What's the best material to study to improve? UPDATED

I saw an interesting discussion on Twitter today regarding what level of study material is best for player improvement. It started with this tweet from Richard James of the Chess Improver blog:

https://t.co/5bEIQTi4YA My latest Chess Improver blog post: 9 lessons from a 9 move game between two kids.

— Richard James (@chesstutor) May 8, 2016

The post is worth reading as a jumping off point for the discussion. What followed on Twitter was a discussion of the topic between Richard James, Jonathan Bryant, and Matt Fletcher.

(Click image to expand view.)

Actually there's even more to the discussion, but the current version of Twitter makes it difficult to get all of a discussion captured in a few screen captures, unfortunately.

More to the point, does anyone else have an opinion on this? I have some of my own, but I'm still mulling them over.

UPDATE: Connor Eickelman writes:

The players above never mention what about their play they are trying to improve. What exactly is the goal of their training? Are they seeking a perfect panacea, able to cure every type of chess flaw? If so, I doubt they will get too much from low-rated games, unless said games are to prove a specific point (how to play against the IQP, for example).

If their goal is general improvement, they would be better off choosing a master from the past, preferably the opposite of your playing style, and go through a book of his best games, annotated by a good writer (500 Best Games of Chess is a good example, but that specific book is not limited to one player.) Then, go through the games, covering the next move with a note-card, and try to guess the next move. Act as though you are playing the game. If you want to be serious about it, get your clock out and set it for a long time control.

If they are trying to improve one specific component of their play, they should study that specific point in their play. If they are weak in theoretical endings, I fail to see how a book about class players slugging it out will help them in that regard. Even fifty games decided by a hanging piece or a two mover will not teach you the Lucena or Philidor! Just my $0.02.

Jonathan Bryant's rating would be a little under 2000, Matt Fletcher's rating would be in the 2000s, and Richard James rating would be in the 1900s. James also does a lot of tutoring and has written some books. I would hazard that none of them are particularly concerned, for a variety of reasons, about how to improve their own games. 

The question is, "How can weaker players improve efficiently?" Giving a 400 ELO player a book on Rubinstein probably won't help them much. Giving them one of Igor Stohl's books on modern chess masterpieces certainly wouldn't. However, going over games of players slightly stronger may help a lot. This is the topic at hand.