i already use 12 cpu like a number of others and have found better moves than R4.
> If this is correct, than if the cloud can surpass your system by at least 150 Elo by virtue of improved software and clustered hardware
Yes, but you're talking about the whole cloud, that one which Vas thinks to charge 100 euro/hour per usage. I really doubt the "more often than not" will be worth more than human's imagination, what I haven't seen is the cluster finding a move that humans with interaction couldn't find, though that is a challenge Lukas would need to set up.
Of course we don't really know whether Cluster Rybka on Lukas' hardware is 150 Elo better than R4 running on your new machine (but Lukas should know).
> I think that 150 is probably a reasonable guess.
Seeing the moves that the Cluster has been able to find very fast, yes, seeing the draws at the WCCC, I have my doubts.
I very much doubt anyone is going to buy the whole cloud to play at any point, it's just too pricey, so this discussion is probably moot.
>I haven't seen is the cluster finding a move that humans with interaction couldn't find
You should compare to a cluster with human interaction.
> I've always wondered how many IDeA cores would equal the current cluster. Maybe double the number of cores in the cluster?
That sounds like a pretty good guesstimate. With that many engines running you'd probably have to analyze a half a dozen projects at a time just to operate efficiently. There's bound to be some kind of performance hit when you're overseeing four-hundred engines (e.g. managing the queue, writing results to disk etc.)
what is more interesting is that i have noticed that more computing power means that corr players have started to play more and more simultanious games.
i see mistakes being made because of it. too much depency on the computer, too little own AN.
(unless maybe if you talk about the level of the correspondence world champion).
I believe that human imagination is overestimated and it is almost impossible to beat rybka4 or stockfish1.9 with 4 cores at long time control(there may be few GM's who can do it but usually even people who play slightly better than rybka win games at the high level because the opponent plays weaker than the moves of rybka at long time control).
My experience in correspondence games that started in 2008 (against 26 opponents with average rating clearly above 2500 ICCF rating) is that I do not lose a single game but have problem to win even against opponents that are weaker than rybka and the common case is that I get a small advantage both with white and even with black in some cases when the advantage is not enough to win(I got +3 =17 in these 26 games but I am not sure if I can win another one(and I think that I am going to offer a draw in some unfinished games) when I am practically sure that I do not lose).
Note that I do not use a good hardware and I could probably play better with better hardware.
I also did not use a lot of time and I probably perform slightly worse than previous tournaments also because of playing 2 tournaments at the same time that is something that I never did earlier.
Here is my best game when I won thanks to rybka(note that 10.Qd3 is suggested by Rybka3 based on my memory)
32.Bh5 is not the choice of the computer but I suspect that I could win also with 32.Bh3 and I simply did not like opposite color bishops.
I believe that the opponent played weaker than rybka.
1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be2 e5 7.Nb3 Be7 8.O-O O-O 9.Be3 Be6 10.Qd3 Nbd7 11.Nd5 Bxd5 12.exd5 Nc5 13.Nxc5 dxc5 14.c4 e4 15.Qc2 Re8 16.Rae1 Bd6 17.f4 b5 18.b3 Qd7 19.Bc1 Re7 20.Bd1 Qf5 21.Re3 Rae8 22.Bb2 h6 23.Bc3 Rb8 24.Rh3 Nh7 25.Rh4 bxc4 26.bxc4 Nf8 27.Bg4 Qg6 28.Rh3 f6 29.Qe2 Qh7 30.Rg3 Kh8 31.Re3 f5 32.Bh5 Ng6 33.g3 Reb7 34.Qg2 Rb6 35.Kh1 a5 36.Ree1 Be7 37.Qh3 Nf8 38.Be5 Bd6 39.g4 Nd7 40.gxf5 Bxe5 41.fxe5 Nxe5 42.Rxe4 Nd7 43.Re8+ Rxe8 44.Bxe8 Nf6 45.Bf7 Rb7 46.Be6 h5 47.Qe3 Qh6 48.Qxh6+ gxh6 49.d6 Rb8 50.Rg1 Kh7 51.Rg6 Ng4 52.Rg8 Rb1+ 53.Kg2 Rb2+ 54.Kf3 Rd2 55.Rd8 Ne5+ 56.Ke3 Rd4 57.d7
i know you have not got big hardware and how you achieved your ratings is way beyond me(ie you are a much better OTB player than myself)
best regards Paul.
i only made this thread as i was a bit bored the other night.
If you are bored, we can play one game, without time control, play with freedom.
I can see how National's suggestion of relying more on human ideas in the opening could bring better results.
>I can see how National's suggestion of relying more on human ideas in the opening could bring better result.
I don't purpose to suggest that I in anyway am knowledgeable regarding this subject , but I would think that relying (using the word "more" is pejorative, in that more for someone-could be less for another.)on one's innate talents would depend on ones innate talents. With that said, using your own initiative in the opening is a coin flip at best and a disaster at worst.
But, let me add that that is the fun of the game.
when +1 =11 is enough for the GM norm.
I believe that there are few games that I missed something in the middle game or in the endgame that maybe better hardware could find.
Another point is that I do not know to do a smart choice of the opening to play for a win and that I am not very interested in correspondence chess.
games when I maybe missed a win in the europe championship(you are invited to use a long computer analysis to check it)
> I am not very interested in correspondence chess.
This was the most unexpected comment from you, considering all you've done.
Still, I believe you could perform better by repeating the methods that brought you to a 2600 rating. You might think it's following rybka blindly but I doubt it. You obviously have natural talent.
I believe that correspondence chess today is basically games of computers when humans have very little value(I believe that in more than 80% of the cases a team of human and computer is not going to beat Rybka4 with no help)
I believe that today rybka4 on good hardware without help can be a GM at correspondence chess.
I think OTB chess is more interesting and I consider getting fide rating of 2051 as better achievement then getting ICCF rating above 2600.
There is still plenty of evidence of human influence in correspondence chess.
I know of one correspondence GM who used only Genius of richard Lang for years and only one year ago started to use rybka(after he lost some games in a tournament and he was still above 2500 even after the bad tournament).
My rating in 1984 was approximately de1800, so I started playing in the
weakest categories. I got to play a group of semi final of the Brazilian
championship in 1993 and got a master norm. I was wrong at work and in my
personal life and decided to quit. I gave all my chess books and thought
I would never get interested in chess.
Today I play on Playchess, which already affect my personal life and work,
because I like to improve my lines with the help of engines, and this costs
time. Let's see if I can find a way to play without sacrificing my personal
life and my work.
I think somebody whose name I can not remember, has already talked about it here.
Paulo Ferreira Soares
years ago: "If you know an excellent lawyer, teaches chess for him and you will see that he is
no longer excellent."
But you're right. The hard part is knowing when to stop playing and go to your normal life.
maybe i am being stupid here,but around this position engines want to play Bg5 as you will have noticed Mark.they dont seem to understand that i want to play on the Queenside against the castled king.thats why i have realined my troops.why should i give black a free move to put his QR in a better position.maybe they are right after all i am a mere human.
You actually mention one of these cases, where you avoid one move that leads to opposite colored bishops. I have a position in one of my current cc games where I think R4 is not correctly evaluating king safety for my opponent. These situations are not rare and seem to me to call for human guidance...
I don't think it's any surprise that the leaders have a more aggressive black repertoire.
> what is, in your opinion, an agressive black repertoire for corr chess?
If white is determined to draw, there's not much that black can do about it. Having said that, I think it's possible for black to increase his chances of reaching the middlegame with more than a "puncher's chance" at winning. For example, against 1.d4 the best way for black to play a sharp, full-contact game is with the Triangle system (pawns on e6 d5 c6 before Nf6) which has the potential to transpose into a number of combative lines like the Abrahams, Botvinnik Semi-Slav, Anti-Moscow Gambit, the Marshall Gambit and a half dozen other lines where the e4 advance results in pandemonium. About half of the time you'll end up in a boring Qc2 Semi-Slav of some sort but if I can reach a fighting middlegame against 1.d4 half of the time I consider that a great success.
From black's perspective, the beauty of the Triangle system is that it often transposes into a variation with a reputation that's based almost entirely on OTB praxis. If the engine room's Najdorf fetish has taught us anything it's that OTB theory in sharp positions is ripe for exploitation. For obvious reasons, I don't want to get too specific but I will say that I think correspondence players would do well to seek out openings where that tension exists, where there's an incongruity between OTB consensus and correspondence results.
I am still looking for something of the same nature against 1.e4, other than classical playchess lines...
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill