First Kramnik vs DeepFritz, In Progress

An anonymous reader writes "Reigning world chess champion Vladimir Kramnik played the first match in a series of eight against the world's strongest chess computer. 'After the game Vladimir Kramnik said that he was never worried about losing the typical Berlin endgame that arose in his first game against Deep Fritz. The World Champion is the master of this line and Fritz was unable to take advantage of the white pieces.' There is live coverage of the event at the main website." We've mentioned this match a few times before.

Comments

No longer in progress: Draw

[Phouk]

"Kramnik was never worried about losing..." out of context is a bit misleading: Kramnik didn't win either, it was just a draw.

(For those who don't read the articles... ;) )

Re:No longer in progress: Draw

[Mr. PJR]

There's a saying among chess grandmasters--"Draw with black, win with white."

Anyway, I have a copy of Deep Fritz--the same program Kramnik is playing against. It's a comercially available chess program, designed to run on multiple processor computers. If you have a spare 90USD lying around, pick up a copy. It's a brutally strong program. You can buy it here: http://icdchess.com/cgi-bin/store.cgi (I have no affiliation with ICDchess, other than as a satisfied customer)

Re:No longer in progress: Draw

[Mr. PJR]

Would someone please elaborate on this? I have seen this in other comments and do not understand the reasoning

When playing the black pieces you have a disadvantage because black always moves second. Basically, the player of the white pieces has a one-tempo advantage and can to some extent determine the shape of the game (by choosing which variation of the opening will be played). Strong chess players can take advantage of the first move advantage when playing white by forcing black to defend or play an opening which the black player isn't strong at.

With these disadvantages, black's behavior is usually to hold on and try to draw. Then to try and win when he has the white pieces (because the players alternate colors).

Re:No longer in progress: Draw

[xintegerx]

Maybe because owning Deep Fritz is like having a Grandmaster caged up in your own home, one who tirelessly makes near-"perfect" moves?

Wouldn't it also be a great tool for analysis and to check what the best move, in a game you have played previously, would have been?

Wouldn't it improve your game?

People would buy Deep Fritz would use it for training. I've heard this about playing chess:

1/3 of your competition must be the same level as you are (makes for an even, fun game.)

1/3 of the competition should be stronger than you (to challenge yourself to advance and learn--why not learn from DeepFritz?)

1/3 of the practice competition should be weaker than you are (a match where you kick butt and raise your self-esteem after losing to higher competition [DeepFritz])

And, it doesn't take a genius to figure any of the above out (from your sig ;)

Not much of a contest...

[Anonymous Coward]

For those that are interested, the verdict among the chess world is that the computer is going to be exposed as a joke in this match. There are certain positions in which computers are very strong (tactical positions -- where each player has many choices over the next few moves and there are dramatic consequences), but there are equally many where they are not (positions in which long term planning is necessary and individual moves seem purposeless). Kramnik is not just strong -- he knows how to steer the game. The first game he had black and was thus trying to draw. So of course he immediately turned the game into a slow, boring game in which the computer's power was useless. Kramnik has shown previously that his anti-computer play is top notch, and you can look for it to win the match for him without problem.

Kasparov lost...

[Ulumuri]

Because of unfair playing conditions, and also because he didn't have access to DB before the match. Therefore, matches between Kramnik and Fritz will take place every other day, be adjourned after 60 moves, and Fritz will not be reprogrammed between matches.

Currently, opinion is siding with Kramnik. GMs Nigel Short and Raymond Keene predict a Kramnik win.

The game went as follows:

Deep Fritz(2807) - Kramnik,V [C67]
Brains in Bahrain Man-Machine Match. Manama (1), 04.10.2002
1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6 4.0-0 Nxe4 5.d4 Nd6 6.Bxc6 dxc6 7.dxe5 Nf5 8.Qxd8+ Kxd8 9.Nc3 h6 10.b3 Ke8 11.Bb2 Be7 12.Rad1 a5 13.a4 h5 14.Ne2 Be6 15.c4 Rd8 16.h3 b6 17.Nfd4 Nxd4 18.Nxd4 c5 19.Nxe6 fxe6 20.Rxd8+ Kxd8 21.Bc1 Kc8 22.Rd1 Rd8 23.Rxd8+ Kxd8 24.g4 g6 25.h4 hxg4 26.Bg5 Bxg5 27.hxg5 Ke8 28.Kg2 ½-½

Re:Kasparov lost...

[damiam]

Deep Blue was fed all of Kasparov's previous games, so it knew exactly what to expect and how to optimize its strategy. Kasparov had never seen a game played by Deep Blue. It's common among grandmasters to review that past games of your opponent to look for their weeknesses. Kasparov didn't have this chance, which put him at a disadvantage.

Re:Kramnick will win it

[Mr. PJR]

Don't expect a computer to ever win a blitz match, because computer's just don't have the insight to play well in those circumstances, which is where human innovation shows through.

I have to utterly disagree with this statement. Blitz games are quick games, such as a time limit of 5 minutes per side per game Blitz games are where computers are strongest, where their tactical ability, coolness under pressure, and lack of obvious mistakes shine through.

In quicker games, even weaker chess programs can anihilate strong human grandmasters. It's the longer games where humans are able to hold their own.

What happened...

[Espectr0]

...with Deep Blue? I would like to see a match between 2 top-playing chess computers for a change

Re:What happened...

[WiKKeSH]

Though this doesnt mean that Deep Fritz has ever played Deep Blue, I got this info from Deep Fritz's player stats http://www.brainsinbahrain.com/about/more_stats.ht ml:

1995: Won the world computer chess championship in Hong Kong, ahead of Deep Blue

First Post was...

[certron]

1.e4 followed by e5

then
2. Nf3 Nc6

Some notes on chess notation:
http://chess.about.com/library/ble21brd .htm
(it is common to omit the pawn designation, it seems)

Some opening moves (which was this one?):
http://chess.about.com/library/ble50ndx.ht m?PM=ss1 3_chess

The whole match:
1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6 4.0-0 Nxe4 5.d4 Nd6 6.Bxc6 dxc6 7.dxe5 Nf5 8.Qxd8+ Kxd8 9.Nc3 h6 10.b3 Ke8 11.Bb2 Be7 12.Rad1 a5 13.a4 h5 14.Ne2 Be6 15.c4 Rd8 16.h3 b6 17.Nfd4 Nxd4 18.Nxd4 c5 19.Nxe6 fxe6 20.Rxd8+ Kxd8 21.Bc1 Kc8 22.Rd1 Rd8 23.Rxd8+ Kxd8 24.g4 g6 25.h4 hxg4 26.Bg5 Bxg5 27.hxg5 Ke8 28.Kg2 ½-½

Hey, I learned something from the above links.

btw, where does 'DeepFrtiz' the name come from? The team flag looks to be Germany, but where did the name come from? They are using an 8-CPU Compaq machine, also. (Good thing chess opening moves are public domain... otherwise the US Fritz would be making sure they didn't fall into the hands of free citizens! erm. or something like that...)

Gratuitous Star Trek ref

[T-Kir]

...aiming for the draw; whereas the computer would foolishly (if it wins) try to win each successive...

Kind of reminds me of an ep of TNG, where Data lost a game against a Grand Master (when his strategy was to win), but won when he went for a draw (and the Grand Master gave up out of frustration).

Live coverage.....

[jimson]

An in related news, live play-by-play coverage of paint drying!

Trivial

[heikkile]

Chess is a trivial game - computers beat most of humans most of the time, and even on the top level it is a very close call. Given the speed computers advance, it is only a matter of time before no human can beat a computer in chess.

Machines have beaten man in many trivial games (tic-tac-toe. 100m sprint, weather prediction, etc). They have also failed in several "obviously easy" challenges (speech interfaces, AI, ...)

Before they play GO, I will not worry about my job.

Re:Trivial

[Chicane-UK]

Hm.. well you have to bear in mind that on challenges such as basic games and 'the 100m sprint' there only needs to exist a certain number of rules for a machine to compete / win. But when you talk about things like AI and speech interfaces.. well thats a whole new ballgame. Speech took mankind thousands of years to develop - and even now I would say we dont use it as effectively as we could.

And the less said about AI the better.. I dont think you can really apply rules and reasons to intelligence!

I dont' have time now,

[3-State Bit]

but one thing I rememer hearing much about karpov, back when kasparov was beaten, was that he, though not world champion, would have made a more interesting match against computerland, because of the fact that he focuses less on tactics (trying to out-think the computer by looking at combinations into more moves ahead) and more on abstract, pattern-based (such as in go) strategy, at which computers suck. Kasparov proved (insofar as you believe playing conditions were fair) that computers can out tacticate people, but perhaps a person whose style leans more toward abstract strategizing ("I want to keep this column open, because I feel it will be very important later" versus "I want to force the computer to lose that pawn, because I think I can pull off a combination in 43 moves")

i do need to go, but here are some things for children of this post to do:
o Look up some original reference (I saw many, many) that talked about how Kasparov's playing style is perhaps less suited to showcasing humanity's superiority to computers than Karpov's was)
o Look up whether Kramnik most resembles Karpov's or Kasparov's style.

One last thing.
Is it still true that in Go, computers play with a 14-move advantage and still lose to people who aren't even world-champion? Go is a game in which, because at each point in the game, it is unclear what groups of stones are alive and what are dead, pattern-based thinking is much more important. Would Karpov (and perhaps Kramnik) have made a better Go player than chess player?

When I come back, I'll add more to the thread, to anyone who wishes to discuss it.

Re:I dont' have time now,

[Scarblac]

i do need to go, but here are some things for children of this post to do:
I'm not a child of your post, but I'm a drunk chess player.

Look up some refernece that talked about how Kasparov's playing style is perhaps less suited to showcasing humanity's superioty to computers than Karpovs's was.
Kasparov is lethal when he has the initiative. He wants complex, tactical, attacking positions, and he's better at them than any human. Unfortunately, these positions tend to depend on calculation, which is what computers shine at. Karpov, like Kramnik, is more about prophylaxis, which is preventing any active options the opponent may have.

Although it must be said that at top level, all these players have a universal style. You can't become the world top player with a purely positional or a purely tactical style. Give Karpov a position that calls for a tactical solution, he's likely to play it. Put Kasparov in a quiet, strategical position, he'll usually know exactly what to do.

The differences show, mostly, in the choice of openings. They like different setups. Karpov choses the Caro-Kann (1.e4 c6) vs 1.e4, which is a very positional, defensive opening. Kasparov goes for the throat with the sharpest lines of the Sicilian (1.e4 c5). [if you're not a serious chess player, please believe me, that one square further makes a huge difference].

Kramnik plays the Berlin (1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6). An opening that gives White a positional advantage - just not enough of an advantage to win. It typically leads to an endgame that's better for White, though, in the hands of grandmasters, not yet winning. And he knows it well. There's no way a computer will understand all the subtleties in these quiet positions, Fritz isn't going to beat him here.

On the other hand, Kasparov actually lost to Judit Polgar, the world's highest rated woman, in the recent Russia vs Rest of the World match. Kasparov had a huge plus score vs Polgar beforehand, but he was tired, thought he could get an easy draw in that line just like Kramnik does. But he couldn't (a report of the match, including comments on the Polgar-Kasparov game, is at Chess Cafe). He just doesn't have the feeling for defending those worse, yet not yet losing, passive positions.

So the difference in style in small, but it's certainly there. And Kramnik's is much better against computers.

Re:I dont' have time now,

[legLess]

Blockquothe the poster:

Is it still true that in Go, computers play with a 14-move advantage and still lose to people who aren't even world-champion?

Oh, yes. Computers go programs are not serious opposition for anyone other than a weak to mid-level amateur. Here's a quick run-down of the go handicap system, for those not in the know: for each point of rank, or strength, difference in the players one stone of the weaker player's is placed on the board in a a predetermined position. Rank goes from 50 kyu (can't spell "go") through 1 kyu to 1 dan, then to 9 dan, then 1 dan to 9 dan professional. Thus a 5 kyu would give a 15 kyu a 10-stone handicap, and in theory, a 9 dan professional could give 68 (20% of the board) stones to someone who'd never played the game - and still win.

So when the poster says "14-move advantage" he means "14-stone handicap," which is huge. It's worse than that, though. A couple years ago, a dan-level player (a woman, not that it matters) beat the current computer go champion after giving it 27 stones. I can't find a bloody link right now, so you'll have to take my word for it.

Go is a game in which, because at each point in the game, it is unclear what groups of stones are alive and what are dead, pattern-based thinking is much more important.

Go is all about pattern recognition. The game is huge - easily the most complex game that people have created (where "life" is not defined as a game :). The board is 19x19 - 361 places to play - and all the stones have equal value. It's not possible for a computer to look 1% as far ahead in go as in chess.

Would Karpov (and perhaps Kramnik) have made a better Go player than chess player?

Who can say? They're very, very different games. I've played go for years, and every now and then I play chess with my brother. It feels very cramped, legalized, and formal. Go flows like a river.

Newsflash! Deep Fritz 'seriously damaged'.

[wackybrit]

I just mirrored a news story from the BBC that states Deep Fritz has been destroyed in a possible terrorist attack on the conference. For some reason the BBC removed the story minutes after it was published.

Gah!

[CyberDruid]

I am totally serious when I say that slashdot needs a "-1 wrong" or "-1 factual error".

• True - Kramnik is a staunch defender.
• Computers are not big on taking risks to begin with (they hardly ever sacrifice material for instance) and they don't really "play for a win", but if the operators wanted it to play more drawish, that would not be a problem, provided that they are allowed to adjust some positional parameters.
• A 5 min game would be extremely difficult for Kramnik. Quick games are basically just about calculating tactics, since the deeper aspects become hidden behind both sides poor play. A human excels in stuff like planning and sometimes logical reasoning, which both takes some time to do. It is a well known fact that computers don't improve their play much when given longer time (programmers will recognize this problem as "the exponential wall").

On a side note: In this game Kramnik drew easily because he could do some logical reasoning that no computer has ever done. He understood that in the final position, the computer could manoeuver around as much as it damn well pleased, there were simply no legal moves that could ever threaten anything. A computer will have great difficulty understanding this, since the calculation of variations will not show this simple visual fact.

In my opinion Deep Fritz will never beat Kramnik in a Berlin Defence. The team could try to deviate earlier, perhaps by closing the position with 4.d3, but this will also be easy play for Kramnik. They could also skip the Ruy Lopez altogether and play 3.Bc4 (Italian) or 2.f4 (King's gambit) instead, but these moves are not so common among the extreme elite. Kramnik would probably equalize comfortably against these moves. IMHO the team should try either switching to 1.d4 (at least for one game, to see where it leads) or just try to head for equal but tactically complicated positions after the King's gambit or the Italian, mentioned above. Playing 1.c4 or 1.Nf3 would probably be unwise. Kramnik knows these waters extremely well and could probably easily steer the game to a dull and totally safe position.

My money is on Kramnik, he will probably not lose a single game.

computers better at blitz

[hayne]

dh003i wrote:
its not a 5 minute game. Don't expect a computer to ever win a blitz match, because computer's just don't have the insight to play well in those circumstances, which is where human innovation shows through.

Actually, the exact opposite is true. Computers regularly beat even the best human players at fast time controls (blitz) since the humans are much more prone to making mistakes when they don't have time to think a lot. This is not merely my opinion - I think you will find few people who are familair with computer chess who would think otherwise. For example, here's what Robert Hyatt (author of Cray Blitz & Crafty) said in 1999( rec.games.chess.computer )

So today, game/30 is no longer safe and the computers are probably better there. Game/60 is also becoming more difficult for the humans, although I think they can do pretty well at this time control. But at 40/2, where there are no "blitz" time controls at the end, the GM players begin to show exactly why micro chess programs are not yet GM-level players at this time control. They simply know "too much" it seems, and they are able to exploit weaknesses they see while the programs are usually quite oblivious to what is going on...

I wrote the world's greatest chess program

[Rui del-Negro]

You may not be aware of this, because They tried to cover it up, but I once wrote the world's most powerful chess program.

My approach was simple: to compute every possible move in every possible game, and come up with the perfect sequence. It took 14 years to do, on a 700-CPU supercluster, but finally we 'solved' chess. The database was huge. The program was unbeatable.

Unfortunately it was also rather boring. The human would make the first move and, invariably, the computer would spend 4 hours sorting through the database and finally declared:

Checkmate in 14705 moves. I win.

RMN
~~~

Now...

[ymgve]

...imagine a Beowulf cluster of DeepFri...

AUGH! Stop beating me!

All your pawns are belong to...

NO! Not the baseball bat!

How long til somebody mods DeepFritz to run Lin...

Ack! Ack! ugh...you win...

Re:How many people can beat the computer?

[Scarblac]

Can anyone suggest some good chess strategy books?

"How to reassess your chess" by Jeremy Silman is probably the most-recommended chess strategy book, but it's not for beginners, more for somewhat advanced club players. A cheap, all-round good book to start with may be "The Mammoth Book of Chess", by Burgess and Nunn. Go to Amazon for reviews by people and sample pages, they're good for that sort of thing.

Of course there are always the game sites the offer chess online.

The best for Slashdot geeks should be FICS, at http://www.freechess.org, with its command line interface and geeky audience (usually 400+ players online). The best Linux client to play there is eboard.

Incidentally, SCID is a *great* GPL'ed chess database, originally for Linux but also ported to Windows, that makes Chessbase obsolete as far as I'm concerned.

Hope this helps.

8 processors

[vlad_petric]

The most interesting fact, IMHO, is that Deep Fritz runs on only 8 processors! A draw with the World Chess Champion is thus a major achievement, as, while it probably still uses a brute force approach (minimax is after all brute force), it's nowhere near Deep Blue in terms of computing power. It is definitely revolutionary in that respect.

The Raven.