The best opening for white against King's Indian.

This is the best response to 1...Nf6 (I think), because it holds a firm grip over e5, restricting black's chances at the e5 pawn break:

Scorpion, I disagree. This is a fairly insipid line. Having said that, Smyslov played it a few times.

I don't know which line is objectively best, but I strongly suspect the Mar del Plata is the leading candidate.

Here's a fairly recent example:

The problem is that the white big pawn chain is too ambitious, and allowed black to play his thematic ...e5.  White also moves too many pawns.  Since it was played by Kramnik there's obviously still objective merits to the opening, but black was allowed his thematic ...e5 and ...f5 pushes.  12...c6 attacks the pawn.  After 20.Qb3 black has 20...Nxe6, not only winning the pawn, but controlling the d4 square too.  20...Kg7 just wastes a move, and allows white to exploit the weaknesses down the d-file with Rfd1!  c5!  pressuring the weak pawn and Rxd4! ensuring the lightsquared bishop stays on the board and getting the passed pawn on the sixth with a strong, lasting initiative.  27.Rxb7 was more accurate, as the knight, rook, and pawns control valuable squares while after Bb2 white is still better but lacks the bishop support and control over that diagonal after the pawn moves.  27.Rxb7,Qc8 28.Rc7,Qd8 29.Ne7,Nxe7 30.dxe7,Qxc7 31.exf8Q+,Rxf8 32.Bxf8 with an extra piece, or 27...Rb8 28.Rxa7,Ra8 29.b6 with two passed pawns with great support. 

Yes, Black was "allowed" the thematic pawn pushes. But he lost anyway. Tigran Petrosian believed that Black is objectively lost in the King's Indian, even with the thematic counterplay.

Petrosian played both the Mar del Plata and the Saemisch as white against the KID.

But Petrosian himself played a KID from the black side of the board.  I even have Fritz looking at a KID game Petrosian played (vs. lines with the big center with f3.)  It is Platz, Hans vs. Petrosian, Tigran V 0-1  E86 Budapest (17) 1952 

Although to be fair since the game was played in 1952 maybe he had the opinion later?  I think in any sound opening black draws with best play.  Heck, I'd even go far as to say draw with best play even in the Grob, though I certainly wouldn't play it.  He was also losing the game until white played f4 too late and started slipping by moving his pieces back. 

I think both players saw that the e6 pawn could be taken. Turn off your engine and take a look at the game. The gambit was sound. Don't trust your engine to determine whether a gambit is sound.

Regarding 27.Bb2, White may have had better, but 27.Bb2 was also good enough to win.

If you can't analyse it yourself, maybe let the engine run for much longer. I'm pretty sure the engines will agree that Kramnik was winning with 27.Bb2. And I'm pretty sure they will also agree that White had an advantage if Black played 20...Nxe6. Black needs to look earlier in the game for improvements.

Pellik, Yeah, there are definitely improvements. I don't think 13.Bg2!? is Kramnik's final answer in the KID, just a really cool line he discovered.

Theory in the King's Indian continues to develop rapidly. My point wasn't to argue the specific lines of the Bayonet, merely to show that the Mar del Plata offers white greater chances for an advantage than Scorpion's insipid line with Bf4 and e3.

White should respect Black's e5 and f5 plan, but not be paralysed in fear of it.

Sorry, the game was actually against Terpugov, Evgeny A43 Moscow 1951, my bad.  The other game is one I'm looking at in a Word document doing De Groot exercises (with typing thoughts on positions instead of recording out loud).  I'll post it when Fritz is done chewing it. 

I would have played e4 first off and forced them into a variation of Alekhine's Defense...problem solved...

It is still good enough to win but 20...Nxe6 is at least a good try.  People have lines they prefer over others and I can respect that, but one of white's main goals in the opening is to restrict black's plans. 

If white black takes the pawn after Rxb7 then 27.Rxb7,Bxd6?? 28.Bb2+,Nf6 29.Nxf6,Rxf6 30.Qf7 when black has to sacrifice material to delay checkmate.

Was an okay calculation exercise at least though I doubt black would have taken the pawn.  In one game there is just so much to look at; noting potential plans for both sides, weighing static vs. dynamic advantages, value judgments over what weaknesses are exploitable, and among exploitable which ones get the highest priority, how to favorably transition into the endgame, etc.  

Pellik, you have " a few GMs" playing that line?  And they play 21...Rad8 rather than Radjabov's 21...Kh7?  In blitz?  In some ways, that's very impressive. But I play a very similar line over there and have almost never seen anyone play that as Black. In blitz or otherwise.

I did run into that in correspondence chess though. I couldn't find anything decisive for White against best play by Black.

Here's an example of a successful blitz game against an FM. The stem game is Korchnoi-Hulak (1988, I think). There are of course major improvements for both sides. But before you say too much negative about Black's play, I reached the same position after 15.Bxa7 in a blitz game against a GM. Unfortunately, I was only able to draw that one despite the overwhelming advantage I had out of the opening.