It's the home stretch for the candidates and several have their backs against the wall. Bareev is down two against Leko and has looked entirely hopeless. Maybe he doesn't like Mexican food? He won't have to eat any if he loses game four today. Rublevsky has been hanging on by his toenails against Grischuk and will need to show something with white today. I'm still in shock that Grischuk let him off the hook the other day. I'd just been saying how good he was at finishing off attacks. Kamsky has played very tame stuff against Gelfand so far and is down one with two blacks in the final three games. He'll have to force the action now that he's behind. If he can't get his clock under control it won't matter what he plays. (He had around three minutes for the last 20 moves of game three.) Shirov is also down one, against Aronian, but every game has been a slugfest and he'll has white in game six.
You can listen to a free game of the day from the candidates over at Chess.FM. ChessBase has round analysis from GM Marin. Fabulous Freddy is in Elista, so onsite reporting should improve. Higher-res pictures would be nice, eh? The world has moved on from 450-pixel shots that are compressed to death. Servers are faster, transfer capacity is greater, and we're on broadband. Let's see some big pics! The official site is cranking out all sorts of analysis, but isn't much on the interviews and press conferences.
With a draw in game 5, Deep Junior just won the match against Deep Fritz after wins in game 3 and 4. I've never understood why official/important computer events use different hardware. It's bad enough with all the work the "trainers" put into the books. When one machine is running twice as many processors as the other, what's the point? They should all play on a standard platform known well in advance. I know that would be annoying for developers to an extent, but they don't seem to mind grabbing the fastest hardware they can at the last minute. Anyway, it was pretty cool to see them play the triple poisoned pawn sac Najdorf line that's been hot lately. I bet lots of GMs will be giving that one a look.
Didn't have time for much chess since game three. There was a Dissenters' March in St. Petersburg on Saturday and another is tomorrow in Moscow. There were other goings-on on the homefront as well. I've finally added to the list of things I don't have in common with Pushkin, Rimbaud, and Byron. Whew!
Fritz plays d4, Junior plays e4. Go figure why they use different books.
"I've never understood why official/important computer events use different hardware.[...]When one machine is running twice as many processors as the other, what's the point?"
I can give you the standard counter-argument: There IS no such thing as equal hardware -- or, equivalently, there is no such thing as "pure" engine strength. Two examples:
1) Anthony Cozzie, the author of Zappa, spent a lot of his development time improving the "scaling" as more cores are added. This means that his engine benefits relative more than other engines when more cores are added. Now, how do we decide which "standard" hardware to use in a tournament? Anthony would find it unfair if only 4-core systems were used. It is not HIS fault if other engine authors decided to spend less development time on MP scaling. It is not unreasonable to speculate that the Fritz team would ALSO have used a 16 core system vs. Junior, IF ONLY THEIR PROGRAM COULD HANDLE IT.
Perhaps one could suggest that if a team wants to use hardware X in a match, they must provide an additional X to the opponent team, so that they can use it if they like. But then we have the other problem:
2) What about Hydra? Should Hydra be disallowed from computer tournaments/matches because it uses specialised hardware?
The "point" of these matches is not to decide who is "objectively" stronger. Just as it is not the point of a human WC match to find out who has the highest Elo. We have rating lists to answer that question. The point is to find a winner in a concrete, high-profile event where both teams/players are allowed to bring everything to the table.
This is also related to the opening book issue. It is funny that you applaud the nice game in the topical PP Najdorf when one considers that we would not SEE such a game if there were no or only limited books. Well, we discussed the opening book issues at lengths at
http://www.chessvibes.com/?p=900#comments
I can just add a few points:
1) Strong opening books are important for performance, but they simply cannot decide a match in favor of a much weaker engine. That is, they don't "cancel" the engine battle.
2) It would have been absurd to insist that Deep Blue should play without book vs. Kasparov. Opening books are simply a part of the "chess machine" equation.
3) Opening books are used to steer the engine into positions it plays well. Just as in human chess, performance is measured in exactly those openings that the player chooses to play, and knowing what traps to avoid in order to get there is an essential part of the "chess equation".
Essentially, matches with unlimited hardware and opening books ensure that we get to see state-of-the-art machine chess. It's not flawless, it's not "fair", but it's the best that mankind is able to bring to the table at the moment of the event. THAT is exactly why these matches are interesting.
[Sure, in order to see the "unequal hardware" criticism repeated endlessly, I agree that it would have been a pragmatic decision to use equal hardware in the Junior-Fritz match. I am just trying to point out why this wouldn't necessarily have been more "fair".]
Happy birthday then! (must be yesterday)
Saludos,
Jan
Happy birthday then! (must be yesterday)
Saludos,
Jan
This was a publicity stunt to advertise Junior--the program that beat the one that vanquished the world champion!
Kamsky burned an hour in the first 10 moves. A repeat of the last game?
Just checking: Gata does know that this isn't a correspondence match, right?
I guess JUNIOR accepted a Game 5 draw to win the match: looks like White wins - any thoughts?
Junior plays the better match. In all games, it has sacrificed material. Pure art.
It is good to see the computers playing some interesting opening lines.
This should at least give the top players something to think about for the future.
Happy b-day, Mig! Go Gelfand!
What is happening in Shirov-Aronian?? I have CoffeeHouse, can't see the comments. Aronian should be able to draw, right?
Aronian's 51...Re2+ instead of Rxf4 aims to keep Shirov's R trapped. Seems a drawing strategy that should work, and Black doesn't seem to have a way to win.
Aronian is two pawns down, but has managed to stalmate the white rook.
Can Shirov get in c4 at any time to start moving his b-passer?? The rook has been trapped more impressively!
Aronian plays like he wants to prove something. Maybe that playing endgames in a tactical way is possible. Not really old Capablanca school.
Aronian's 61 ...Nh6 Interesting. Is he considering ... Rc8 and Rg8 to prepare a breakout? Is there some way to free the K? Ah. 62 ...Nf7. So thinking draw, or a different try?
This must be draw. I don't see how Shirov can make progress. His only chance was c4, but the square is blocked more than ever.
Shiro can play 65.Kb2 and put Levon into Zugzwang, he has to move the rook. After that Alex plays 66.Bc5 and then sacing the pawn with c4. This might work.
Draw!
what a defense by aronian!!! one of the best i have seen by any top flight player.
Surely Shirov should have played c4 while he could. I guess he was trying to get an improved version of it, but that didn't work out very well..
This game needs further analysis I guess. I also do not understand why Shirov didn't play c4. The draw was always in his suitcase.
I thought that was the whole point. Watching it live we were trying to figure out since move 50 how he would get c4 in.
Deep Junior is one of the most aggressive chess playing entities out there. It sure gave up a lot of pawns. Deep Fritz's 5 pawn mass on the Queen side looked quite menacing. The chess games are quite interesting, with both sides relying on seemingly indirect threats to parry the opponent's tactics. These will be a challenge to annotate.
I can see a place for having matches with each engine being allowed to utilize "Unlimited" platforms (the most powerful configuration the programmer/operator can obtain), as well as for having "Absolute" matches or tournaments where the engines all must use the same platform. If you want to go the latter route, you may also want to limit the engines to using the same opening book and endgame tablebase, so that it would be a true test of the strength of the chess engine, and the skill of the programmer. The stock opening book could be released months ahead of the event, giving each programmer a chance to optimize the engine to play within the limitations of that opening book.
One novelty that ought to be considered is having the chess engines play Thematic events, where all of the games would commence from a particularly interesting "Tabiya" (at which point the "book" would be disabled), of a position that is now the subject of Elite GM scrutiny, such as the Triple Poisoned Pawn Sacrifice line in the Najdorf Sicilian. Essentially, you'd be generating an instant, and fairly authoritative, monograph, on a very topical variation.
I wonder how Deep Junior would fare in a Rice Gambit thematic?
The only reason I can think of (why Sirov didn't play c4) is that he was afraid his king would be cut off the third line with Ra3 for example. But I have big doubts this fear was real.
"I can see a place for having matches with each engine being allowed to utilize "Unlimited" platforms (the most powerful configuration the programmer/operator can obtain), as well as for having "Absolute" matches or tournaments where the engines all must use the same platform. If you want to go the latter route, you may also want to limit the engines to using the same opening book and endgame tablebase, so that it would be a true test of the strength of the chess engine, and the skill of the programmer."
DOug, this is what the rating list groups are doing on a daily basis :-)
And this is my point. These matches should be something more than something you can just simulate on any home computer. Otherwise, they are completely redundant.
The thrill in these matches is seeing state-of-the-art computer chess displayed. And all the better that the chess is also interesting by virtue of strong opening books made by experts.
I hope Junior team takes the Rybka challenge, without one-point advantage indeed, but draw odds is ok, if Vasik is so sure about Rybka. I'm sure it would be a very close match.
Latest pics at chessbase quite amusing. Leko, Bareev winning in the fashion stakes - Gelfland and Kamsky nailing the 'other' look...
The official website really sucks. After almost 4 hours since the last game finished, there is not any report posted by the organizers! There is a short and incomplete interview from Kamsky-Gelfand and Leko-Bareev 3rd game and nothing from the other two matches.
"'I've never understood why official/important computer events use different hardware.'
"I can give you the standard counter-argument: there IS no such thing as equal hardware--or, equivalently, there is no such thing as 'pure' engine strength."
I'll have to try this approach when (cross fingers) I get into research and wind up designing and conducting a sloppy and improperly controlled experiment. "I had every reason, sir. I could not be sure what variables beyond my control might affect results so, what the hell, I figured why not throw in one or two variables more? Let's bring everything to the table and let 'er rip. Not flawless, not fair, but interesting, don't you think so? So do I still get my grant money?"
Bringing up Hydra is an interesting dodge as well. "Sir, I didn't think it was worthwhile to conduct a fair and controlled test of A and B, because we couldn't have done the same thing with completely different entity C."
The most impressive game I have ever seen by Aronian sure he was lost on several occasions - but the resistance he displayed smply exhausted Shirov. I have only seen Kasparov display this kind play. Fantastic. I believe only Kramnik has the ability to defeat Aronian in a match at the moment.
Ernest, very funny comment, but you seem to miss the point:
If Zappa and Hiarcs are run on a 1-core system, Hiarcs is stronger.
If they are both run on a 32-core system, Zappa is stronger (I don't even think that Hiarcs can run on a 32-core system).
Zappa would say that it is Hiarc's problem that they didn't do enough to work on MP-efficiency, and would protest any competition that is limited to, say, 2 cores.
Likewise, I have doubts that Deep Fritz 10 can even run 16 cores at once (or, at least, do it effectively), while Junior can. So, it can be said that it was software that made the difference, not hardware.
Anyway, I would not necessarily object if "equal hardware" rules were used. I am just trying to say that "equal hardware" is more equal for some that for others. Heck, there are engines that only work on one core -- does that mean that all competitions should be restricted to one core systems only so that these engines are not "handicapped by inferior hardware"? Or, put another way: Should Junior be restricted to 8 cores just because Fritz cannot handle 16 cores?
I don't know if Deep Fritz can only handle 8 cores, but I find it quite likely.
I found curious that when Deep Junior started defeating Fritz (AND NOT EARLIER), Chessbase began putting the argument that Junior was running in a much better hardware. Lame excuses??
(You assume in a high profile event each participant tries to prepare in the best way
... you don't see very often humans losing a match and starting right away to say: "I was sick")
The only reason to put this in context is just for arguing that perhaps there is another explanation to Junior's victory different to the fact the chess program is just better. Indeed, Junior has proved its superiority in the Computer WCC last year, so I guess this match was one-sided since the beginning.
Considering Carlsen just drew Aronian and Shirov is precisely one game behind in their match, any statement of Levon's match greatness may be a bit...premature and overexaggerated. I think Aronian is terrific player (match or otherwise) but his on-the-board results do not yet suggest placing him in head-above-others category.
"Heck, there are engines that only work on one core -- does that mean that all competitions should be restricted to one core systems only so that these engines are not 'handicapped by inferior hardware'?"
Actually, yes. But not just for that reason. In this case, anyway, it must be remembered that the competitors in this Deep Fritz and Deep Junior aren't the equivalent of (say) an America's Cup yacht, something built exclusively to compete and entirely for some private entity's amusement, like Deep Blue or Hydra. Fritz and Junior are commercially available programs meant for public consumption. They're mean, or supposed to be meant, for any chess-mad schmoe running Windows to spend fifty euro on. Hence both Fritz and Junior *should* be designed to run well on Joe Schmoe's expected hardware, not the overpowered server-class machines that software engineers and other children have wet dreams about.
Of course chess software, being what it is, should take advantage of what power is there, but like any other commercial software with taxing requirements, it should *not* be expected *only* to run well on hardware that almost nobody owns. Its performance should fall off gracefully according to the lesser power of the machine running it. Thus I contend that, because both Fritz and Junior are marketed to people like us, the only fair test should use hardware that people like us are likely to own. The biggest geek I know still is only up to a dual-core machine.
In any case I stick by my earlier contention as well that there's no earthly excuse to throw in more variables into a test than necessary. Unless it really was about nothing more than who could spend the most money, which is why the America's Cup is such a bad joke.
"In any case I stick by my earlier contention as well that there's no earthly excuse to throw in more variables into a test than necessary."
But it's not a test... It's a match...
In human WC matches, we are not aiming to find the highest Elo rated player -- we are aiming to find the WC. In an engine match, you are aiming to find the winner of the match, not aiming at measuring which engine is "objectively" stroger. The testing groups are already doing the latter.
But I am repeating myself...
"In any case I stick by my earlier contention as well that there's no earthly excuse to throw in more variables into a test than necessary. Unless it really was about nothing more than who could spend the most money, which is why the America's Cup is such a bad joke."
Well, all of the Yacht entries spend obscene amounts of money on the design, fabrication, and sailing of their boats. Indeed, the cost to be competitive is about 100 million Euros, minimum. The American entry cost about 250 million $$. However, there is no direct correlation between which team spent the **most** money, and which team will win. These boats still need to be sailed, and the skill of the crew is what provides the winning margin. Nevertheless, it still is a huge, obscene, scandalous waste of money.
"DOug, this is what the rating list groups are doing on a daily basis :-)
And this is my point. These matches should be something more than something you can just simulate on any home computer. Otherwise, they are completely redundant."
These matches would be more than what one could simulate on any home computer. One of the major reasons why official computer competitions garner more attention is that the engines are operated by
their respective teams, in controlled and equitable circumstances, according to rules that are laid out ahead of time. The results of a match between engines that takes place on a home PC might be interesting to the person who takes the time to run the match, but thus far has not earned widespead attention in the chess community. What happens with the rating list groups is a bit opaque.
Congrats at surviving this far!