Well, there's a lot of reasons, ranging from "He's not Bob Geren" to "he's intelligent and he communicates well," but today I want to focus on a specific skill. A manager can manage "to win" or manage "not to lose" and Melvin has shown a shrewd ability to assess which to do at different times.
Some managers refuse to take chances and eschew the "conventional wisdom". They are afraid of having to answer difficult questions from the media after the game and so they take the safest road even when it leads to the intersection of Dullsville and Lose. Other managers -- Joe Maddon comes to mind -- seem almost obsessed with defying "the book" and will give Sucky McWeak a 3-0 green light, put 6 infielders on the left side, put LOOGY McSouthpaw in right field for a hitter, and will call for the "delayed fake-leg-before-wicket safety squeeze," seemingly outmanaging themselves as often as they outmanage their opponents. (Note: I like Maddon and think he's a very good manager. I just also notice him "overmanaging" as much as anyone.)
And then you have Melvin, a manager who often follows the script because the script is often tried and true, but who isn't afraid, every once in a while, to defy the book in favor of a decision that just makes sense at the time.
In New York, Melvin was managing to win, not afraid to lose, when he ordered the intentional walk to Robinson Cano with 2 outs in the 9th and Cano representing the winning run. This was managing to win because Melvin knew that if Grant Balfour could get out the next hitter he faced, the A's would win. Never mind that a double would now lose the game instead of tying it -- who can Balfour most likely get out? That wasn't Cano, the Yankees' best hitter, a LH batter and one hitting in a ballpark with a short porch in RF. It was Vernon Wells, having an excellent season but still the better matchup for Balfour.
Melvin factored a ton of things in at once here. The LH/RH consideration, the short porch in RF, and the fact that a tie (e.g., a Cano single) would position the A's to face Mariano Rivera in the 10th with Oakland's best relievers already spent. Don't play to avoid losing here: Play to win. Melvin did, and the A's did.
Now fast forward to last night, when Melvin shrewdly managed "not to lose". Going to the 9th trailing 5-3, Melvin knew one thing for sure: If the A's don't score at least 2 runs in the 9th, they lose. So he maximized the A's chances to score by pinch-hitting John Jaso for Adam Rosales to start the 9th, even though he did not have a backup middle infielder available.
Boy did it pay off. Jaso walked, the A's rallied for 3 runs, and if Oakland had a bizarre infield alignment (Moss at 3B, Donaldson at SS, Sogard at 2B, Jaso at 1B) in the bottom of the 9th, protecting a one-run lead, well: At least there was a bottom of the 9th to protect. Just don't lose it in 8.5 IP. Melvin realized this was not a time to manage for anything but "not to lose (in 8.5)". He, and the A's, managed not to lose.
That's two very unconventional moves in 49 games from a manager who otherwise often rolls along using his "plus" relievers in "their roles" and maximizing platoons "by the book". He is neither too gung-ho to push the envelope, nor is he afraid to push the envelope. And perhaps most importantly of all, he knows when to play "to win" and he knows when to play "not to lose". And the A's might be 2 wins better off because of it.