If you compare ESO balancing to a game where they actually do things properly (Rainbow Six Siege), you'll notice something: the developers actually back up their gameplay changes with data and metrics collected in game and of course, they actually bother to collect that data to begin with.
Example in R6: Oh, an operator in rainbow six siege has a 75% win rate? Let's look into why they're winning so much and adjust accordingly until we get a reasonable stat on their win percentage
Example in ESO: Oh, it seems there's 3 or 4 forum posts about X class or X ability being OP (most likely posted by the opposing class which is counters), lets destroy that class/ability so people stop complaining.
Literally the easiest metric to collect in this game seems completely ignored: Class Distribution. IE, Almost every end-game trials group DPS is a NB, healer is a Templar and tank is a DK. Hmmm... maybe, just maybe, there's something about those three classes that make them vastly overperform in their roles? Or are ZOS just waiting for a couple of forum posts to butcher those classes in those roles?
The proportion of people playing Sorc/NB compared to Templar/DK/Warden in PvP is ridiculously skewed. I would honestly say there are 100 NB/Sorcs for every 15 Templar/DK/Wardens. Hmmm.. maybe there's something weak about Templars, DKs and Wardens? Maybe?
Collect some data on your game.
While you're at it, ring up AWS and get some proper infrastructure so your game doesn't crash every 15 minutes in Cyrodiil.
"Rock is too OP, please nerf"
Sincerely, Scissors.