Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Some conclusions after the draw odds match
- - By Roland Rösler (****) Date 2008-01-08 00:25 Edited 2008-01-08 00:32
Rybka wins the match 6-2 and I don´t expect this result! I thought Joel would win. So why he doesn´t? First I have to say that Joel makes a very good job (okay, loss in game 1 wasn´t necessary). I see 3 determinates for his loss:
1. Openings are more important than I thought.
2. Time pressure for the human was enormous. This correlates with point 3.
3. Chess is too complex for humans. Joel was outplayed by moves which were out of his horizont (but no real blunder!). This is interessting because years ago human can outplay the engines because of their horizont effect.

For further draw oods matches I will suggest the following changes (I know, there is nothing planned like this):
1. Human can see the opening moves (like in the Kramnik vs. Fritz games). We are not interessted in opening batlles but in Rybka skills. So the Rybka opening book will be short and fine.
2. TC in the next matches against humans should be 120' + 120". This is only in favour for humans (Rybka takes her time basically in the second and fourth game, where nothing was to count). And only one game per day and only 6 games in a row and then a rest day for the human if the match has to go on.
3. No changes in opening book and engine during the game! This isn´t fair.

I´m sure, these rules doesn´t really help in the long run (less than 2 years). All these callings for Kramnik or others is in my mind only nonsense and expensive. Maybe Kramnik makes the difference in this match. But I´m sure, Rybka can play against him in two years with no book (and 32 or 64 cores) and he can´t draw the match (I think, he can´t draw one game).
What we have seen in the draw odds match is, that ~ 3000 Elo for Rybka isn´t an illusion. And this was to prove.
Parent - - By Banned for Life (Gold) Date 2008-01-08 00:47
I would like to see a draw odds match with alternating color where Rybka played without a book. To prevent repeated games, something could be added to randomly perturb the evals during the initial N moves.

As far as Kramnik is concerned, I think he is exceptionally talented when it comes to drawing games and I think he would have won this draw odds match in a convincing manner.

Parent - - By dareapa (**) Date 2008-01-08 00:55
What ever happened to the idea of checkmating ones opponent?? Isn't that how we win at chess?/
Parent - - By Banned for Life (Gold) Date 2008-01-08 01:10
Normally you win at chess by being ahead at the end of a tournament or a match. This frequently includes strategies such as getting draws with black, or drawing later games when this is good enough to win the tournament or the match. Many world championship matches had the champion maintaining their title if the match ended up drawn. Playing all draws would be one way for the champion to keep his title under those conditions (of course he might not be a very popular champion :-)).

Parent - By dareapa (**) Date 2008-01-08 01:15
Because I am weak at chess relatively speaking, I do not have the ability to draw whenever I feel like it so draws in my case is by accident. When I push a pawn on an opponent it is to try to checkmate him or make him resign no matter whether I'm the black or white pieces.
Parent - - By Roland Rösler (****) Date 2008-01-08 01:42
You can see it in my post, that I also think, that Kramnik (or even Leko) make the diifference in this match. But this are only (expensive!) snap shots! And I don´t believe in your convincing manner. I´m sure, that Kramnik would made a convincing 6-0 win against Deep Fritz 10 in a draw odds match 14 month ago, when he lost 4-2 against Fritz (where Fritz was playing on a quadcore system 32-bit  but only uses 2 cores).
I would gladly see real attackers like Topalov, Moro or Shirov against Rybka. Win or die! Come trough or lose! Not ever this humility behaviour! Kill her!! I´m rather sure, it doesn´t work, but we never see it by very good players (okay, Jaan can make it in pawn odds match). You take the initiative and losses on the way are irrelevant. You have to mate!
I think, without book and 32 cores there will be no same games! If yes, you can make 4-5 ply opening book for draw odds matches with white or black.
Parent - - By turbojuice1122 (Gold) Date 2008-01-08 01:56
Deep Fritz 10 had to have used all four cores in that Kramnik match because it was getting nps counts of about 9 Mnps, which dual-core systems simply couldn't do at the time.  However, it had roughly the same playing strength as would Deep Fritz 10 on a single-core system due to a bug in the program that screwed up actual playing strength on quad cores.  The Deep Fritz 10.1 update rectified this mistake, and the updated version is about 75 elo stronger than the version that played against Kramnik; the update is irrelevant for single and dual-core systems.
Parent - By Roland Rösler (****) Date 2008-01-08 02:24
Thanks! Kramnik lost 4-2 vs. (Deep :-() single! 32-bit Fritz 10. Not so bad for Fritz. Fritz 10 Elo performance >2850!
Also much thanks to you by your attempt to focus the 32-bit single core playing strength (in other posts). This is really important!
Parent - - By Banned for Life (Gold) Date 2008-01-08 01:59
As you suggest, the flip side of the coin would be to disregard draws and focus only on, say,  how many wins a GM could get before Rybka would get 6 wins in a match. Even in this type of match, I'm not sure that the best strategy for the GM wouldn't be to play for a draw and look to take advantage of any unforeseen opportunities. I suspect that a Topalov or a Shirov would have have their attacks refuted one after another and would end up with nothing.

Parent - By Roland Rösler (****) Date 2008-01-08 02:42
I see, we agreed :-). Topalov always with white and he has eight attempts to win at least one game (and he can look to Larrys opening book)! How are the bets? Better than for Kramnik, who has to make 4 draws (with look to the opening book)? I´m not sure. We can make a vote :-) if it´s interesting, because we will never see such games.
Parent - - By lkaufman (*****) Date 2008-01-08 04:40
     I agree with most of your comments, but I don't understand why you feel that changes in the opening book during the match are not fair. Without that right, we would be helpless if Joel learned that a line we chose was faulty and he could just repeat his play. Joel also has the right to use his own Rybka between games to find improvements on any line played. Since Joel is a stronger player than I am, if anything he should benefit more than we do from the experience of the match (talking only about openings).
     In the only repeated variation of the match, I changed a move (7...c5) to another one (7...Re8) that I had intended to put in the book before the match but never got around to doing so.
     You are right that openings proved very important, not in terms of who knew them better or of getting good positions as Black, but in terms of which side achieved the type of position that favored him.
Parent - - By turbojuice1122 (Gold) Date 2008-01-08 04:58
There's a bit of an argument ongoing in the Hiarcs forum about how "major" were the changes in Rybka during the match?  Were they major enough that Joel was basically playing a very different version in different games and would have been better off with a different sort of thought process or style than he would have been in other games due to these changes?
Parent - - By mjlef (***) Date 2008-01-08 10:21
But people change too.  Joel was not the same Joel after each game, since he learned too.
Parent - - By JhorAVi (***) Date 2008-01-08 11:56
It's the opposite..  Joel was getting tired and tired after each game.
Parent - By turbojuice1122 (Gold) Date 2008-01-08 15:09
I think that actually we're dealing with both factors here.
Parent - By lkaufman (*****) Date 2008-01-08 15:06
No, the changes were minor except for the contempt setting, which I increased from 50 to 60 to 75 during the match. But this would not affect how Joel should play. It's just like in a normal human match to a set number of games, as the side who is losing becomes increasingly desperate to win with each draw. In this case the change was not motivated by the results so much as by simply getting a feel for how high contempt could go without causing weird play. In other words, if I had time for a private test match with someone first I would probably have settled on one value for the whole match.
Parent - By Vasik Rajlich (Silver) Date 2008-01-08 14:55
There are all sorts of possibilities. Another approach is some sort of centaur-type play. Humans are still superior to computers at many things.

Up Topic Rybka Support & Discussion / Rybka Discussion / Some conclusions after the draw odds match

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill