• Subscribe to our blog and never miss a new post! It's easy, fun, and thrilling! (Well, it's easy - the jury's still out on the rest.) Just enter your e-mail address in the box below, then click the button.

    Join 154 other followers

Shredder wins 2010 World Computer (Chess) Software Championship

The chess program Shredder, brainchild of programmer Stefan Meyer-Kahlen, has won the 2010 World Computer Software Championship.

The event, administered by the International Computer Games Association (ICGA), was held in Kanazawa, Japan from September 24 through October 2. It was hosted by the Japan Advanced Institute of Science and Technology. The World Computer Software Championship (WCSC) was one of several events held as part of a larger group of computer gaming events. The WCSC was open only to PC software programs (not dedicated machines); the nine participating programs ran on identical hardware platforms.

The WCSC was announced a nine round Swiss system tournament, but with just nine programs participating it became in reality a single round-robin tournament. Meyer-Kahlen’s Shredder software won the event in convincing style, finishing with seven points and without a single loss:

Tournament crosstable generated by ChessBase

(This crosstable was generated by ChessBase, which doesn’t included “byes” in the score calculation. This explains the apparent discrepancy between the crosstable above and Meyer-Kahlen’s official announcement, which claims eight points out of nine for Shredder.)

The time control was G/45 with a fifteen second “Fischer” increment per move. All standard tournament chess rules applied to the event. Computer monitors were placed so that each operator could see the monitor of the opponent. The operators’ only duties were to input moves, respond to computer requests for clock information, and synchronize the computer’s clock with the official clock time of the standard (shared) timepiece.

This will doubtless sound very odd coming as it does from a guy who makes his living from computer software, but I generally don’t find computer vs. computer chess games to be very interesting. Such games are often long, drawn out affairs in which one chess engine wins by “grinding out” a victory. As I was playing through Shredder’s games from this event, though, I found a really interesting game from Round 7. In this game Shredder plays White against the last-place finisher Hector for Chess. This game fascinates me for a variety of reasons, not the least of which was a very pretty trap laid by Hector — one which might have turned the tables and saved the game had Shredder fallen into it:

Shredder – Hector For Chess [E04]
WCSC 2010 Kanazawa (7), 01.10.2010
[Steve Lopez]

1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.g3 dxc4
Now that White has “telegraphed” his intention to fianchetto the light-squared Bishop, Black takes the c4-pawn.

5.Bg2
As for 5.e3, advancing the pawn with the intention of playing Bxc4 makes White’s fourth move meaningless and a waste of a tempo.

5…Nc6 6.Qa4 Bb4+ 7.Bd2 Nd5 8.Bxb4 Ndxb4 9.0-0 Bd7 10.a3 Nd5
A very strange move from Hector, one which returns the pawn. 10…b5 is the “book” move in this position.

11.Qxc4 Nb6 12.Qc2
12.Qd3 is another good idea, with the intention of overprotecting the d4-pawn.

12…f5 13.Nbd2 0-0 14.Nb3
This is another overprotection idea; either Knight can now swing into action even after the d4-pawn comes under attack.

14…Nd5
There are two problems with this move. Black’s Rooks are still unconnected — he’s way behind in development. The other is that the Knight
becomes a target on d5.

15.Nc5 Rb8 16.Rfe1
Preparing the e-pawn’s advance.

16…Qe7
Black’s Rooks are finally connected, but White’s already well into his own middlegame (as we’ll see) and is comfortably ahead positionally.

17.e4 fxe4 18.Rxe4 Nf6 19.Re2 Nd8
If Hector was a human player, I’d say he was anticipating the coming dogpile on the isolated e6-pawn (after Ra1-e1 and/or Nf3-g5) and was
seeking to prevent it by overprotecting the pawn. It’s a good idea, but it comes with strings attached: Black’s position is slowly being tied in knots.

20.Ng5 h6 21.Ngxe6 Bxe6 22.Bh3

Position after 22.Bh3

It just keeps getting uglier for Hector. The Bishop is pinned to the Queen and is itself caught in a pretty serious crossfire.

22…Qf7 23.Nxe6 Re8
Not 23…Nxe6 because of 24.Bxe6 with a deadly pin. Black would have no choice but to play 24…Qxe6 25.Rxe6 and White’s up a Queen and
pawn for a Knight.

24.Rae1!
Shredder intensifies the pressure instead of settling for exchanges (which would still leave it a pawn ahead. Face it, Black’s just losing…). If 24.Nxd8
Rbxd8 White’s still winning here — he is a pawn ahead after all — but now has his own isolated d4-pawn to look after. There are a number of
options here, but 25.Rae1 is likely best, as it encourages exchanges which can only help White at this point. 25…Rxe2 26.Rxe2 Re8 (26…Rxd4
NO!!! 27.Be6 Qxe6 28.Rxe6 Kf7 29.Qb3 and White can now pretty much pick any path to victory.) 27.Re5+-

24…Nc6
Attacking the d-pawn, but this is just a futile gesture on Black’s part. If this was a human player, I’d say he’s posturing here.

25.Bf5 Kh8 26.Bg6

Position after 26.Bg6

This is just brutal. If this was a boxing match instead of a computer chess game, the referee would be stepping in to stop it.

26…Qd7 27.Bxe8 Rxe8 28.d5 Qxd5 29.Nxc7
Pinning the Queen and Rook and forcing a trade.

29…Rxe2 30.Rxe2+- Qd8
30…Qf7 would have been better; it still attacks the White Knight and doesn’t give up the initiative.

31.Ne6
Attacking the Black Queen, forcing a response, and maintaining the initiative.

31…Qd5 32.b4 Qd6 33.Qg6
Threatening Qxg7#.

33…Qd7 34.b5 Nd4 35.Nxd4 Qxd4 36.Re7
And we again have the threat of Qxg7#.

36…Qd1+
The move 36…Ng8 was arguably better just for simplicity’s sake.

37.Kg2 Qd5+ 38.f3 Qa2+
This isn’t just a “spite” check and actually conceals a tactic. There’s a subtle drop of poison hidden in this move — driving the White King to h3
actually has a point.

39.Kh3 Qg8 40.g4
Now White can’t play 40.Rxb7 because of 40…Qc8+ forking the White King and Rook, and turning the tables. This was a really nice little trap. I’m impressed with Hector for finding it and with Shredder for not falling into it.

40…b6 41.Rxa7 h5 42.gxh5 Qe6+ 43.Kh4 Qe1+ 44.Qg3 Qxg3+
Black, unfortunately, really doesn’t have anything better here.

45.hxg3
White’s pawns “are no longer deranged” (as they’d have worded it a century ago); the fat lady can start singing any time now. It would be interesting
to know what criteria Hector’s operators used to decide when to resign a game.

45…Kg8 46.Kg5 Nh7+ 47.Kg6 Nf8+ 48.Kf5 Nh7 49.g4
From this point on, mate is forced no matter what Black does.

49…Nf6 50.h6 Ne8 51.g5 gxh6 52.gxh6 Nd6+ 53.Kg6 Nc8 54.Rf7 1-0
The mating sequence goes like this: 54…Ne7+ 55.Rxe7 Kf8 56.Re2 Kg8 57.Re8#

Despite the fact that Black was behind for most of the contest, this was a very interesting little game; aspects of it are much more subtle than are normally seen in games contested between computer chess engines.

Congratulations to Stefan Meyer-Kahlen and Shredder for a fine accomplishment in winning the 2010 World Computer Software Championship!

ChessCentral, the leader in cutting-edge chess, offers multiple versions of Shredder:

Shredder12 (PC), Deep Shredder 12 (multi-core/multi-processor PC version), Shredder 12 (MacIntosh download), Deep Shredder 12 (MacIntosh download), Shredder 12 (Linux), Deep Shredder 12 (Linux)

Have fun! — Steve

About these ads

13 Responses

  1. I’m not sure I understand this contest… Specifically, Deep Junior is the only other engine I’ve even heard of..

    In other words: Where’s Rybka? :)

  2. Rybka won both the “free for all”, and “blitz” tournaments where hardware was unrestricted.

    Note “Identical hardware” doesn’t equal “fair hardware”, if it was run on one core, it’s unfair for programmers that have worked to implement multicore support, if it was ran on a 8core machine, it’s unfair for the programmers that haven’t been able to program multicore support, since they don’t have the resources, or have never tested on such “big” machines, or don’t care to implement it, so it’s as if they played on hardware 8 times slower, etc.

    By the same token, it would have been unfair to have Rybka participate on the software tournament just because all the other people can’t run on a 200 core cluster, though Rybka would have won anyway, it just didn’t participate because this tourney was irrelevant, some reasons:

    There are rating lists that run thousands of games, that show the engines’ strength on equal hardware, thousands of games are more significant than the few games played here, so to know who’s best, go here, or here, or here etc.:

    http://computerchess.org.uk/ccrl/4040/

    http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20BestVersion/rangliste.html

    http://ssdf.bosjo.net/

    If equal hardware is what matters, then people can run such matches at their own home, and see what engines are best without having to see a championship.

    What makes these tourneys interesting, is knowing the highest possible level of unassisted chess is attained at those time controls, one sees the latest, strongest beta version of the programs face each other, and one can enjoy the games without wondering if one can or not get that (the same way one doesn’t wonder about if one could buy Magnus Carlsen to the computer). One can see who’s the very best possible chess entity on the planet.

    Limiting the hardware misses the point.

    • Rybka chose to not participate in the identical hardware tournament.

      I chose to write about the identical hardware tournament.

      Life is full of choices.

      Steve

      • Yeah, but people is left wondering “Where’s Rybka?”, while it can be said “it was at the non-identical hardware tourneys (both long time control and blitz) of the WCCC, where she finished in first place.”

        Note my previous reply was a critique of the WCSC tournament, not a critique of the article, that has a well commented game and all.

      • Thanks for the clarification.

        My point is that I see no harm in giving credit where credit is due, in this case to Shredder, without feeling some sort of obligation to mention any of literally hundreds of other non-participatory chess engines. If people are curious about the absence of Rybka (or Fritz, or Hiarcs, or Stockfish, etc, etc.) the folks to ask are the engine’s programmers.

  3. Well, it’s sort of a bummer, reading here it seems there’s a WCCC and a WCSC and who knows now what else? If I had a good program I’D PARTICIPATE in all the world tournaments — anyhow, it would seem that Shredder has built up a good reputation and if a user like me has a “normal,” computer with just one, two, three or four cores, then Shredder is the winner.

    Right?

  4. Vyrton is wrong that you can trust the rating lists. He is a Rybka Forum censor who ensures that the public is in the dark. All the rating lists he gave deliberately exclude better engines than Rybka, due to fantasy beliefs about their origin/legality, or whatever the power brokers declare. This slur was completely exploded by a long 30 page report that gave a complete analysis.
    Better rating lists appear at open-chess.org and a multiple of various Russian sites, which are claimed to involved in piracy so that they can be censored. Houdini is #1, and from a FIDE-rated Belgian. Rybka chose not to compete in the regular class any more, and will largely appeal to the high-tech geek crowd with oodles of disposable income who can pay for mega clusters.

  5. Vyrton is also wrong that Rybka cluster is the “best chess entity” on the planet. because the Freestyle guys can top it.. COmputer intelligence is not really here yet, and adding a good human can add 150 elo or more.
    The short 9 game tourneys are more about having a superior opening book, and that is where Rybka excels. This can play a major role in deciding who wins, and is one reason why the Houdini maker didn’t travel to Japan.

  6. Rondo is a new version of Zappa, developed by a guy from Texas. Thinker has been around for awhile, its a guy from Canada (Perkson?) my memory says. Jonny is some sort of academic project that can use huge hardware (800 cores), but never lives up to it. The others I don’t know either, and are hobbyists I suspect. If Rybka can’t beat these guys on equal hardware, what’s the point of it?

  7. Thanks Steve: that’s a really interesting article.
    As for who enters these events: commercial programmers use computer chess tournaments as marketing tools. It therefore makes sense for them to enter their programs in events they are likely to win. It’s a shame that consumers assume the program that beats other programs most often is the best.

  8. Re: Post above…yea, it’s really a shame that people ‘assume,’ a program that beats others is the best.

    !?

    Next, I was going to ‘assume,’ that the NFL team with the highest rating was better than the team with the lowest.

    • I think you might have misunderstood his point. See Trample’s post immediately above it.

      The fact that so many of the “big boy” commercial programs declined to participate in an event run on equal hardware is quite telling.

      — Steve

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 154 other followers

%d bloggers like this: