Originally posted by Gaggio23
This whole article feels really unrealistic to me. First of all, data analysis is mentioned but never mentioning from what type of event is taken (bo1, bo3, play, ranked, event??). If you look at untapped data, which is partial but still it's data that we can access and should be a realistic sample, the meta share is really polarized towards ajani/bombardment decks, which even after the nerfs still feel unfair to play against.
When looking at format data for balance, we take all of the competitive modes into account (usually ranked and events, and always looking at both Bo1 and Bo3). We don't use data from the Play queues for balance, because the deck-based matchmaking there skews results.
Third-party data (like Untapped) is only a small slice of the playerbase, and it's also non-representative. The more competitive/hardcore/experienced players are more likely than the casual players to use a tool like this, and they're also more likely to play the meta decks, so it provides a distorted picture.
To answer a question from a parallel comment, when we're looking at deck popularity & winrates, the system uses a flexible clustering algorithm, so it pulls in decks that are similar, not just identical.