23 June 2018

Another Purported Problem with Castling

I ended the previous post, A Purported Problem with Castling, with the observation:-
Another general rule is that castling considerations are more complicated in chess960 than in traditional chess, even if specific start positions might be less complicated.

It turns out that the observation also applies to the rules for castling. The 'Purported Problem' post was based on hundreds of comments to a Chessbase article, of which the last half might be called the Petrarlsen - Celeje debate. For more than a month the two parties discussed the pros and cons of various castling rules, all of them alternatives to Fischer's chess960 rules. I won't enter into that debate -- this is a chess960 blog -- but there were a number of thought-provoking points that are worth discussing. For example:-

In certain [chess960] positions, the [castling] King moves toward the center.

A little thought is enough to determine that these are positions where the King starts on the b-file and castles O-O-O. This brought me once again to an old post on my main blog, Introduction to Chess960 Geometry (March 2009), where I developed the following table. Pardon me for taking a snapshot of the table and its discussion, which is easier than copying everything separately.

From that table we learn that there are exactly 108 positions where the Kings start on the b-file. When one of those Kings castles O-O-O, it starts on the b-file and ends on the c-file. While that might be a (small) disadvantage, it is counter-balanced by the simultaneous activation of the Rook that started on the a-file. Since all start positions with a King on the b-file imply a Rook on the a-file, there is another important consideration for the opening plan: How to activate that Rook? There are four ways to do this:-

  • Castle O-O-O [moving the King toward the center]
  • Castle O-O [switching the King from the b- to the g-file]
  • Moving the King to the second rank
  • Moving the Rook along the a-file

The strategy that a player chooses -- remember that both White and Black have the same options -- depends on the overall strategy for developing all eight pieces. Each one of those four choices for King/Rook activation has an impact on the choices for the other six pieces and vice versa.

That last observation might seem obvious, but a second observation might be less obvious. Of the six files where a King can start the game, the simultaneous considerations for castling (King safety and Rook activation) present six different scenarios. For example, in those positions where a King starts on the d-/e-files, the choice of moving the King to the second rank is less attractive than when the King starts on the b-file. On the d-/e-files, the King is more exposed to attack than on the b-file, and moving to the second rank doesn't really change this.

The special considerations for castling arise because a single move alters the position of two pieces of the same color. In the Petrarlsen - Celeje debate, Petrarlsen (a chess960 antagonist) argued that the chess960 castling rules must respect the exact same goals of King safety and Rook activation found in the traditional setup (SP518 RNBQKBNR). Celeje (a chess960 protagonist) was more flexible. What would Petrarlsen say about the start positions When Castling Undevelops a Rook (September 2010)?

It's curious that even in SP518, the consequences of castling to the two sides are not completely equivalent. Castling O-O-O moves the King one square off-center and develops the Rook to the center. Castling O-O moves the King closer to the corner, but doesn't bring the Rook to the center. Both scenarios often require a second move to correct the deficiency of the castling move. Castling O-O-O is often followed by Kc1-b1, while castling O-O is usually followed by a further Rook move.

Similar considerations apply to all of the other 959 start positions. If we want all chess960 positions to mimic the traditional position, we might as well just play only that position.

16 June 2018

A Purported Problem with Castling

My previous post, Comments on Purported Problems (May 2018), continued a mini-series on a Chessbase article, 'The problem with Chess960'. After commenting on the article itself (follow the 'Purported Problems' link to find my original post and comments), I selected some thought-provoking discussion points from the more than 200 comments made against the Chessbase article. One in particular is worth a discussion on its own:-
Petrarlsen 3/2/2018 03:31 • In one of the Carlsen - Nakamura game, the two players castled on move one, and I rather think that every game played with this position between top GMs would follow the same course; this feels more like a farce than like the beginning of a serious game.

The game reference is to the recent match, 2018 Carlsen - Nakamura (February 2018). The commenter continued,

What can be the meaning of castling if the players castle on the first move and wouldn't even consider playing any other move?

Ignoring the arrogant assumption about knowing what the players considered or didn't consider, let's look at the position. The PGN is available via my '2018 Carlsen - Nakamura' post. It turns out that the position was used in games 11 and 12 of the match, when both games were drawn.


SP324 NBBRQKRN

In game 11, with Nakamura as White, the players continued 1.O-O O-O. With Carlsen as White, the first moves were 1.c4 O-O 2.O-O. Carlsen at least tried 1.c4, but then castled on his next move.

Looking at the position more carefully, what are the logistics behind castling O-O-O? Taking White as the example, we have to move the dark-squared Bishop and the Queen to enable O-O-O. Both moves can be prepared by d2-d4 (or d2-d3), after which there is a further question of where to place the Bishop or Queen when it moves. Neither piece has an obvious developing square, and the Bc1 is perhaps better on the long diagonal. So it takes at least four moves to play O-O-O, after which there is no particular advantage in the resulting position. The move O-O, on the other hand, looks natural -- the King is a little safer and the Rooks are connected. Isn't that what castling is supposed to achieve?

Rather than being 'more like a farce than like the beginning of a serious game', the move O-O is the first idea in what promises to be a long, tight struggle. It also adheres to the principle of playing the obvious moves first. As for the other start positions where O-O is possible on the first move, I once counted these positions in Introduction to Chess960 Geometry (March 2009, on my main blog):-

There are 90 start positions where the players can castle O-O on the first move. [...] There are 72 positions where the players can castle O-O-O on the first move.

I am sure that not all of these 90 positions are as straightforward as SP324. As a final thought, in a post titled On a Losing Streak (May 2015), I wrote,

After the game, my opponent said, 'I have found from past personal experience, castling on the first move makes for a very difficult game!' I could hardly disagree with him, especially when it is followed by a dubious gambit.

While that opponent might have exaggerated his personal experience, it is a more useful general rule than to claim that a player should always castle on the first move whenever possible. Another general rule is that castling considerations are more complicated in chess960 than in traditional chess, even if specific start positions might be less complicated.