Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / It's Ehlvest Again!
1 2 3 Previous Next  
- - By lkaufman (*****) Date 2007-06-06 23:31 Edited 2007-06-06 23:35
     I've made my choice for the 11,000 dollar challenge match (see "match offer" for details) in which Rybka will play only Black and without book after move three. It will be GM Ehlvest (again), FIDE rating 2643. He is currently #4 in the U.S. and #70 in the world, and has been as high as World #5 around 1990. Although he lost the previous pawn-handicap match to Rybka, he showed great mental toughness and character in scoring 2 out of the last 3 games in that match despite having already lost the match and all chance for a money prize. He learned a great deal from that match, and will fully understand the difficulty of the current challenge. The dates for the match are set for July 5 thru July 9, with the precise schedule to be decided later. Unless something changes, the match will be played from my home in Potomac, Maryland. Internet coverage will be arranged as soon as is practical. The version used will be 2.3.2 with whatever improvements we can make in the intervening weeks, possibly with some special "anti-human" features added.
    Some people might say, if Ehlvest couldn't win with an extra pawn, how can he win this match with even material? There is some truth to this, and indeed Vas and I expect Rybka to win this match, but here I'd like to make the case that Ehlvest does have some important advantages as compared to the pawn-odds match.
1. He will have twice as much basic time, and three times as much increment time as in that match. This should make a big difference. Rybka will only get half of his time (the previous match was played with equal time).
2. He will get all White pieces, whereas in that match he played all Black pieces, also a big factor at high level.
3. He will be playing a position (the normal opening position) which he has played thousands of times in top level competition for over thirty years, whereas in the previous match he was playing positions he had never before faced.
4. He has the benefit of extensive knowledge of five hundred years of opening theory, while the computer will be totally on her own resources after move three.
5. He has the experience of the first match behind him.

     In view of all of these factors, I think it is at least likely that he will improve on his 31% score in the first match. The big unknown is the three move book. Rybka must either play main lines, hoping to find her way through them without falling into known theoretical traps, or play inferior defenses that will leave White with a substantial advantage right from the start. Either way, it's a big burden.

     Thanks again to the main sponsors (I got the ball rolling, but others put up most of the prize fund). I'm looking forward to an exciting match, and to learning the answer to the question "How well can computers play without opening books?".
Parent - - By turbojuice1122 (Gold) Date 2007-06-06 23:55
This is excellent news--thanks to you and everyone who helped to put this together--it will definitely be very interesting from a fan standpoint.  I wish both sides the best in putting together some really fascinating and high-level chess games.
Parent - - By Uly (Gold) Date 2007-06-07 00:12
Ehlvest Vs. Rybka II.3.2:

This time it's personal!


I'll be ultra interesting to go over the game(s) Ehlvest'll win :)

Also, all my money is in that Rybka will want to play Nc6 as soon as possible regardless of position ;)
Parent - By lkaufman (*****) Date 2007-06-07 04:15
     Just as a curiosity, ten years ago I actually defeated Ehlvest myself in the World Open, at a time when his American rating was #1 at an astronomical 2750! Sure enough, we met twice in subsequent tournaments and naturally he won. Will he also get revenge against my computer for beating him? We shall see!
     Regarding Rybka's tendency to play Nc3 or Nc6 even when it looks inappropriate, I did change some values in 2.3.2 to try to discourage this. She still loves these moves, just not quite as much as before.
Parent - By Roland Rösler (****) Date 2007-06-07 01:20
Oh, I like this choice very much! You have elected the very best. And this second chance for Jaan is very dangerous for Rybka. Have a look to your bank account. It will decrease!
Parent - By Lee Ma Hong (**) Date 2007-06-07 03:48
2. He will get all White pieces, whereas in that match he played all Black pieces, also a big factor at high level.
3. He will be playing a position (the normal opening position) which he has played thousands of times in top level competition for over thirty years, whereas in the previous match he was playing positions he had never before faced.
4. He has the benefit of extensive knowledge of five hundred years of opening theory, while the computer will be totally on her own resources after move three.

I am not so sure this will be a big factor. I play correspondence chess, and some time back when I started play in a tournament, I set Rybka on infinite anlysis during the opening phase just to see what it would come up with. Amazingly, it found the main lines.

IMO the only way Ehlvest can make something out of the opening is: (1) if he elects closed positions and play along purely strategical lines with minimal piece play for black,  or (2) he opts for variations that simplify the position so the endgame is reached as soon as possible

Ultimately the experience will further benefit Rybka, as apparent weaknesses its play will be re-programmed and improved, hopefuly incorporated in version 3.
Parent - By ozziejoe (**) Date 2007-06-07 08:17
that's great. I think Ehlvest will definitely have a fighting chance, if he can get rybka into some deep strategic openings. rybka might go a bit astray without an opening book.
Parent - - By grolich (***) Date 2007-06-07 09:22
According to your tests, how much does the lack of an opening book affects an engine's rating?
Parent - - By Carl Bicknell (*****) Date 2007-06-07 10:12
I'm looking forward to this because it will be "real chess", but, seriously, he stands little chance and most of the "advantages" listed above wont help him that much.

muhahahaha...death to the human
Parent - By lkaufman (*****) Date 2007-06-07 13:26
I am not sure why you put "advantages" in quotes; they are very real, and would surely make Ehlvest a huge favorite against any human opponent (I don't know how we would disable the human's book after move 3 though). But maybe you just mean that they aren't enough to offset 450 Elo points.
Parent - By lkaufman (*****) Date 2007-06-07 13:22
I haven't run this exact test (maybe I will), because we are using a 3 move book, not no book, but I can say that the damage should not exceed one hundred Elo, because we can always open with a "pass" move like 1...h6 to avoid theory, and the value of a tempo in the opening is around 80 Elo at top level (the difference between +40 as White and -40 as Black). Assuming we try to do better than this, we should limit the damage to less than this, perhaps 60 or so. Of course this is on top of the 40 point handicap for playing Black. However playing without a book is worse as Black than as White; the combined damage may well be more than just a hundred points. It is quite likely that Rybka will find herself in theoretically losing positions after a few moves in many games. Finally there is the practical aspect that Rybka will be using time in the opening while Ehlvest may be playing moves from memory.
Parent - By h1a8 (***) Date 2007-06-09 22:42
I have very good reason to believe that an engine without opening book vs. another engine with opening book is at a significant disadvantage only because it will have less time to think. That is to say, the engine with opening book in many cases would start to think after the 7-10th move which causes a tremendous time disadvantage (especially in blitz).
And for those who say that the disadvantage lies more in the weaker opening lines chosen without the opening book instead I will have to disagree (even though it might be at a very slight disadvantage here).
When rybka 1.01 beta 7(I think) first came out I made an opening book out of her natural opening moves ( or preprogrammed moves that she would do anyway without an opening book for the saving of thinking time). The book was about 7-10 moves long (I don't quite remember) to even up the thinking time between both engines. I mainly did this because rybka didn't have her own book at the time so this was a way to give she her own book (a natural one). Suprisingly, rybka scored significantly better (probably 20-30% better) with this book than without (even though she killed everything in sight without an opening book by still a big amount).
Parent - By lkaufman (*****) Date 2007-06-09 23:45
Now I have run this test, by running a hundred blitz games between 2.3.1 and 2.3.2 with 2.3.1 using various commercial books and 2.3.2 using no book. The result was exactly tied, fifty-fifty. Since the test results so far put 2.3.2 about 35 Elo ahead of 2.3.1 in blitz, that would be my best estimate of the value of an opening book in blitz play between computers. In slow play, it should be more, and with a human involved, much more still, since humans understand "similar" positions and computers don't (yet).
Parent - - By Vasik Rajlich (Silver) Date 2007-06-07 10:22
Thanks Larry for organizing this. Thanks also to the sponsors, and to GM Ehlvest. I'm looking forward :)

Parent - By Graham Banks (*****) Date 2007-06-09 05:48
Yeah - me too.
Don't worry about all the detractors Larry. I suspect there would have been moans and groans no matter who was chosen, so you were on a hiding to nothing really.

Regards, Graham.
Parent - - By Uri Blass (*****) Date 2007-06-07 11:28 Edited 2007-06-07 11:31
Can you give us some details about number of players who accepted the challange?

I understood that except ehlvest you had 2 GM's with the rating 2550-2600 who also accepted.

Did you have more GM's who accepted the challange and do you expect another GM to play in case that Ehlvest lose the match
and in that case for what prize?

Did you ask the sponsors about it?
I think that it may be a good idea to have a plan to continue these matches if possible because even if ehlvest is going to lose it does not mean that other GM's(even if their rating is only 2550-2600) cannot do better.

Parent - - By Carl Bicknell (*****) Date 2007-06-07 12:41
Kramnik. That's who you need. Find out how much money he wants to get beaten by Rybka.
Parent - - By Harvey Williamson (*****) Date 2007-06-08 22:39
You can start the bidding at $500,000
Parent - By Michael Waesch Date 2007-06-08 22:44 Edited 2007-06-08 22:47
I wouldn´t pay anyone even 5 bucks to play chess if they can´t do it for love. Commerce has harmed any sports severly, no matter if that´s the way our rotten world works.

Just play me. For free and into the bargain you will additionally win.

... and don´t tell me that wouldn´t be as interesting as watching top players like Kramnik ...

I am also very fond of letting me mate in 1!

Parent - - By lkaufman (*****) Date 2007-06-07 13:50
Six grandmasters accepted the challenge. In order of rating, they were Hikaru Nakamura, Jaan Ehlvest, Joel Benjamin, Emanuel Berg, Arthur Kogan, and Sergei Kudrin. Because of the large rating gap between the first two and the last four, I felt that I must choose one of the top two, with practical considerations playing a role in the final choice. As for the future, if Ehlvest loses the match, I would say that it partly depends on the score (and the games); a close match might convince the sponsors to give another GM a chance, whereas a one-sided victory for Rybka would indicate that only true handicap games can be competitive. At least that would be my view as a 10% sponsor of the event. I know that two of these GMs, Benjamin and Kogan, have special experience with computers (Benjamin was the main training opponent for Deep Blue) that might give them better chances than their ratings would predict. We will surely discuss what to do next.
Parent - - By Banned for Life (Gold) Date 2007-06-07 14:24

A match with stated rules against Ehlvest will be very interesting, but I'm surprised you didn't choose Hikaru "my brain is better than Rybka six days a week" Nakamura. I suspect a match with Nakamura would garner more publicity and no doubt many more memorable quotes.

Parent - By Carl Bicknell (*****) Date 2007-06-07 15:04
Benjamin has a rating of 2576...he's pretty good.
It would be interesting to see his comments about Rybka compaired with Deep Blue
Parent - - By lkaufman (*****) Date 2007-06-07 16:42
     You are surely right about the publicity angle; there were other considerations I won't discuss here. In the event of a second match, both Hikaru and Joel Benjamin are likely candidates if they remain interested.
Parent - - By tigershark (**) Date 2007-06-07 19:44
I'm pleased the match is taking place. But with Nakamura, you'd have got massive chess publicity over all the major blogs (Susan Polgar etc); and maybe even got ICC to relay. Nakamura gets 1000s of people who automatically "follow" him when he logs onto ICC. You had an opportunity to market Rybka to them all and you've blown it

The Rybka team seem to take strange decisions with marketing

Convekta instead of chessbase, Ehlvest instead of Nakamura; less than eye catching logo on software box if you don't take the download option etc etc

It's a shame. 'Cos the product is fantastic
Parent - - By lkaufman (*****) Date 2007-06-08 00:27
This match was not sponsored by Convekta, but by private donors who were mostly more interested in a high quality match than in publicity (I think). If Convekta wants maximum publicity, they should sponsor the next match, and then they can choose the opponent. Although I am on the Rybka team, my job is to make Rybka even stronger, not to promote sales directly.
Parent - By tigershark (**) Date 2007-06-08 05:50
and you think Nakamura can't deliver a high quality match? Right.....
Parent - - By Felix Kling (Gold) Date 2007-06-08 08:21
I think for a lot of people Nakamura would have been more interesting, also nakamura's rating is higher...
I would have made an other decision, Nakamura has also some experiences playing with/against Rybka from the freestyle tournament.

At least in the second match we want to see nakamura :)
Parent - - By lkaufman (*****) Date 2007-06-08 12:45
Let's just say that there were non-chess factors that determined the final choice. If there is another match like this, Nakamura would be welcome to play if he wishes. Otherwise, Joel Benjamin wants to play the next match, whether it be this type of match or a pawn handicap match.
Parent - - By Michael Waesch Date 2007-06-08 16:49
Non chess factors determined decisions concerning chess? Sounds weird. And I also would like to know if the sponsors did agree since it´s their money ...

Parent - - By Uri Blass (*****) Date 2007-06-08 17:50
I think that it is obvious that  the sponsors agreed.
I even guess that the sponsors prefered Jaan Ehlvest and not Hikaru Nakamura because it is the only reason that I see to choose
Jaan Ehlvest when most people prefer Hikaru Nakamura.

Parent - By Michael Waesch Date 2007-06-08 18:11
Sorry, but I see absolutely no transparency here and I also see no reason not to inform us all about the process that lead to the choice of Jan Ehlvest again, especially since quite some people preferred Nakamura. And I also think it is not sufficient to just point out that Nakamura is an avid Blitz player pulling every trick to fool weaker players, since this isn´t a Blitz match. I have played Nakamura on a simul on ICC and he just won without doing weird moves trying to fool me.

Parent - - By Banned for Life (Gold) Date 2007-06-08 18:34
I think you guys are in violent agreement on this.

Certainly Jaan is a very strong player and it should be an excellent match, BUT I think its pretty obvious to all that playing against the very self assured GM Nakamura would have generated much, much more press for Rybka, very possibly even mainstream press.

One might assume that since Nakamura lives far away from Larry, there might have been logistical problems in putting together a match against Naka. Still the potential reward for both sides is so high that I'm really surprised at the decision.

From Naka's perspective, a loss would be expected and not that newsworthy whereas a win would elevate his standing in the chess community and might even make him well know outside of the chess community.

From Rybka's perspective, this match might get picked up by mainstream media, especially if Naka looked like he had serious chances to win.

In the best case, a Naka win would revive interest in the man/machine contest and would likely lead to a real sponsor for a rematch.

I agree with the poster above that from a marketing/sales perspective, this was a major blunder by the Rybka team. You never know if you will get another opportunity like this in the future...

Parent - By Michael Waesch Date 2007-06-08 18:43
You made some very good points and I also think that it was a big blunder nor to chose Nakamura.

Parent - - By Uri Blass (*****) Date 2007-06-08 20:14
I am not sure if the rybka team decided about the opponent.
It is possible that it was simply a decision of the sponsors who prefered to pay less money.

In that case  the only possible claim of blunder by the rybka team is the claim that it could be better if Vas payed the missing money that sponsors did not want to pay.

Parent - - By Michael Waesch Date 2007-06-08 20:36
Ah and that´s the reason why Kaufman is chickening out by just saying that reasons not relayed to chess lead to the decions to take Ehlvest again. In that case that´s just another big mistake for leaving the audience with the wildest speculations.

Quite a lot concerning this matter looks very suspecious now.

Parent - - By Banned for Life (Gold) Date 2007-06-08 21:24
If I were Vas, I would have 'invested' a few thousand dollars for a match against Naka. Think about it; CB had to find a sponsor to pay Kramnik a half million dollar guarantee plus a half million dollars for winning. Of course Naka isn't world champion, but he probably generates more excitement than Kramnik all the same.

Anyway, Larry worked hard to make this happen, and it will be a good match, but without the publicity and the great sound bites that Naka would generate.

Parent - By Michael Waesch Date 2007-06-08 21:41
If you ask me, I just assume that this will only a repetition of the last match - in every which way:

I.) Rybka will be handicapped which also was criticized last time, even the handicap isn´t so big this time.

I agree with the posters who stated that if (grand)masters of chess want to participate in man vs machine battles, they shall show up with everything they have to offer for a standard chess game and the machine will come up with eveything it has to offer. Talking about hadicaps alone does harm the game significantly, let alone really playing such nonsense.

II.) Again the machine faces an opponent who is neither suitable for the audience (people simply don´t want Ehlvest again, again and again) nor for the initial reason IM Kaufman stated. With Ehlvest he will not prove once and for all that non-handicapped matches against engines are pointless by now.

In addition to that, GM Nakamura really and obviously seemed to be the better choice. His self-confidence and sometimes even arrogance would really put fire on the board. If he won it would be a really good advertisement for chess and if he lost people at least could see Rybka putting a damper on the self confidence of a gm who sometimes seemingly over estimates his capabilities, as long as I can judge as plain Patzer.

III.) Again this match will be most likely relayed by an almost unknown server with no coverages on Playchess and ICC where most of the computer addicts live, leaving the match to be seen again by a fistful of people.

In addition, I already mentioned it, there is a huge lack of information about the whole process of aquiring the grandmasters and about making the final choice which will lead to the wildest speculations, reducing the benefit of the match even more.

All in all, I am not sure if I am going to watch it under these circumstances.

Parent - - By gala.martin (**) Date 2007-06-08 21:44
I think Nakamura is willing to challenge rybka. So we will likely have a match in the future. A 3 moves book could be really dangerous in such a situation, since Nakamura is an expert in opening traps for chess engines.
Parent - By Michael Waesch Date 2007-06-08 21:48
I am pretty much baffled how easily people here assume that we will be able to privately raise such funds again for another match and another on and ...

Parent - - By Vasik Rajlich (Silver) Date 2007-06-09 14:09
The comments here are interesting. I like Nakamura as much as the next guy :), but let's keep in mind that Ehlvest is also a strong player, in the same general class.

If we want to "put on a show", I'll just send some girls in bikinis to Larry's house for the photos :)

Parent - By Milton (***) Date 2007-06-09 14:19
I live not too far away from Larry in Baltimore County Maryland and I have a camera... :-)

Parent - - By lkaufman (*****) Date 2007-06-08 21:43
    Since you are all so curious about the choice, I'll just say that I did initially select Naka, but he was rather unresponsive to emails about dates and I concluded that a definite match with Ehlvest was better than an uncertain one with Naka, especially since he has not been doing well in tournaments lately and may have very little rating edge over Ehlvest on the July 1 FIDE list. With Ehlvest we can be sure that he will complete the match to his best ability regardless of the score, and will not blame any losses on minor external factors, like Kasparov accusing IBM of cheating. Maybe this would also be true of Naka, but I don't know that. From my perspective, a cocky personality and an emotional nature are negatives for a match like this. Despite this, Hikaru is still welcome to be the next opponent if we do hold another such match.
Parent - By Michael Waesch Date 2007-06-08 21:49
I disagree. It´s the primadonnas of chess, like Kasparov, that attract to the game, not the honest and deligent ones.

Parent - - By Banned for Life (Gold) Date 2007-06-08 22:52
I'm with Mike on this one. There was incredible buzz around the Kasparov vs Deep Blue matches, due in large part to Kasparov's cocky personality. Naka has the potential to generate a lot of excitement too and if he pulls out early, you would still get most of the great publicity (and keep most of your money). Of course its harder to set a match up with this type of personality, but the huge increase in interest in the match would certainly seem to justify the additional work.

Parent - - By Michael Waesch Date 2007-06-08 23:33
Indeed. I just have the impression that team Rybka is just going the easiest way: Pick the easiest to handle GM, don´t search for sponsors that can contribute more attractive prize funds, just write a simple letter to FIDE and wait what happens, just pick the Convekta server for relaying and don´t care that no very much people will watch ... and so on.

I just don´t think that this will have much, if any, effect.

[Irony on]

For the next match I propose the following:

I. The human player

Let´s pick up any talented player who will make sure that there are no complications with him at all. The conditio sine qua non will be that no one knows his name and everyone has to look it up. In addition to that he must be an honest worker with the most boring lifestyle one could imagine possible for assuring that there won´t arise any tensions during the match. The fall asleep percentage of the audience should thereby reach about 95%. Anything below will be considered as a failure.

II. The handicaps

Since humans are no longer able nor willing to really take on machines on an even level, we make the following concessions to the human player:

II.a: The human always plays with the white pieces.
II.b: The human receives queen odds and may place a Knight from outside at any free square on the board for one occasion the human player can choose arbitrarily.
II.c: Thinking time for the human is 120 minutes for the whole game.
II.d: Thinking time for the engine is 120 seconds for the whole game.
II.e: The engines must not run on a system higher than a PIII.
II.f: The engine is not allowed to mate the player.

III. Location of the game

The location of the game is the home of the player and it is the sponsor´s responsibility how they get their technical equipment there. In addition to that, the sponsors have to hire a waiter who will run to get the things needed when the player snips with his fingers. We find that fair and just for taking the challenge to such a very very hard match.

IV. Prize funds

The minimal prize fund for such a match is at least a gazillion U.S. dollars. No less!

IV.a: If the human player wins, he gets the full prize fund.
IV.b: If the human loses, the engine gets nothing.
IV.c: So in every case, the human gets the money.

V. Publicity

The positive effects of the match are to be milked to it´s last drop. It shall be told to the audience that player and sponsors did take a huge financial risk by agreeing to such a match and are therefore to be praised for their efforts. The public must believe that they all do it in good sportsmanship and for olympic ideals only and must be driven away from the truth that all is only for the money at all costs!

VI. Additional last minute agreements

VI.a: If the human player has a lost position, he is either entitled to pull the machine´s plug or demand that colors are switched.
VI.b: The human is allowed to use any source of information available for him, especially everything he finds in the toilet.
VI.c: The engine is not allowed to use any other method of move creation than its own algorythms.
VI.d: If the human player is too tired to play on, he is entitled to interupt the match as long as he likes.
VI.e: If there are technical difficulties preventing the engine from playing on, the engine thereby concedes defeat.

VII. After the match

We strongly believe in the superiority of man and his god-like abilities to handle everything to it´s best and so we expect a crushing defeat for the silicium based challenger that dared to question man´s chessic superiority.

So the player is expected to explain his glorious victories after the match in detail, so every human on this planet (or at least the 2 guys watching on our unknown server) can furthermore believe in man´s good faith and as deterrent example for other arrogant soulless things in order to prevent any upcoming similar challenges to man´s supremacy.

For a very unlikely loss of the human player, he is supposed to cry like a baby and blame secret grandmaster assistance for his loss. We will the happily arrange another match of this kind with even more concessions to the human player.

[/Irony OFF]

Parent - By Lee Ma Hong (**) Date 2007-06-09 00:52
okay, you can be the organizer of the next match ...
Parent - - By Uri Blass (*****) Date 2007-06-09 04:23
In that case it is misleading to say that Naka accepted the match.

I think that acceptence of a GM should include 2 dates:
1)dates of choosing the GM to play.
2)dates of the match.

1 should be equal for all GM's
2 can be dependent on the GM and different GM's may prefer different dates.
Parent - By dareapa (**) Date 2007-06-09 04:28
No.. Naka accepted but he was not chosen to play rybka for reasons stated in this thread.
Parent - - By Hetman (*****) Date 2007-06-10 19:53
I am not sure what the people prefer. You know that from ...where ? I have not made any poll or voting, you either.
The way to check that is to offer the tickets for the match Nakamura-Rybka and Ehlvest-Rybka . Then from the count of bought tickets we would know  something.
I am curious if money got from tickets will satisfy any GM.
I like chess but i know that it is not show game and any money coming to chess area  are welcome.
Parent - - By Michael Waesch Date 2007-06-10 19:58
There is surely nothing wrong with choosing Ehlvest again from the game´s point of view. He is a solid worker and he plays on as hard as he can, not giving up no matter how far he might be behind and this for those measly bucks offered (compared what is usual unfortunately for such games).

But it´s also true that the more glamorous or arrogant players attract more audience.

Up Topic Rybka Support & Discussion / Rybka Discussion / It's Ehlvest Again!
1 2 3 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill