What makes any game fun to watch and follow? I'd argue the central factor is the competitive balance struck between skill and chance. A balanced game is one in which the players and teams with the most skill rise to the top, but one where there's enough uncertainty thrown in that there's room for surprises and upsets. In any given game, there is a favorite and an underdog, but there's enough suspense about the result that neutral viewers tune in anyway. In traditional sports, fans love dynasties, and they love watching dynasties fall. An ideal competitive scene gives you both.
But there are additional factors that are unique to esports, where the rules and nature of the game frequently change. These have bearing on the competitive balance, but are also independent of them. For instance, I'd suggest that the ideal Dota 2 game from a viewing perspective takes 30-40 minutes, similar to the viewing time of a standard episode of television or a half of soccer. I'd suggest that Dota 2 is best when there are a wide variety of strategies that can win; pushing strats, ganking strats, split pushing strats, lategame 4 protect 1 strats, and so on. I'd argue that average Dota2 game should be reasonably close for 75% of the time, and once a team takes a strong lead, it shouldn't belabor the point, unless the team at a deficit stages a comeback. It should be hard, but not impossible to do this.
I've watched every edition of The International, and two tournaments in particular strike me as being poor: TI4 and TI6. In both tournaments, the meta was out of whack. I remember TI4 being characterized by unstoppable pushing strategies, and snowballing leads. I remember TI6 as being a tournament of "rubber-banding" games, where no lead was safe to a high ground reversal, and "epic" comebacks became so commonplace that they stopped being interesting or impressive. While TI4's overall results were still pretty "fair," with the best teams making it to the finals, TI6's results were goofy, with two qualifier teams making the finals, and half of the invites missing out on the winners bracket. It's no surprise that TI7 features just one of the players who played in the TI6 finals, and that player is only in Seattle as a standin (and also is, for my money, the best of those ten players).
That's why this TI7 is so refreshing. The favorites have looked like the favorites, and five of six made the winners bracket. But there's still room for surprises, and the groupstages featured plenty of examples of top teams dropping games to underdogs. Moreover, we saw every type of game in the groupstage. There were massive stomps and prolonged defensive battles. We saw some comebacks, and one really extraordinary reversal that I saw (TNC vs LGD, game two). High ground is a challenge to break, but not impossible. Laning advantage is important, but not game-deciding.
What's best I think is that we've seen a lot of room for the personalities of different teams to emerge. Some teams play with sacrificial, team oriented #1 positions, and others play with dominant classical farming #1s. Some rely on dual cores with a powerful farming mid laner, while others use a mid laner in the classical tempo setting and playmaking role. There is tremendous flexibility in how teams lane and how supports position themselves to get an early advantage. Almost every hero has been picked, and many can be played in multiple positions. The current meta is, to my eyes, an almost perfect blend which allows individual player skill and team strategy to really shine, and it's made watching this TI an unexpected treat.