I think there is a balance, early game should not be the only thing that matters, which is were we are right now in HS. If you don’t draw well within the first 4 cards, you simply lose.
Historically, the importance of early game is because of the broken nature of the game if it goes on long enough. Standard was typically exempt from this brokeness in the past because of a greater effort by the team to balance it. But with the amount of discover & random effects in the game, every extra turn you play (for example as say control warrior) is another turn that your opponent has to high-roll some incredibly game-winning effect that just destroys you.
Mathematically, even if that probability is small, it increases exponentially with number of turns. For example, say the probability that your opponent discovers a game-winning random effect is p per turn. Assuming each turn is independent in terms of rolling a game-winning effect (fair assumption), then the probability of losing thanks to this effect can be modelled as
1-(1-p)^n
where n is the number of turns played. Here is a simple table showing how fast this probability increases with n, for different values of p.
p=.01
n | 1 - 0.99^n
1 | 0.01
2 | 0.0199
3 | 0.029701
4 | 0.039404
5 | 0.04901
6 | 0.0585199
7 | 0.0679347
8 | 0.0772553
9 | 0.0864828
10 | 0.0956179
p=.05
n | 1 - 0.95^n
1 | 0.05
2 | 0.0975
3 | 0.142625
4 | 0.185494
5 | 0.226219
6 | 0.264908
7 | 0.301663
8 | 0.33658
9 | 0.369751
10 | 0.401263
So you can see that this is VERY sensitive to the value of p. If its 5% per turn there is nearly a 40% chance by winning a game via this effect by T10, and at 1% its much more modest <10%.
Personally, I think the real value of p is somewhere between 1% and 5% per turn, and its what really holds back control decks and longer games, and why everyone feeds into the aggro, go-face paradigm, as its a much more certain way to win. Hence the importance of early game over everything else in HS.