The main characteristic of the Averbach is that White delays the development of the g1-knight for some time. This will allow White to control important squares such as g4 and h5 and take active measures on the kingside himself. White will often consider castling queenside in the Averbach.
———————————————————
The starting position is usually reached following 1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Be2 0-0 6.Bg5. As in most King’s Indians Black will need to make a choice early on how he will try to challenge White’s center and obtain sufficient counter play.

Advancing the e7-pawn to e5 is often a very poor idea when White plays an early Bg5. Folllowing 6…e5?? 7.dxe5 dxe5 8.Qxd8 Rxd8 9.Nd5, Black would lose at least a pawn due to the double threat against the pinned f6-knight and against the c7-pawn.

6…Be6? moves this bishop to a poor square where it may be attacked with d4-d5 soon. Having the bishop on e6 also makes it impossible to develop the queenside knight to c6 next and gets this bishop in the way of the e7-pawn in many lines. You should not move this piece until you know better where this bishop will be placed best.

6…Ne8? retreats this well-placed knight to a poor square on the back rank. You should try to develop a new piece or look for counter play in the center.

6…Nc6 is a fine way to develop the queenside knight and tries to pressure the dark squares d4 and e5 in the center. Black also hopes to provoke White to continue with 7.d5 next. The game Korchnoi-Kasparov (1988) then continued with 7…Ne5 8.Nf3 Nxf3+ 9.Bxf3 h6 10.Be3 e6 11.0-0 Re8 and an unclear position. Good, we are instead looking for a different post for this piece.

6…Nbd7 is a solid move that prepares to advance with …e7-e5 soon. This has done fine in practice for Black. After e.g., 7.Qd2, Black should continue with either 7…c6 8.Nf3 e5 9.0-0 or 7…e5 8.d5 Nc5 9.f3 a5. Good, we are instead looking for a different post for this piece.

6…Na6 is an idea borrowed from other King’s Indian systems. The main purpose of this move is usually to play …e7-e5 soon and gain access of the c5-square for this knight. This move has done quite well in practice. Also interesting is to develop the queenside knight with 6…Nbd7 or 6…Nc6 instead. 
After 6…Nbd7 7.Qd2, Black should continue with either 7…c6 8.Nf3 e5 9.0-0 or 7…e5 8.d5 Nc5 9.f3 a5. If Black instead moves this knight to the aggressive c6-square, play in the game Korchnoi-Kasparov (1988) then continued with 7.d5 Ne5 8.Nf3 Nxf3+ 9.Bxf3 h6 and an unclear position. 

7.h4 shows White’s aggressive plans on the kingside in this line. The more popular 7.Qd2 doesn’t cause Black many problems after 7…e5 8.d5 c6, while the sharp 7.f4 is also interesting and could lead to similar positions as we know from the 5.Bg5 challenge.
———————————————————
Although Black’s kingside is sufficiently protected at this time, Black has to play with considerable care.

7…h5 is a committing move to play, but is probably playable. After 8.Qd2 e5 9.d5 c6, White had a slight edge in Horvath-Loskutov (1996).

7…Be6? moves this bishop to a poor square where it may be attacked with d4-d5 soon. Having the bishop on e6 also moves it in the way of the e7-pawn in many lines. You should not move this piece until you know better where this bishop will be placed best.

7…Bd7?! moves this bishop to a mediocre square before it becomes clear what Black’s best plan is. Developing this piece should be delayed.

7…e5 challenges White’s center and is the logical follow up to playing …Na6 on the previous move. Black’s best alternative on this move is probably 7…h6, when after 8.Be3, play could later lead to similar positions as in our line. 
Instead 7…c5 8.d5 Nc7 9.Qd2 e6 10.h5 does well for White in practice. This is mainly because Black has less space, is not able to exchange any pieces and must also worry about his king.

8.d5 doesn’t allow Black to open up some lines in the center. Note that White no longer benefits from trading pawns on e5 and queens on d8. Following 8.dxe5 dxe5 9.Qxd8 Rxd8 10.Nd5 Rd6, Black’s a6-knight protects the key squares c7 and c5.
———————————————————
Now that the center is closed White hopes to spend all of his energy on generating a kingside attack.

8…h6 forces White to break the pin against Black’s f6-knight one way or other. One drawback of playing an early h2-h4 is that the g5-bishop cannot retreat to h4. Note that the immediate 8…Nc5? runs into 9.b4. Since Black’s f6-knight is still pinned to the queen in that line, Black could not win the e4-pawn.

9.Be3 retreats this good bishop to the most active square. The exchange 9.Bxf6? Bxf6 would give Black the bishop pair and control of the dark squares. In the King’s Indian, White hardly ever benefits from trading his dark-squared bishop for Black’s f6-knight.
———————————————————
Black now has the option of playing …g6-g5 should White advance with h4-h5 later on.

9…Nc5 attacks White’s e4-pawn and gets this horse to a more active place near the center. Black could also look to open some files on the queenside with 9…c6 instead. 
Stopping White’s kingside plans with 9…h5 has been a popular choice by many strong players. This is better than on the previous move since White’s dark-squared bishop is then on the less aggressive e3-post. Play could continue with 10.f3 c6 11.Nh3 cxd5.

10.Qc2 is White’s best way to protect the e4-pawn. Instead 10.f3? would lose control of the h5-square and allows 10…Nh5, while 10.Bf3 a5 11.g4 Bd7 12.g5 hxg5 was unclear in Zitin-Shulman (1995). 
Taking Black’s c5-knight with 10.Bxc5? dxc5 would be a poor idea since White’s dark-squared bishop is his best minor piece. Black could then reach a fine position with moves such as …Ne8-d6 and …f7-f5.
———————————————————
Black still needs to develop a clear plan.

Sacrificing a knight on e4 only gives Black two pawns for a piece following 11.Nxe4 Nxe4 12.Qxe4 Bf5 13.Qf3 e4 14.Qg3 Bxb2 15.Rd1. This will not give Black enough compensation.

10…c6 looks to open the c-file now that White’s queen is on this file and can become a target for Black’s rooks later on. This is the most dynamic move that also gives Black’s queen new options along the d8-a5 diagonal. Instead 10…a5 would stop White from playing b2-b4 next. But as he will soon see, this is not necessary at this time. 
Black could also try to win the bishop pair with 10…Ng4. But after 11.Bxg4 Bxg4 12.f3 Bc8 13.h5, White would have a small edge. If Black tries to strengthen his control of the g4-square with 10…h5, White has done well with 11.f3, e.g., 11…c6 12.b4 Na6 13.a3 or 11…a5 12.0-0-0 Nh7 13.g4.

But does White have enough pieces on that side of the board to justify such an ambitious plan? If White tries to push back Black’s c5-knight with 11.b4?, Black can play the surprising 11…Ncxe4! 
Following 12.Nxe4 Nxe4, White must avoid taking the e4-knight with 13.Qxe4?? since after 13…Bf5 14.Qf3 e4, both the rook on a1 and White’s queen would be attacked. Instead 13.dxc6 f5 14.cxb7 Bxb7 would give Black a promising position.
———————————————————
Finding harmony between one’s pieces is very important when play shifts from the opening into the middle game.

11…Nxh5? allows Black to give you a poor pawn structure on the kingside with 12.Bxh5 gxh5 13.Rxh5. You would then have to worry about the weak h6-pawn.

Keeping the kingside closed with 11…g5 is not as popular as our main line, but is a reasonable idea. Following 12.f3 a5 13.g4 Bd7 14.Nh3 a4, play was unclear in Bareev-Kasparov (1992). We are instead looking to open up the position.

Taking White’s h5-pawn with 11…gxh5 looks odd, but seems playable. After 12.Bxh5, Black should continue as in our line of play.

11…cxd5 opens up the position a little more. This will give Black much better chances to keep White busy in other parts of the board and take him away from his intended kingside attack. Keeping the kingside closed with 11…g5 is not as popular, but is also playable. Following 12.f3 a5 13.g4 Bd7 14.Nh3 a4, play was unclear in Bareev-Kasparov (1992).

12.cxd5 is White’s only good recapture. Clearly 12.exd5? Bf5 allows Black to take over important squares with gain of time.
———————————————————
An important decision must be made.

White is threatening to win a pawn by playing 13.hxg6 fxg6 14.Bxh6 next. If you want to sacrifice the h6-pawn, you need to find a much more active continuation.

12…Nxh5? allows Black to give you a poor pawn structure on the kingside with 13.Bxh5 gxh5 14.Rxh5. You would then have to worry about the weak h6-pawn. The game Meduna-Mrva (1996) continued with 14…f5 15.exf5 e4 16.g4, and White was better.

Keeping the kingside closed with 12…g5 plays things safe and leads to an old main line following 13.b4 Ncxe4 14.Nxe4 Nxd5 15.Qd2 Be6. After 16.g4 a5, play was then unclear in Alburt-Rizzitano (1986). Good, we are instead looking to sacrifice the h6-pawn and gain an edge in development.

12…Bd7! seemingly ignores White’s threat and tries to take advantage of White’s lack of development. This is more ambitious than closing the kingside with 12…g5. Then 13.b4 Ncxe4 14.Nxe4 Nxd5 15.Qd2 Be6 is an old main line. After 16.g4 a5, play was unclear in Alburt-Rizzitano (1986).

13.hxg6 challenges Black to demonstrate the point of his seemingly careless play. Instead 13.b4 Na6 14.a3 Rc8?! 15.Qd2 was better for White in Logovoi-Degraeve (1996). But instead 14…g5 should be fine for Black.
———————————————————
The more the position opens up, the more exposed White’s king can become.

13…fxg6 is Black’s only reasonable move and opens the f-file for Black’s rook on f8. He cannot afford to lose another pawn in addition to the h6-pawn. Instead 13…Qa5?! pins White’s c3-knight to White’s king and thus threatens to win a pawn with 14…Nxce4 next. But after 14.gxf7+ Rxf7 15.Bd2!, White keeps the initiative and an extra pawn.

14.Bxh6 happily snatches Black’s h-pawn and hopes to open up Black’s king. Instead 14.b4 is an interesting alternative. Then 14…Na6 15.a3 h5 16.f3 has been played several times and would lead to an unclear position. This might be White’s last chance to improve on our featured line.
———————————————————
The way to achieve compensation for the sacrificed pawn involves being able to gain control of the dark squares.

Your move allows White to retreat his dark-squared bishop from h6 to e3 or to g5. Black would then be left with the inferior dark-squared bishop. This is so because the pawn on e5 is blocking the vision of this piece.

14…Bxh6 trades dark-squared bishops. This seems to bring White’s rook to a more active post, but Black has seen a few moves ahead. Note that Black has to trade these bishops if he wants to get compensation for the lost pawn. For example 14…Qa5? 15.Bd2 is much better for White since Black’s bishop on g7 is rather passive with Black’s pawn on e5 blocking its vision. 
Black had an interesting alternative in 14…Rc8, which moves this rook to the open c-file and threatens to win the e4-pawn next with 15…Ncxe4. White’s c3-knight would then be pinned to White’s queen on c2. If White protects the e4-pawn with 15.f3 or 15.Bf3, Black could then continue with 15…Bxh6 next.

15.Rxh6 is clearly the only reasonable move.
———————————————————
Black must be a little careful that White won’t be able to get his queen anywhere near Black’s king.

15…Kg7 protects the g6-pawn and attacks White’s rook at the same time. Instead 15…Qa5? 16.Qd2! would allow White to keep a more active piece coordination as in our line. It also sets the trap 16…Nb3? 17.Qg5! 
Black has reasonable alternatives in 15…b5 and 15…Rc8. Both try to pressure White’s e4-pawn. White could keep an edge, however, by meeting 15…b5 with 16.b4 Na6 17.a3 and 15…Rc8 with 16.Qd2.

16.Rh1 is White’s best retreat for the attacked rook. Note that White is unable to get any reinforcement on the kingside.
———————————————————
White still hasn’t completed development and won’t be able to find a safe haven for his king.

16…Rh8 ends any hopes White may have had for a kingside attack. This move gave Black fine compensation for the sacrificed pawn in our featured game Kachiani Gersinska-Schmaltz (1994). However, even stronger appears to be 16…Qa5!? This stops White from playing b2-b4, pins White’s c3-knight to White’s king and thus threatens to win a pawn with 17…Nxce4 next. White can break this pin with 17.Qd2 and menace the queen check on h6. Following 17…Rh8 18.Rxh8 Rxh8, Black would have reasonable compensation for the pawn, e.g., 19.Rb1?! Rh1 20.Kf1 Qb4! 21.a3 Qd4. Note that White could not safely castle queenside in these lines as Black would quickly take over with moves such as …Rc8 and …b7-b5. 
White met 16…Rh8 with 17.Rxh8 Qxh8. Then castling queenside is still unwise because of a quick …Rc8.

So White tried her fortune by storming ahead with 18.b4 Na6 19.Bxa6 bxa6 20.0-0-0. But after 20…Rc8 21.Kb2 Rb8, Black had an edge due to his pressure against White’s unsafe king and due to his more active pieces. Note that instead 19.a3 Qh1 20.Kf1 Rf8 may give White better chances for survival. But this entire line appears to give Black too much compensation for the pawn. White is simply unable to find safety for his king. This is quite ironic since it is White who tried to attack Black’s monarch early on. But we learned that a lone rook is no match for the opponent’s king in most situations. The open h-file actually helps Black in many lines and made it undesirable for White to castle kingside. 
Once White exchanges his best minor piece, the dark-squared bishop, his position becomes much less dynamic. Black is correct in giving up a pawn in this line to be able to trade his passive dark-squared bishop for White’s counter part. Black’s king ended up perfectly safe on g7. Both sides have many other options on move 6 and 7. White should probably take a close look at playing 7.f4 or 7.Qd2 early on or try to improve with 13.b4 or 14.b4 later on. Black has several promising alternatives along the way, including closing the kingside with 11…g5 or 12…g5. But as we saw, this is not necessary. Also 9…h5 and 14…Rc8 would give Black a fine game.