Chess: Kramnik vs Deep Fritz (again)

Well, after today’s game 2, I think I will never again be embarrased when I blunder in a game.

Holy cow, talk about stunning!!!

Kramnik lost game two on a horrendous blunder that most chess players would have seen. Granted it goes to show you just how mentally exhausting it has to be to play against Fritz.

Wow. Is humanity doomed?

– IG

The game of chess has been solved for all positions with 5 pieces (including both kings) or less left on the board. This is known as endgame tablebases and is achieved by generating every single legal position with tese pieces and sorting them into a database.
Some 6 piece positions have also been calculated.
This has taken years, so I think that positions with 20 or 30 pieces are still a long way off being solved.

According to chessbase.com, Kramnik made the move (Qe3) allowing mate in one, stood up and was about to wander off for a drink when the machine’s operator played the mate.

Just an oversight. Kramnik was probably too relaxed after the earlier complications, and assumed there was nothing to see.

As for being doomed, certainly computers will soon be unbeatable.
Postal chess is still played, despite the fact there is no way to check on the palyers using computers.
People still run races, despite the invention of bicycles!

It’s still reasonably easy to detect whether a bicycle is being used.

I take your point about postal chess. But surely no one believes it’s still a test of human-only skill?

I meant my “is humanity doomed” with a bit more humor than originally came across. Sorry :wink:

– IG

Here’s the final position, right before the checkmate. I guess when you’re looking for mate, it’s stunningly obvious. I’m not a chess expert by any stretch, but I don’t think I’ve ever even seen a mate at Grandmaster level.

There is also the problem that to store a 32-man tablebase, your hard-drive would need to be inconveniently planet-sized. :slight_smile:

They happen from time to time. Players of World Champion rank have left their queen en prise, one player has left a mate in one on the board and the other player missed it, and so on. The blunders are all there waiting to be made, except by computers.

Mind you, sometimes it works against the iron monsters. There was a famous instance once of a computer playing a losing Rook sacrifice because it had spotted something like a mate in three against it, and only giving up the Rook delayed it. The mate was obscure enough that the watching masters were dumbfounded by the senseless sacrifice, and in such a case a human might have opted for playing a nonchalant move and hoping the opponent missed it.

Join the club

Having computers play chess better than almost all humans hasn’t stopped humans playing chess against humans; just as the invention of the bicycle hasn’t stopped foot races.
I suppose that some people would be embarrassed to play chess at all when a computer could find all their mistakes.

My mate is playing in the current World Postal Championship. He doesn’t use a computer and says he trusts the other players.
Personally I wouldn’t!

The pont is that top players don’t check every position for mate. They have a feeling about how the game is going. OK, if they feel threatened they will check for mating lines - but there isn’t time to do this every move.
Kramnik had been pressing Fritz and expected a winning ending once the queeens came off. Qe3 threatened mate by Black, so Kramnik just stopped analysing.

Grandmasters usually resign before mate.
Here are some blunders! :eek:

World Championship match Steinitz v Chigorin - Chigorin allows mate in 2 by 32.Bb4?

http://www.chessgames.com/perl/chessgame?gid=1036366

World Championship Qualifying Gligoic Book - Gligoric misses 33. Bf6 mate (he does win eventually)

http://www.chessgames.com/perl/chessgame?gid=1163005

Monte carlo tournament Von Popiel v Marco - Marco resigns instead of playing 36. … Bg1! winning the queen

http://www.chessgames.com/perl/chessgame?gid=1099222

But funniest of all is this …

Sztern - Lundquist
Australia, 1983

DIAGRAM
White: K/b1; Q/e3; R/c1; N/e5; B/d3; P/b2,d6,f4,h3,h5.
Black: K/g8; Q/b6; R/a3,e8; N/d4; P/b7,c5,g5,g7,h7.

28…Qxb2+
0-1

In the diagram position, Black offered White a draw. White, as was White’s right, asked to see Black’s move before giving an answer. So, Black played a queen sacrifice. White was so surprised by the move, which forces checkmate (29.Kxb2 Rb3+ 30.Ka2 Ra8+ 31.Ba6 Rxa6#), that White resigned! But White completely forgot that White had been offered a draw, which White could have accepted, instead of resigning!!

http://www.geocities.com/rookknightrook/ChessArticlesDavidCohen.htm#A0517

Fox and James in ‘The Complete Chess Addict’ estimate the number of legal positions as 2 * 10^43.
How long before computers can store that much?

World Computer Championship 1977, Duchess v Kaissa - Black plays 34. … Re8 to avoid 34. … Kg7 35. Qf8+! Kxf8 36. Bh6+ and next Rc8+ mates.

http://www.chessgames.com/perl/chessgame?gid=1162183

[mode=geek]That’s the one, I recognise the sequence tho’ I’d forgotten the game.[/mode]

Well, say you need about 100 bits to store a tablebase position, and that your hard drive can store one bit on an individual atom. You’d need a hard drive of 2 * 10^45 atoms. The earth has somewhere in the neighborhood of 8 * 10^49 atoms, so your hard drive would need to be 1/40,000 the size of the earth, i.e. a sphere with a radius of 235 miles.

It’ll be a while. :slight_smile:

I see. But isn’t the presence of the white knight on f8 enough to at least flag something in the Grandmaster’s head that “hey, don’t forget about this guy, there’s a possible mate at h7 if you don’t pay attention.” To me, it seems on the level of forgetting a piece and getting back-rank mated.

I can see a grandmaster not actually looking for mate, because mate is so rare. But shouldn’t he at least be looking for checks? In my much lower-level experience, checking can be a valuable way to buy time, and hopefully eventually put you in a position where you can threaten something else at the same time. Or, at least, you can force your opponent to make a move he would rather not, to defend against it. This seems especially the case, if the queens are still in play with the board that sparse. It seems to me that Kramnik should at least have considered “What if he checks me with that queen; how do I respond to that?”

The only time you usually ever see a Grandmaster allow mate in 1 is during time trouble. GMs clearly look for mates, they look for checks. But they are fallible, they can crack under the pressure. Almost every chessplayer suffers foom chess blindenss, the momentray inability to see something obvious. I suspect with GMs it happens very rarely, or at a deeper level.

This makes an interesting point, about how top players are able to see so much so quickly.
Basically they use pattern recognition, such as ‘three unmoved pawns in front of a castled king = possible back-rank mate’.
If you show a grandmaster a real game position, he will have it memorised in a few seconds. :eek:
When he reconstructs it, he will put a group of pieces on together (= pattern), then do another group. (For example ‘White King g1, Rook f1, Bishop g2, Knight f3, pawns f2, g3, h2’ is instantly recognisable.)
Now if you put pieces on the board randomly, the GM is :confused: because the usual patterns are absent.

Every experienced player knows the pattern ‘White Rook a7 knight f6 ; Black king h8’. (The Black king is threatened with Rh7 mate, and cannot escape by himself.)
I don’t know of a standard pattern with White knight f8 (how did it get there without being captured?).
So my educated guess is that if the Knight was on f6, Kramnik would have been more alert!