The topics of concern tend to be due to things being ‘too strong’ or ‘too successful’, so the effect tends to be in concerns for stuff being too high, but the result can work at the other end if people are looking at stuff being 'too weak" (though in many cases, players are content for some things to remain lower)
Part of the visible issue of parsing sites is that they do skew the numbers, esp when filters are applied. Their estimation systems are prone to inflation because exposure across its spectrum is assumed, and the exposure to the playerbase tends to be wide and shallow, with select early/intentional uploads setting a benchmark of adjustment to the system.
What Reinhardt posted regarding stats is particular to populace opinion or clinic trials, so the analysis, hypothesis, skew, study and so on for the impact of the stats are different compared to game compilations. Part of the biggest issue of statistical analysis is that people tend want to apply them to a broader range or conclusion than the stats really apply.
In that regard, the stat tends to be less about the ‘analysis’ and more the conclusion at hand. So, to some degree, concerns of high/low depend on the intent of the study and the impact the ‘skew’ could have, esp in regards to standard deviation.
In regards to the topic itself, yes, its likely that any hero that is excessively overpopular will see something toned downward to either discourage the popularity, create a placebo that they have been “nerfed”, or act as a filler change to appease outcry till the meta changes regardless of that hero.
Some heroes do get ‘temporary’ nerfs, and then get buffed back to, or even above, their release levels. In that regard, from a numbers perspective the concern of being “over powered” could be offset by the reality that numbers were later increased (“buffed”) but the previous problem and outcry did not occur.
So in that regard, the labeled concern for “over powered” would be less in regards to the stats “proving” the concern is objective, but rather one of perspective and adaptability.
Blizz tends to consider some stats these sites don’t filter very well (such as premade games or exact hero level) and use that to influence if the outcry to things being “overpowered” are based on values, designs, or gameplay trends.
While they aren’t ‘perfect’ or whatever for the game, they are the ones making the changes, so some things influence their choice and actions more than others; the capacity for another site to post numbers that defer from their own may only influence them to do superficial changes per the placebo, or similarly, they may look to offset the “frustration”, but keep a similar power level to the hero (to keep the feel/balance the same) and do a side-grade adjustment instead.
Deathwing was going to be popular regardless; he’s been a high demand hero and a heavily requested “game breaking” addition that has has needed for awhile, while then released at a vacation period were lots of new players come back to the game to experience the hero… and potentially make divergent results that end up as noise to third party parsing that can’t filter it out.
So yea, per popularity and outcry, changes will likely continue to be made; the previous dev notes on his last balance hinted as such to be expected. However, some of the extent by which people argue the case to be more pronounced than may be the case could find themselves disappointed when said tuning comes.
What [they] use to form their opinion may not be the same material that blizz uses to conclude what sort of change they’ll make to the numbers/design/etc. Some posters tend to get so caught up in arguing with others that they forget that Kara or Oj or Zenas or whomever aren’t the ones that actually change the game.