Blizzard has a lot of data to back up their statements.
Hearthstone Game Designer Max McCall recently took some time to talk about the strength of the Shaman class. In his post, he describes the current situation where the class as a whole is more popular than they'd like, but believe it has proper counters in today's meta and will hopefully be brought back into check. Its win rate isn't astronomically high and McCall says archetypes like Renolock have been partially to thank for this.
He's come under fire, however, as some forum users are now saying that Vicious Syndicate's data proves that Aggro Shaman is only slightly unfavored against the decks McCall suggested. As such, he's taken the time to retort, explaining how that isn't enough data to understand the meta as a whole especially when it's biased towards players who are higher ranked since they're more likely to use deck tracking software.
See his full post below.
I chose Reno Warlock for my example because it performs substantially better than Dragon Priest against the midrange Shaman variants, and those midrange variants outperform aggro Shaman against the field and are favored against aggro Shaman specifically to boot.
Perhaps I should have called out Dragon Priest as a Shaman counter instead of Reno Warlock, since Dragon Priest is slightly stronger than Reno Warlock against specifically aggro Shaman, and aggro Shaman is the most popular variant right now. Regardless, Reno Warlock is a marginal favorite - by half a point, but still a favorite - against aggro Shaman. But against slower Shaman decks, you'd rather be playing Reno Warlock than Dragon Priest.
Vicious Syndicate does good work, but they obviously don't have as much data as we do. We look at every game played among the top 1.3% of players, in all regions.* That roughly corresponds to rank 4 and up. We know every card in every deck and sort those decks into the archetypes you see on the ladder. Our BI team is extremely sophisticated, and we can drill extremely deep into small variants of those archetypes and see how small deck tweaks impact certain matchups.
Many conclusions can be drawn from the same data, but we are extremely confident in our assertions about specific matchups when the data is aggregated in this way. I do note, however, that at nosebleed Legend, some individuals do substantially overperform. This was most obvious when Patron was very strong. Skilled Patron players at or near rank 1 Legend were winning almost 80% of their games after a couple hundred games, while the mere mortals at medium Legend ranks only won about half their Patron games. So, yes, some players win a lot in matchups that are ostensibly unfavorable - but in general, when we describe a matchup, we're talking about the high-level metagame at large.
*Yes, we have looked at lower ranks as well. The numbers were virtually the same. It's computationally expensive to look at every game at every level, even with sampling, but we do recheck every so often to ensure that the high level metagame is still representative of lower level play.