Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / What about a Hybrid Team...!
Poll How much 'Kramanand' Team will score against Rybka in a 20 game match? (Closed)
20 1 5%
between 16 to 19 1 5%
between 12 to 16 2 10%
between 8 to 12 6 30%
They surely will lose! 10 50%
- - By Priyanvada (***) [in] Date 2007-07-12 15:02

             In my view One human person who may be a superGM is always prone to human like mistakes(Remember Mr. Kramnik Allowed mate in One, which is almost never expected from a person like him!) which is normal thing for us, but this does not conclusively tells us that Computer Software has surpassed Human potential in playing chess!

    Consider my following 'Dream case"...!

                What if some day Mr. Kramnik and Mr. Anand make a team and sit together to play against Rybka,
with whatever best hardware configuration available/possible....
      Since Human blunder or fatigue is almost eliminated ...and super-creativity of Vlad and super-accuracy of Vishy are combined This team may represent human potential much much better than any individual human player,...

        So why not carry out a "match" like this...!

                          How much do you think in a 20 game match, Kramanand will score against Rybka!

1. Exact 20
2. between 17 to 20
3. around 15
4. between 12 to 15
5. Around 9 to 12
6. They will Surely loose the Math? (Ohh!...since this is also a possibility I have to mention it)
Parent - By JhorAVi (***) [ph] Date 2007-07-12 15:42
Rybka in an advanced INTEL multiprocessor configuration = 32 CORES working together vs just two GMs :)
Parent - - By lkaufman (*****) Date 2007-07-12 16:29
Two players consulting might not actually play any better than the stronger of the two playing alone, as they will just disagree a lot. At best, they might play 20-25 Elo stronger. So, if we give them 2800 together, and if we modestly give Rybka only 3000 against humans, they should lose at least 15-5.
Parent - - By JhorAVi (***) [ph] Date 2007-07-12 16:47 Edited 2007-07-12 16:54
Since the two GMs will be discussing with each other about their ideas, it will be less ackward if they're also allowed to move the pieces in another chess board. That method can solve their language differences and translates to several ELOs higher..

With regards to disagreement issues, Maybe tossing a coin will solve it :) :)

Another idea is to make only 1 stronger GM as the main contender while the other weaker GM will just help like guadian suggesting ideas and alerts incase a blunder may arise. The stronger GMs move will always prevail so not fall into disagreement problems.
Parent - - By M ANSARI (*****) [kw] Date 2007-07-12 20:00
Kramnik and Anand would be stepping on each other's toes and would probably play worse as a team.  Tactical errors are not usually the problem of super GM's.  Their games are usually based on plans, where tactics do not really come into play.  Choosing the correct plan is totally based on how the GM feels and how much he prefers a position over another.  For example Kasparov would probably choose a position which is dynamic with a lot of room to move around, while Karpov might prefer a solid positions which has no weakness.  You will find that Anand and Kramnik would not always agree on which plan is best.
Parent - By Kapaun (****) [de] Date 2007-07-12 21:49
Yes, they would play worse together, that's my opinion exactly. Too many cooks spoil the broth.
Parent - - By SR (****) [gb] Date 2007-07-13 08:47
The question is meaningless unless you specify the available time. I interpret the question as if the hybrid team essentially have unlimited time. In this case I think they might score 8 or even more points. Its not unrealistic the hybrid team would be able to win at least one game, but in attempting to play for wins the team might loose 2-3 games.
I am not sure they will play better as a team as if they played individually. An efficient approach would be to have each player think say 30 min on a critical continuation (in separate rooms), and then meet and discuss their findings, and then make a decision.

In correspondence chess (where the human abstain from  use of computer) I still think the best humans (corr rating 2700) might be roughly at the same level as Rybka.
Parent - - By billyraybar (***) [us] Date 2007-07-13 11:51
Same level as Rybka?  Slim chance.  Just like a really really smart person couldn't ace one of these tests -- or (made by members of the giga society -, regardless of how long one could ponder it.   
Parent - - By SR (****) [gb] Date 2007-07-13 20:40
I find your response very cryptic. Maybe you could elaborate. I notice that none of the links you provide has anything to do with chess.
Parent - By billyraybar (***) [us] Date 2007-07-13 21:09
I do not believe unaided humans would be play at the same level as unaided computers in correspondance chess.  I agree the the gap will tighten in correspondance games.  I just have to believe that computers are far better at tactics that humans are are strategy.  Plus, chess is 90% tactics, right?  ;)
Parent - - By lkaufman (*****) Date 2007-07-13 22:23
I believe that the best human players, say Kramnik or Anand, would indeed have very good chances in a match with Rybka if both sides took several days per move (unaided) and the human player could truly devote several hours a day to this. But of course this won't happen, and as far as I know the top actual correspondence players are at best on a par with ordinary grandmasters in their chess skills, so I doubt that these players could do the same.
Parent - - By SR (****) [gb] Date 2007-07-13 22:43
Yes, none of the top corr players (e.g. former corr. world champions etc) are super GMs and some (e.g. Jorgen Sloth) are even just average IMs. In my judgement the standard of world championship correspondence is in general VERY high. Of course super GMs are in a different league over the board, but top corr players often analyze positions to great depths and objectively speaking I think the standard of a world champion corr game is somewhat higher than the standard of a world championship over the board game.

But, yes the over the board champions are usually way more talented than the best corr players.  
Parent - - By Uri Blass (*****) [il] Date 2007-07-13 23:00 Edited 2007-07-13 23:15
Jorgen Sloth is not a player that I can consider as  a top correspondence player(edit:the name is slightly different in the table but I could not find a correspondence player with the exact name Jorgen).

He scored only 7/16 in the world championship based on the following link

His iccf rating now is only 2532 and my guess
is that most correspondence players with ICCF rating above 2600 have fide rating that is below 2200 or do not have fide rating.

In my judgement the standard of world championship correspondence is not high and most of the top players simply do not play there.
The level of the game is of course higher than OTB chess but the main reason is the fact that humans use engines to help them.

Note that the only tournament of GM sloth now is
100 Years Danish Chess Federation A (started at 6/1/2004)  and he scored 2.5/12 in that tournament
He started that tournament with rating of 2576 so even before this tournament he was not what I consider as a top correspondence player.

I have iccf rating of more than 2600 and I do not consider myself to be a top correspondence player.

Parent - By SR (****) [gb] Date 2007-07-14 00:09 Edited 2007-07-14 00:13
I mentioned Jorgen Sloth since he has been WORLD Corr champion, and my point was that though he is not such a strong player over the board he was strong enough to win the world  championship (before most people began to get Computer  help). 

I suppose he refrain from using computer and probably are less serious than he was when he won the 1975-1980 world championship. He might not be a top player today, but he certainly was 30 years ago in the mid 1970s. The 2600 iccf rating you talk about sounds somewhat inflated.
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-07-14 07:05
I wonder what sort of scaling Kramanand would achieve. Probably, it would require some parallelization work :)

Up Topic Rybka Support & Discussion / Rybka Discussion / What about a Hybrid Team...!

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill