Greengard's ChessNinja.com

Yours and Mainz

| Permalink | 29 comments

The annual Chess Classic, which used to take place in Frankfurt, begins in Mainz, Germany this week. (Joimans pronounce it MYnts and who am I to disagree?) An overview at ChessBase here. The centerpiece is an eight-game rapid match between multiple defending champ Vishy Anand and Teimour Radjabov. The Boy from Baku II has always been a tough rapid player and he's been working and improving his overall game in the past year. Oddly, these two haven't met over the board in nearly three years. Their last decisive game was Radjabov's spectacular win with black at Dortmund 2003.

The Ordix Open is usually one of the strongest open rapid events you'll ever see. Radjabov's win last year put him into this year's match with Anand. Aronian, Morozevich, Grischuk, Bacrot, and Shirov trailed him. That's strong, and it should be just as tough this year. On the circus side, there are also several shuffle chess events. In Mainz they call it "Chess960" for the number of possible positions using the Fischerandom version of the rules. (The king has to be between the rooks and there has to be a bishop of each color.) Aronian plays Svidler in what they are calling the Chess960 world championship. The FiNet Open is also shuffle chess. There are also several human-machine events. Radjabov plays Shredder and Svidler plays a Chess960 match against Spike.

The trouble of transmitting and storing shuffle games is still around. Importing PGN causes trouble with the "illegal" castling, so these games are often given in two or three pieces, depending on if one or both players castle. HIARCS 10 and a slipstream upgrade of ChessBase 9 (and all Fritz 9 generation programs, I believe) has solved this for the ChessBase data format. Importing PGN is still a problem as far as I can tell. I played a Chess960 game in Fritz (with the HIARCS 10 engine) and exported it and replayed with no trouble in ChessBase 9. But when I moved that game to a PGN database and then back into ChessBase it gave the castling move and subsequent moves as notes, not moves. I hope the Mainz folks distribute games in ChessBase format too. (And that ChessBase fixes this, since it's not the PGN format itself that has a problem. See below.) I know there are a few freeware programs that can replay 960 games, links anyone? How about freeware engines that can play it?


It looks like registration for the rapid open and the chess960 open don't close until sometime thursday right before they start, but is there any word on who will be participating? You suggest that it'll be "just as tough this year" as last year - do you know who will be playing? Thanks.

I'd like to see 960 become 480: it would be useful to distinguish between true variety and mere reflections.

e.g., if the starting position is K on d1/d8 and the Q on e1/e8, other pieces the same, you're still playing classical chess.

The King should always be on the Kingside (e thru h files). Or take a walk on the wild side...but pick one.

Nick, an early list I saw included Mamedyarov, Morozevich, Shirov, Grischuk, Bareev, Kasimdzhanov, Naiditsch, Harikrishna, Almasi, Volokitin... Also Sargissian, who seems to be one to watch.

Just pronounce the "ai" as the I in I :o)

"Yours and Mainz"

hehe, good pun, mig :)

On a side note, Morozevich has been playing Fischerandom blitz on ICC all this week to prep for the event :)

It comes to me as a surprize that the PGN format is unable to represent a Chess960 game.
Is the format itself the culprit, the fact that there is no standard agreed upon or because software engineers haven't bothered to fix their PGN readers?

I am reluctant to use proprietary formats because of vendor lock-in (such as .doc format for documents). Can the ChessBase format be read by other software?

To Bill Brock's comment: in chess960, mirror image starting positions do play differently because of the castling rules (OOO always ends with the king on c1, OO always ends with the king on g1.) So there really are 960 possible starting positions. I've played a game online with the standard start position, except reversed (K on d file, Q on e file). It gave me a headache, trying to play a standard opening in mirror image while keeping in mind the funny castling.

The problem with pgn is it assumes that 0-0 or 0-0-0 means the king will go to c1 or g1 and the rook will go to d1 or f1 (with corresponding black squares of course), pgn needs a tag in the header to determine whether the game is fischer random, and then it also needs to know where the rooks and king are, and enforce the castling rule... It isnt a huge deal, just a lot of work for every engine/game viewer.

For the last year or so I've had good Chess960/Fischer Random results using the Arena interface and the Hermann engine ( http://www.playwitharena.com/ ); I think some of the other engines that come with Arena understand Fischer Random also.

Hermann isn't enormously strong, but it's good enough to point out my blunders and make me feel like the idiot I am. If you want something stronger, you might try Spike ( http://spike.lazypics.de/index_en.html ) which apprently can be setup to work with WinBoard (not tried this myself, however).

I'm not well-informed enough to say much about the problems of recording Fischer-Random games in PGN, but I've never had any problems importing Fischer Random PGNs into Arena. I note the Wikipedia has a bit on the subject at http://en.wikipedia.org/wiki/Chess960#Recording_games_and_positions

>>> The problem with pgn is it assumes that 0-0 or
>>> 0-0-0 means the king will go to c1 or g1 and the >>> rook will go to d1 or f1

I thought in Fischer random 0-0 and 0-0-0 did place the King and rook in the same position as in classical chess, and only the starting positions are different?

Regardless, is there really a problem with PGN? It seems to me it's the programs that interpret PGN for Fischer random that have the problem. Hardly any are designed to know what the rule for castling is in that case.

PGN shouldn't have a problem with 960 because the position of the king and rook after castling is always identical to what it is in normal chess, regardless of where the king and rook's original positions. It just keeps it as 0-0 or 0-0-0. But various programs have trouble importing it because it looks like an illegal move to the parser. I didn't mean it was a problem with the format itself.

I discuss this FEN-PGN castling issue in detail in my book "Play Stronger Chess by Examining Chess960" (search Amazon.com for 'chess960').

FEN is the portion inside .PGN that captures the position as it is before the start of the moves given in the rest of the .PGN.

Contrary to what Mig might have implied, Fritz9 has no problem interpreting .PGN files from chess960 games (tho Fritz8 cannot). However, the original FEN notation specification from years ago has proven inflexible.
(Over the long history of chess the rules have been adjusted many times, so flexibility is a good thing.)

Fritz9 wants still-eligible castlings to be denoted by the start column letters of the still-eligible rooks.

So instead of KQkq, for Fritz9 chess960 FEN should be like BEbe (if the start position was say QRBK-RBNN S#870).
In essence, Fritz9 has switched to chess960 style FEN, but it still understands the old KQkq.

Reinhard Scharnagl has designed ways to partly salvage the older KQkq. The problem with enhanced KQkq is that complexities creep in for rare cases when both rooks are on the same side of the king on rank 1, in a middle game position. The claim made for salvaging Kqkq is that it increases backward compatibility. On close scutiny that claim is debatable.

The programmer for Shredder, Stephen Meyer-Kahlen, is the main proponent of the new rook columns style.

Gene Milener

The open Ruy Lopez is a specific example of why there are technically 960 start positions instead of just 480.
Note that in the "reciprocal" to the traditional setup, RNBK-QBNR S#534, White has an extra tempo for not having to move his rook to the column on which Black's king and knight reside (column 'd'), after White castles O-O-O.

There is a lot more to say about castling in chess960, but it is far too much to write in a Blog.

Gene Milener

I was referring to ChessBase 9, not Fritz. ChessBase 9 doesn't correctly read 960 PGN, even if it is exported from Fritz 9 (or itself). To go over 960 games in ChessBase 9 you would have to open the PGN in Fritz 9 and save them in CBH, which CB 9 can then replay correctly.

Over the years there have been numerous "opening theme" chess tournaments, such as tournaments devoted to the French Defense. I know for a fact that at least some of these games have been submitted to the USCF and have been counted in the formal player rating process. The USCF knows about this and says it is fine.

So starting from the traditional setup position, consider the following opening for a theme tournament (for brevity only White moves are shown, as Black makes the same "mirror" moves):

1. e23
2. Qd1f3
3. Bf1e2

4. Be2d1
5. Ng1e2
6. Ne2g3

7. Qf3e2
8. Qe2f1
9. Qf1g

10. Ng3f1
11. (any, competitive play begins)

This is 99.9% the same as chess960 setup RNBB-KNQR S#549. Castling options still full. All legal.

Gene Milener

>>> Reinhard Scharnagl has designed ways to partly
>>> salvage the older KQkq. The problem with
>>> enhanced KQkq is that complexities creep
>>> in forrare cases when both rooks are
>>> on the same side of the king on rank 1,
>>> in a middle game position. The claim made for
>>> salvaging Kqkq is that it increases backward
>>> compatibility. On close scutiny that
>>> claim is debatable.

wikiedia has a nice explanation on Fen notation:


After reading that, I know now that "-" indicates neither side can castle, and "KQkq" indicates White and Black can castle both kingside and queenside. Thus "Kkq" would indicate White can't castle queenside.

Why is this convention incompatible with Chess960? You mention rare cases with both rooks on the same side as the King, but that is not a valid position in Chess960.

I'm guessing that you must be talking about a variant of chess that allows castling Kingside or Queenside with the same rook? In that case I can see the KQkq system not working as it does not identify which rook has moved.

Nevermind .. I get it now.

The there is no way to know which rook is at the starting position.

Thanks for the informative post Gene!

RSS feed is not working anymore. Am I doing something wrong?

Goran, according to Mig the rss feed is broken. It is not a priority to fix it. Shame - I use it as well.

Check out free-castling! That's cool too!

Radjabov lasted only 16 moves against Schredder (imminent mate). Svidler lasted 53 moves against Spike, but he was slowly plucked bald without ever having a whiff of counterplay. Human finger nails don't compare to screwdrivers, especially not for rapid jobs.

After losing to Naiditsch 0.5-3.5 yesterday, Harikrishna made a sensational comeback with 4 wins today, winning the match 4.5-3.5, wow!

if you watch some of Svidler's games at chess 960. he seems to aim for normal chess positions.. and often succeed's. Also, if you look at Hari's 4-0 wipeout of Naditsch today and compare it with his play yesterday, it seems he figured that out as well. Objectivly, I doubt it is the best strategy but in human vs human play it seems to be the best idea (at least in rapid) to try and get a normal config of pieces BEFORE your opponent can.

The position in their third game, looked like some sort of Sicillian Scheviningen with Black Bishop on b6 instead of g5.. Infact, Naditsch tried one move too hard to get a 'normal' Sicillian placing of his Knights (Nc1-b3-d4, Ng1-e2-c3)... 4 knight moves in the first 8 moves... someone tell these guys that the basic principles of chess still apply.. DOes anyone wish to sign me up for a book on chess 960.. come to think of it... IT'S GONNA BE HARD TO GET AN OPENING VARIATION NAMED AFTER YOURSELF IN CHESS 960... with the starting pos changing all the time..

Radjabov tried to get creative/original (aka stupid in chess 960 rapid against a machine) in the first game and you know whappened

final tally over all 4 games is 3.5-0.5 for the screwdrivers over the (2700+) finger nails

Re the RSS feed, if you can change your reader to use the Atom file instead of the RDF, that seems to be working. Most readers can handle both types. I'm actually not sure why 3.3 killed the old RSS feed.

Okay, I recreated the legacy index.rdf RSS feed, so now there are three to choose from. Anything that worked before should now be working again. Let me know how it looks and if it's okay.

Thanks Mig, the rss works again for me now. :)

Twitter Updates

    Follow me on Twitter



    About this Entry

    This page contains a single entry by Mig published on August 14, 2006 6:46 PM.

    Get Your Brit On was the previous entry in this blog.

    4th Staunton Memorial is the next entry in this blog.

    Find recent content on the main index or look in the archives to find all content.