But as always, there are exceptions to these rule. In his classic work "Art of Attack in Chess", Vladimir Vuković points out the special situations in which a player may choose to postpone castling, or not castle at all:
1. Castling is postponed or not carried out at all in favour of some other useful actions, such as an early attack
Reasons like capturing an opponent's piece or ruining his pawn structure,which may benefit you more than castling. Or sometimes, when one is confident that his attack is strong enough to break down the opponent's position quickly enough, then he may choose not to castle simply because he does not want to waste time. I know of players who choose to castle late/not castle at all in favour of an early attack.
Remember this position?
Meesen vs H. Muller
Correspondence Game 1928
|Position after 10. b3|
Yep, it's the same position I showed as the very first example in my "How to attack" series! It is Black's turn, and he can choose to castle on either wing. But because he has the more active pieces, coupled with the vulnerable position of White's king, Black decides not to waste time and launch his attack straightaway:
10... d4! 11. Qxd4 (11. Bxc6+ and Black wins a piece with ...Bxc3+ and ...Bxa1) Rd8 12. Bxc6+ Kf8 13. Bd5 Rxd5! 0-1
White resigned since 14. cxd5 leads to 14... Qxe2#. By choosing to attack rather than castle, Black manages to bring a quick end to the game.
2. Castling is postponed because for the time being it is too dangerous
Thus it would be more logical to eliminate the opponent's threats first before castling. A simple example can be shown in the hypothetical position afrer 1. d4 d5 2. Bf4 Nc6 3. e3 e6 4. Nc3 Be7 5. Nf3 Nf6 6. Bb5. Here Black has the choice to castle kingside, but after 6... O-O White can damage Black's pawn structure by playing 7. Bxc6. Hence if Black wishes to retain his pawn structure he will choose to postpone castling by playing 6... Be7 followed by O-O later on.
Another example follows after the moves 1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. d3 Nf6 5. Bg5 h6 6. Bh4 d6 (D)
|Position after 6...d6|
Here, to play 7. O-O? will be premature and incorrect, for it allows Black to reply with 7... g5! following with a pawn avalanche against White's kingside; this threat must be prevented before White can safely castle. On the other hand, if White were to play another developing move such as 7. Nc3, and then Black still attempts an attack with 7... g5 and 8... h5?!, the attack will be less dangerous because White has yet to castle. In fact, White could even get out of harm's way by castling queenside later on!
3. Castling is postponed when one wishes to create favourable attacking/defending conditions-- for example, he can delay till his opponent castles, before deciding which side to castle on.
Remember what I said in Part 2 of my "How To Attack" series (http://nushsblackknights.blogspot.sg/2013/08/how-to-attack-part-2.html) regarding the castled position? When the king is still in the centre, he is free to choose whether to castle kingside, queenside or remain in the centre. But after castling, he has already committed himself to a protective shelter on one side of the board, and it is difficult to change the nature of this feature throughout the game.
That is why sometimes, a player may choose to delay castling until his opponent castles first, because he wants to see which side of the board the opponent has been committed to before making his own decision on which wing to castle on. If he is aiming for an attack, he will most likely castle on the opposite wing; if instead he wants to avoid being attacked, he will castle on the same side as his opponent.
Such delays usually occur in typical Sicilian positions like this:
|Position after Black's 7th move|
It is White's turn, and he can choose to play 8. O-O. But by doing this he will find it difficult to attack his opponent after Black plays 8... O-O, for his king gets into the way of the attacking White pieces. Thus, rather than castling straightaway, he makes further preparations to castle on the opposite wing-- for example by playing 8. Qe2 followed by 9. O-O-O, in which his pieces get maximum freedom to attack Black's kingside.
Here's another interesting example:
|White to move|
Once again, White can choose to castle kingside immediately, but by doing this he allows Black to castle kingside too, thus averting the chances of White launching a strong attack. So perhaps he should play 1. Qd2 and 2. O-O-O later on? But what if after 2. O-O-O Black replies with O-O-O, and then White suddenly finds that he cannot launch an attack without compromising his own king's position?
So the best choice for White will be to play 1. Qe2/Qd2 first, and let Black decide which wing to castle on. If 1... O-O then 2. O-O-O, or if 1... O-O-O then 2. O-O. White castles on the opposite wing from his opponent in order to get the maximum freedom while carrying out his attack.
4. Castling may be postponed or even abandoned altogether when the centre is permanently blocked
In positions where the pawns obstruct the centre such that it is impossible to break through in that theater, the most logical plan for both sides will be to launch their attacks along the wings. In such a case, the kings will actually be safer in the centre (at least for the time being) as compared to when they are on the side of the board!
We look back to our previous example:
|I'm back! (:|
After 1. Qe2, the most logical reply for Black is also a postponement of castling. Notice something interesting about the position-- the structure of the pawns is such that the centre is blocked; Black has a potential outpost on e4, but it is almost impossible to get his knight there. As a result attacks can only be conducted along the wings, thus the kings are safer in the middle of the board for now. Thus the game might continue as follows:
1. Qe2 Nd8 2. h3 Nf7 3. Qd2 O-O-O
Here Black can safely castle queenside; if White continues 4. O-O-O the possibility of a flank attack disappears, while if 4. O-O Black can launch his own counterthreat with 4... Rfg8, threatening 5... g5.
Although over here the preliminary skirmish has ended with castling, there are also other cases where the centre is so clogged up, the kings choose to stay in the middle throughout the course of the game.
5. Castling is unnecessary when the endgame is already not far off, in which the king will be better placed in the centre of the board
The statement speaks for itself; I guess I don't really need examples to help illustrate the point.
And these are a few cases in which the postponement or total abandoning of castling can be justified. However, to decide whether or not to castle, or when is the right time to castle, ultimately lies on the player's own judgement and the features of the position itself.
Oh, and one final hint: Even if you do have justifications to postpone castling, it is always good to get ready to castle anytime, so that you can do it quickly if the situation around the king's central position gets too hot. There can be a big difference in spending one or two tempi , especially when your opponent is attacking you. Thus, the best policy for one will be to get ready to castle as soon as possible (i.e. clear all your other pieces off the back rank), before deciding whether it should be postponed or not.
"Art of Attack in Chess" by Vladimir Vuković