size; or in an alternative manner deemed most relevant by the Elite community.
Another problem is that using stats from past games assumes that incumbent and previous metas
=were close to identifying the optimal performance of each of the heroes and builds they can accommodate. I see little reason why this would be the case.
This is yet another statistics issue, and the field present a veritable array of theories on how to deal with moving trends from simple moving averages to the GLICKO-1 system mentioned on the same page that has these stats in the first place.