“On Chess 960, Classic, and Computing”

by Michael Sutton

Chess 960 is also known as Fischer Random Chess, after Robert James Fischer, its inventor. Bobby felt that chess was rapidly evolving towards over-analysis.

I remember in the early 1970s attending a Canadian Open with Lawrence Day, with his dark, dark hair. Back in those times, we had proper adjournments, sealed moves, and resumption with new time controls.

What happened? Turns out, over-analysis by chess programs, like Fritz, Shredder and so forth, turned adjournments into pointlessness. If the purists are so insistent that chess 960 is not “real chess”, why did we let computers rule the analysis of classic chess? When did that become acceptable?

I find it ironic that the solution we adopted to no adjournments was to use incremental clocks. This was another invention of Robert James Fischer. He was originally laughed at. I have often said that the only institution more radical than Chess is the Catholic Church. Change comes slowly, and grudgingly to our game.

But change we must if our game is to stay fresh, and remain as a true head-to-head test of human skill. Chess 960 is a great way to achieve that.

During the recent world championship, Vishwanathan Anand had a bank of six grandmasters. Veselin Topalov relied on a 128-node CPU for analyzing opening positions. It often seemed that the player to lose was the first one to go off book analysis. This to my mind is not chess.

Chess should be the best move over the board at the given time, a test of skills in evaluating subtle differences in positions. Not raw brute, ply hammering by multi-threaded algorithms.

Chess 960 itself does not lend itself easily to opening analysis, since the analysis of starting positional factors is usually too subtle to lend itself to simple algorithms.

Based on this, most computer programs rely on opening books. In classic chess, those opening books were developed by human players over hundreds of years of play. Take way that opening book, and the computers have a much harder time. So most classic chess programs are essentially cheating, since they are looking at a book but the human player is not.

I had a long talk with IM Zvonko Vranesic at the Canadian Open. He had been honoured for his historical contribution to chess in Canada. I had beaten him in a simultaneous at Hart House over 35 years ago, and it was a good opening gambit to talk about chess and computing. Zvonko decided on a career as a computer scientist at U of T, rather than push on to the GM Title. He did not regret that, since he was able to teach at the top level as a professor for many years. Many of the current techniques in computer chess algorithms came from his program at the U of T. However, he still feels that the brute-force ply analysis shown by most programs has not captured the true human reasoning process in chess. He agreed with my analysis of chess programs vs. human play. He lamented that chess programs are dominating, but that we are no closer to emulating human intelligence.

About the Game

First of all, the game is not as random as players think. You are only positioning five pieces randomly on the back rank, with two major constraints:
a) the two bishops always go on opposite colour squares
b) the king is always placed between the two rooks, since castling in chess 960 is allowed

I designed a special set of polyhedral dice (see attached .pdf) for the 2010 Canadian Open for placing the first five pieces, with the king and rooks taking up the last three open squares:
a) a white d4 (4-sided die) for placing the white-squared bishop
b) a grey d4 for placing the black-squared bishop
c) a red d6 for placing the first knight on open squares
d) a green d5 for placing the second knight on open squares
e) a gold d4 for placing the queen on open squares

These permutations generate 959 other positions outside classic chess, which in itself is a 960 position. So think of classic as a subset of 960. (It’s true!)

We successfully used the dice in Alexander Shabalov’s first 960 simultaneous. He went +10=3-0. He said he had never worked harder in a simultaneous, but also he said it was the most fun as a GM he had ever experienced.

There are chess clocks now that can compute 960 positions. However, as a bit of chess throwback I would rather roll the dice than have a microchip tell me where to go. It is a matter of taste I suppose.

Strategy

Nothing changes in chess 960 from classic chess. All pieces move normally, things like en passant and pawn promotion work as usual. You rely less on opening traps, and more on sound development principles, i.e.,
a) pawn control of the central squares
b) king safety – i.e., castling
c) good open lines for major pieces
d) getting misplaced pieces to the centre of the board – i.e., knights out of the corners, bishops onto good diagonals

My experience is in some 960 positions you will get weak pawns and poorly placed pieces, but both players are working under the same constraints. Good opening moves to cover these weaknesses are important.

I find the game will strengthen your middle-game tactics, and usually you get satisfying middle-games with both players competiting tactically as equals. It takes a few games to stop hanging pawns and finding out which squares aren’t suddenly covered, but your mind rapidly adapts. I find playing 960 becomes quite addictive!

Most chess 960 games after 10-15 moves, will look like a classical chess position. You won’t be able to easily determine how the game originated.

Castling

A final note on castling, since it is the most misunderstood part of chess 960.

960 essentially reinforces the castling rules in the classic game; i.e.,
a) the king cannot castle while in check, through check or into check
b) the king cannot have moved prior to castling, nor the rook the king is castling with
c) the way between the king’s move, and the rook’s move must be clear, outside of the king and rook themselves
d) in some 960 castling positions, the king and rook may simply exchange places, only the rook may move, or only the king
e) if you castle either queenside O-O-O or kingside O-O the pieces must end up in the classic final position, i.e., Kc1-Rd1 or Kg1-Rf1

Final Thoughts

Enjoy the variant. It is getting better coverage in the chess world. Nakamura is the current World Champion; Svidler had a good run before that. I’m told Nakamura is undefeated in online play. The Championship has been played by top grandmasters since 2001.

It’s sad that Bobby gets recognition for his clocks, but not for 960, which I feel is his greater contribution to chess. No one should apologize for his later apparent madness and eccentricity. You need to look past that and see what gifts he gave to this great game of ours.

Don’t be afraid to ask me questions, or explain any of the rules of play or positioning. I’m glad to help any club member explore this great extension to chess.

This lecture was given by Michael Sutton on November 29, 2010 at the Annex Chess Club