< Earlier Kibitzing · PAGE 54 OF 79 ·
|Mar-22-09|| ||rogge: Norway is the capital of Stockholm, I heard.|
|Mar-22-09|| ||Nimzo64: No it does not, the spelling is Carlsen.|
|Mar-22-09|| ||crwynn: I think Acirce was perpetrating some oblique form of humour so perhaps his spelling there is not to be taken so seriously....|
|Mar-22-09|| ||Nimzo64: Norwegians do not have a sence of humour|
|Mar-22-09|| ||sheaf: chiezze hab a cence of humeuaurx...;-)|
|Mar-22-09|| ||zarg: <arnaud1959>
he made a joke. :) <acirce> know perfectly well the correct spelling, but choose the Swedish one on purpose...
We have a long traditions in making mutual jokes, but are good neighbours.
|Mar-22-09|| ||rogge: Carlsen, Carlssen, Carlson, Carlsson, Carlzon, Karlsen, Karlssen...|
|Mar-22-09|| ||acirce: And then there's User: karlzen (where did he go btw? he was a great poster)|
|Mar-22-09|| ||Appaz: Maybe Pontus Carlsson won a game today?|
That would make <acirce>'s statement valid.
|Mar-22-09|| ||DeepTrouble: Achieve:
<I said close to useless, but of course it has a relative usefullness, when you have the TBs at your disposal. Depending on the position at hand.>
There are open endgames which Rybka 3 handles well without TB-support. The point is that if Rybka can see far enough (and it often can in simple, open endgames), it doesn't need specific knowledge or TBs to play good moves.
<However, <rybka 3 <<< generally >>> handles endgames better than the 2700+ players Carlsen mostly plays against> is a bit of a stretch to me.>
It may sound a bit farfetched, but this view is actually based on my own experiences when analyzing hundreds of games (usually every move after the opening) between 2700+ rated players using Rybka since early 2006. You might be surprised to learn how often even strong GMs make mistakes (or inaccuracies) in the endgame (and even in simpler endgames). I think there are many psychological factors involved here. Time can also be a factor, of course.
<1. I still use a 32 bit 2.1 version of rybka - and it is not up to the standard of the recently released Deep Rybka 3.>
Rybka 3 is significantly stronger than 2.1 (or 2.3.2 for that matter), as shown through independent testing by CEGT, CCRL and SSDF.
<2. Players often play on increment, even in classical games, with just a few minutes on the clock, and are tired and inaccuracies may, and do, frequently occur.>
Yes, but inaccuracies (or outright bad moves) also occur when they're not under time-pressure. One of the computer's strengths is, of course, that it doesn't get tired.
<3. For humans who know the endgame fundamentals, the < strongest move> is the surest and least complicated way to a win>
No, not necessarily. The strongest move could also be the move that leads to most complications. Computers aren't too concerned with tactical or complicated positions, but humans often prefer to simplify things or play simple, but strong moves when they've got the advantage. Magnus often prefers to play practical moves that suits his style, not necessarily the very best move (it can be seen in his games and he's explained this in interviews, after being asked:"Why didn't you play the stronger move X?").
<It even clouds the issue, because the theoretical win that follows from Kxa6 is the safest and most efficient route to a win>
Yes, for a human perhaps. It's probably the reason why Magnus played Kxa6.
<The specific line from this position is quite simple, and if you play it out on ANY ENGINE, incl. Rybka 3, you will see the eval increase to +10 and higher, 6 0r 7 PLY further down "the line." (Kb3)>
The same also goes for b4 of course (I've played through it's PV). And the reason why b4 has a higher eval is that Rybka sees that this move will lead to a bigger advantage in a shorter amount of time. But like I said: Both moves are good. It's just that Kxa6 appears simpler and more practical for a human. But the issue here is whether it was the strongest move, not the most practical, and Rybka's evaluation and PV indicates that b4 gives black a bigger advantage in a shorter amount of time.
END OF PART 1/2 (I had to split up my message; it was too long)
|Mar-22-09|| ||DeepTrouble: CONTINUED:
<4. Terms like <objectively strongest> should be put in perspective, re the OTB circumstances the players play in: even IF the engine finds a so called "better move", it may prove a more risky one, and as a player you want to avoid such lines, and therefore it is not better per se.>
Objectively, it is. That's the difference between a subjectively strong move, and an objectively strong move :)
<5. ... Even as a dry analyst, with your TBs on your lap and a state of the art calculating beast, evals are merely floating on bobbing waves, heavily changing with each passing (4-6) PLY. This often becomes more apparent when approaching certain endgames like the one in Yue-Carlsen.>
I don't simply look at evals, but on the suggested main lines as well. It's natural and perfectly logical that evals increase as you get closer to a clearly won position. If the evals go up and down, it's usually because the players don't follow the suggested moves, not because the computer changes its mind constantly.
<Of course <rybka is <<< NOT >>> close to useless>, in many type endgames, but then again in many it *is*. (I see that <sheaf> has provided some material.)>
I've seen endgames (typically closed) in which even tablebases won't help much, but this is a rare occurence (you can also find examples on the Rybka forum). In practical play, however, I very rarely see Rybka suggesting bad moves when using TBs.
<If I had the time I would post some examples, but then bare in mind I only operate the inferior, dressed down 2.1 version of <Rybka>, so I can not pass specific judgement on Deep Rybka 3 on a Quad core.>
I don't use a quad core, I use an octocore :) (8 cores). I also use the 64-bit version, but this only means that Rybka searches deeper in a much shorter amount of time compared to Rybka 3 (single CPU 32-bit version). It doesn't affect its evaluation. However, the multiprocessor version does sometimes produce other PVs than the single CPU-version, and that's because the parallelization code isn't entirely deterministic.
<Keep <interrogating> the machines, at ANY stage of the game, is my "motto.">
Certainly :) That's part of the fun. Simply looking at evals without playing through the main lines and trying to understand them isn't that interesting.
Anyway, I don't want to start a long discussion about this. I don't post on this forum very often (on the contrary), so I might not be able to follow up this discussion anyway.
END OF PART 2/2
|Mar-22-09|| ||Andrijadj: Ivanchuk-Morozevich was a phenomenal game...|
|Mar-22-09|| ||DeepTrouble: zanshin:
<You are 100% correct and your explanation was perfect. I thought about addressing this point, but I'm glad I didn't because your reply was better than anything I had in mind. In fact, I'm going to copy your post to my forum for future reference.>
Well, I'm glad you found it useful :)
I apologize for posting such a long reply (I'm talking about my last reply, which I had to split in two), but the blindfold-games have ended so I hope the long posts don't look too 'disruptive'.
|Mar-22-09|| ||Bears092: Could someone help me with how these blindfold games work?|
Each player has a laptop with a blank board and a list of the moves played?
Also, how does the strength of the top players change when they are playing blindfold? Would a 2700 player blind be more or less even with say a 2200 player? Or what would an approximation be?
|Mar-22-09|| ||Jim Bartle: I've read the top players' strength drops about 100 points in blindfold. |
And yes, each player has a laptop with a blank board, but I think only the last move appears in writing.
|Mar-22-09|| ||SetNoEscapeOn: <Bears092>
I don't think that they get a list of the moves. The last move simply appears on the screen for a while and then disappears. The notation for the last move may stay on the screen as well.
I don't think the way their strength changes can be described in such a simple fashion. They play like 2700 players who blunder more often than usual. Sometimes they don't blunder at all and produce truly great games.
|Mar-22-09|| ||percyblakeney: Carlsen is doing surprisingly well in blindfold, 3000+ performance with three rounds to go and the only remaining game that looks really difficult on paper is the one against Kramnik. The latter has been in trouble in his black blindfold games though, saving the draw against Topalov and Kamsky but losing against Anand. Of Carlsen's other opponents Karjakin is last in the blindfold section while Radjabov isn't doing too well either. An even score in these three games should be enough to win the blindfold half.|
|Mar-22-09|| ||achieve: <DeepTrouble> Thank you for taking the time (trouble ;)) - to address further several elements of our discussion, which has been highly appreciated.|
You indeed post rarely, but when you do you go about your business very thoroughly, and expertly.
What I enjoy mostly when all is being said and done, is the process of:
being CHALLENGED to IMPROVE ON my overall- and specific- understanding on a subject like this;
<computer's relative strengths and weaknesses, and how to optimize my understanding- and ways of using- the information it provides.>
|Mar-22-09|| ||kamalakanta: <Nimzo64: Norwegians do not have a sense of humour>|
You must be kidding!
|Mar-23-09|| ||alexmagnus: @Carlsen-Carlssn-Carlssen etc.: don't forget this "hybrid": Magnus Carlhammar|
|Mar-23-09|| ||chessgames.com: Thanks to everybody for participating in today's live broadcast from the 18th annual Amber tournament. Coming up in just a few minutes are the rapid games which we will be broadcasting on our Amber Tournament (Rapid) (2009) page.|
|Mar-23-09|| ||Davolni: <Mar-23-09
chessgames.com: Thanks to everybody for participating in today's live broadcast from the 17th annual Amber tournament. Coming up in just a few minutes are the rapid games which we will be broadcasting on our Amber Tournament (Rapid) (2009) page. >
isn't today a rest day??????
what's this about?
|Mar-23-09|| ||alexmagnus: <what's this about?>|
The message was posted tomorrow but went through a timehole ;)
|Mar-23-09|| ||strifeknot: What games, <chessgames.com>? Isn't today a rest day?|
|Mar-23-09|| ||Pawnsgambit: <chessgames.com> Thanks, I am waiting for the secret blindfold and rapid match that is taking place between Kramnik and Topalov. are the games going to start at midnight?? It is already 10:30pm in France.|
< Earlier Kibitzing · PAGE 54 OF 79 ·