Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Joel Benjamin Match
- - By lkaufman (*****) Date 2007-06-15 17:04
     I just received confirmation that Grandmaster Joel Benjamin will play a handicap match with Rybka during the second week of August. GM Benjamin has quite a distinguished record; he has been three times US Champion, six times World Open champion or co-champion, three times U.S. Open Champion or co-champion, and has played on the U.S. Olympic Team in six Olympiads and two World Team Championships, and also in one Interzonal. More to the point, he was the chess consultant for Deep Blue, having earned the job by defeating it 2-0 in a formal match in 1995. 
     The tentative plan is for a pawn odds match similar to the first Ehlvest match, but with a longer time control and increment (one hour + 30"). However, if the upcoming Ehlvest match proves to be close, we will attempt to find sponsorship for a match of that type instead.
Parent - By Carl Bicknell (*****) Date 2007-06-15 19:49
Fantastic!
I honestly think he is the most interesting out of the bunch you listed on a previous thread.

After Rybka has won the match don't forget to interview him about chess today and Rybka c/p Deep Blue etc - it makes an interesting read.
Parent - - By plicocf (***) Date 2007-06-15 19:52
Good news, IM Larry Kaufman!
A request: please, no pawn odds, may be better, if Rybka lost to
GM Ehlvest, less time to the GM and/or a book. If Rybka wins,
no 3 moves book and more time for GM Benjamin or less time to Rybka.

Pauloo Soares
Parent - - By Uly (Gold) Date 2007-06-15 20:12

> please, no pawn odds


I agree. To me pawn odds is not chess, it's just a variation in where one side has a pawn missing. It's like if they played chess 960... Or worse.

At least some positions are legal:

1. Nf3 Nc6 2. Ng5 Nb8 3. Nxh7 Nc6 4. Ng5 Nb8 5. Nf3 Nc6 6. Ng1 Nb8

rnbqkbnr/ppppppp1/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 7


But does that look like chess to you?
Parent - - By Carl Bicknell (*****) Date 2007-06-15 21:20
I have to say Vytron is right.

Could you give Joel a big reward for every game he draws / wins but play him at normal chess?
Parent - - By Uly (Gold) Date 2007-06-15 21:42
Now compare it with the following position:

1. e4 e5 2. Nf3 Bc5 3. Nxe5 d6 4. Nf3 Nf6 5. Bc4 O-O

rnbq1rk1/ppp2ppp/3p1n2/2b5/2B1P3/5N2/PPPP1PPP/RNBQK2R w KQ - 0 6


That looks a lot more like chess to me, I say let Rybka choose what pawn to give up and how (I already explained how long time ago and went unnoticed).

Or you can even make an opening book in where in all lines Rybka sacrifices a pawn or two, then that's still handicap and is still chess.
Parent - - By lkaufman (*****) Date 2007-06-16 04:56
    This position is more or less what you might expect if the handicap is the Black "e" pawn at the start. Some pawn handicaps produce much more normal looking chess than others. In my opinion, the handicap of the "c" pawn is the most normal-looking one; it reminds me of various gambits like the Smith-Morra or the Scotch, without the lead in development. The point of using all eight pawns is simply to provide variety, but I must admit a temptation to restrict the choice to the more normal-looking ones. The most unpleasant one for the computer, especially if it plays Black, is the "f" pawn.
     As for choosing gambits instead, obviously we can't use normal gambits, because then it wouldn't be a handicap. We could devise clearly unsound gambits, but the result would not necessarily be better than just removing a pawn, I think.
     Please feel free to restate your old idea, I don't recall it.
Parent - - By Uly (Gold) Date 2007-06-18 02:51 Edited 2007-06-18 02:55

>> As for choosing gambits instead, obviously we can't use normal gambits, because then it wouldn't be a handicap. We could devise clearly unsound gambits, but the result would not necessarily be better than just removing a pawn, I think.


The "better" result is psychological for all the persons (me included) that don't like pawn handicaps, and I don't because the starting position is not a legal chess one (Or reaching it requires cooperation from both players). So reaching a position with pawn handicap from the starting position would be the solution (For all the persons that dislike piece handicaps for this reason).

>> Please feel free to restate your old idea, I don't recall it.


The idea was about "evaluation handicap", that is, we don't base the handicap on the material on the board, but on Rybka's evaluation. Let's say we want a -1.0 handicap, then Rybka would play bad moves on purpose until her score is -1.0 or higher, then she would play the strongest moves possible for the rest of the game (A book that already does this would be better).

This was a solution at a time in where 78% of the people wanted standard chess ( http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=7316 ), but that would lead to Rybka crushing every time or to Rybka being weak for all the games, so my solution addressed those problems (If other ideas don't work).
Parent - - By lkaufman (*****) Date 2007-06-18 04:10
     The cleanest implementation of your idea would be to have Rybka play Black every game and be required to start every game with 1...a6 and 2...h6 (obviously assuming no 2Bxa6 by White). This gives White a score of +0.54 (based on 5 ply search), which actually equates to a full pawn (in the opening, Rybka scores need to be nearly doubled for the normal interpretation of pawn=1 etc.). This would be a valid handicap, comparable to a pawn (with alternating colors), fully legal, and would not restrict the human player at all. It would also have the advantage that at least Black would not have made any clearly harmful moves. On the downside, there would be no variety unless the human player chose to make it so, it would be too easy for the human to prepare openings using Rybka at home, and some would disagree with Rybka's evaluation of the handicap. After all, Rybka considers 1...a6 to be a poor move, leaving White with a 0.25 (meaning half a pawn by normal count) advantage, but Tony Miles once beat Karpov at his peak with this "handicap".
     More generally, I think most people would be far less impressed with Rybka winning from this position (or any similar opening position which she evaluates similarly) than from a victory in which she starts every game a pawn down.
     A slightly larger handicap would be to have Rybka open every game with 1...Nc6 and 2...Nb8, in effect a three move handicap. I would like to see a match tried with this handicap at some point, though there wasn't much support for it when we took a poll. This would appear to be in line with your idea.
     I suppose we could create a book in which every game starts with a pawn blunder, like 1 e4 b5 2 Bxb5 or 1e4 f5 2exf5 etc., but I'm not sure that most people would find this more appealing than simply removing a pawn (although clearly you would). Also, I find the idea of requiring the GM to open a certain way, even with a good move like 1e4, to be unattractive; what if he never opens 1e4 normally, his chances might be less than another GM who does.
     Another handicap I mentioned was that Rybka would forfeit castling, in effect normal chess where she "promises" not to castle. This might suit your preferences for a more normal game, but this idea did not gather any support.
Parent - - By Uly (Gold) Date 2007-06-18 04:59

>> The cleanest implementation of your idea would be to have Rybka play Black every game and be required to start every game with 1...a6 and 2...h6 . This gives White a score of +0.54 (based on 5 ply search), which actually equates to a full pawn.


Yes, but for variety's sake she would do that on one of the games, and other similar things on other games (What about a6 + c6 + e6?, there should be plenty of ways of doing this, so we get rid of all the problems you mentioned, even if the implementation is not "cleanest").

>> More generally, I think most people would be far less impressed with Rybka winning from this position (or any similar opening position which she evaluates similarly) than from a victory in which she starts every game a pawn down.


Maybe you are right, and people would be less impressed with something they'd want to see more than with something they didn't (pawn handicap), but there's only one way to know. I think we should make another poll to ask what the people want for future handicap matches; if they don't like any kind of handicap then we're wasting our time.

>> A slightly larger handicap would be to have Rybka open every game with 1...Nc6 and 2...Nb8, in effect a three move handicap. I would like to see a match tried with this handicap at some point, though there wasn't much support for it when we took a poll. This would appear to be in line with your idea.


Yes, I'd like to see that as a last resort.

>> I suppose we could create a book in which every game starts with a pawn blunder, like 1 e4 b5 2 Bxb5 or 1e4 f5 2exf5 etc., but I'm not sure that most people would find this more appealing than simply removing a pawn (although clearly you would).


Yes, as long as this is a one person opinion (me) then we could just keep on with the pawn handicaps (Because all the hassle that would be done wouldn't be worth just one person). I hope someone would back me up, about the "we don't like pawn handicaps but we would accept similar handicaps".

>> Another handicap I mentioned was that Rybka would forfeit castling, in effect normal chess where she "promises" not to castle.


It would indeed be very interesting, what about letting the GM choose what kind of handicap would he want? (Or no handicap at all, we got plenty of options).

One thing I've noticed is that sometimes things sound better on paper than what they really are, but we already had pawn handicaps, I'm just saying we could have something different to see if people like it more.
Parent - - By lkaufman (*****) Date 2007-06-18 14:17
As for letting the GM choose the handicap he wants, naturally if a grandmaster proposes a match with any specific handicap, we will give his proposal serious consideration. But obviously we cannot offer a prize fund without specifying the handicap, he'll ask for a queen!
Parent - By Uly (Gold) Date 2007-06-18 21:14

> he'll ask for a queen!


Maybe some day that will be a valid handicap ;)

I want to thank you for taking the time to replying to my posts :)
Parent - - By Banned for Life (Gold) Date 2007-06-18 06:05
In my opinion, the cleanest way to handicap Rybka would be to have Rybka play all out but give draw odds on every game. If this wasn't enough, Rybka could give move and draw odds on every game. This methodology would provide a significant advantage for the GM without making the game appear artificial, and would also encourage the Rybka team to set up Rybka to play very aggressive chess to avoid draws.

Alan
Parent - - By Vasik Rajlich (Silver) Date 2007-06-18 07:55
Draw odds would change the play too much.

Personally, I like pawn odds. It's really simple. If you need two sentences to explain the rules, that's 1.5 sentences too many.

Vas
Parent - - By Banned for Life (Gold) Date 2007-06-18 15:48
"Draw odds would change the play too much.

Personally, I like pawn odds."

These juxtaposed sentences would seem very strange if they didn't come from a Redskins fan. :-)

Regards,
Alan
Parent - - By Vasik Rajlich (Silver) Date 2007-06-19 12:22
Is this because only Redskin fans understand the concept of a paragraph? :)

BTW - pawn odds doesn't change the play that much. And, most importantly, the difference is less and less as the game proceeds.

Vas
Parent - - By Banned for Life (Gold) Date 2007-06-19 19:38
Pawn odds makes big changes in the opening of the game, throwing theory out the window, but doesn't affect the middle and endgame, while draw odds leaves the opening alone, but may require the side giving the odds (Rybka) to take more risks in the middle and endgame to avoid a draw. Most GMs have played for a draw with both colors at some point so this shouldn't be unnatural for them at all. So you get to pick your poison. I suspect that GMs will be more comfortable playing with draw odds and the chess community will have an easier time relating to this type of game. Maybe you can stir up a lot of interest in material handicap games but this seems like it will be an uphill struggle.

Alan
Parent - - By turbojuice1122 (Gold) Date 2007-06-19 20:20
It would seem silly (if this is indeed the plan) to have a draw for the GM worth the same amount as a win by the GM.  I would suggest a variation on the theme some are discussing concerning draws in the Freestyle tournaments: in this case, wins are always worth one point, but a draw is worth a third of a point to Rybka and two-thirds of a point to the GM.
Parent - - By Banned for Life (Gold) Date 2007-06-19 21:00
Or you could count points in the normal way but say that Rybka has to get a score of, say 6 points in 8 rounds to win.
Parent - By Vasik Rajlich (Silver) Date 2007-06-21 12:42
As I see it, the idea of a handicap is that the person receiving the handicap has a higher chance to achieve his objective in each game. This makes the games more exciting for the spectators.

Vas
Parent - By lkaufman (*****) Date 2007-06-18 14:14
     Draw odds, or draw and move odds, will be the perfect way to test Rybka once we have made a major effort to teach her how to play for the win with either color against weaker opponents (i.e. everyone). It is not a good option now, because Rybka will generally play to draw with Black unless the opening book lands her in an unbalanced position, which won't happen if White is seeking a draw.
     I have two goals for Rybka relative to human opponents. One is to be competitive with grandmasters at the traditional pawn and move handicap (f7). The other is to be able to defeat any human player in normal chess most of the time, even when he plays to draw and has White. Clearly your test is one of the two Rybka must pass to prove that my goals have been reached. We're just not nearly ready yet.
Parent - - By lkaufman (*****) Date 2007-06-15 21:37
     Shortening the book is not an option, because we need three moves just to provide some variety and prevent prepared games. If the Ehlvest match is close, then increasing the time ratio makes sense, if we can get sponsorship. If it's not close, the only alternative to material handicaps would be (as suggested below) a match in which the goal of the GM is simply to get draws, i.e. a draw odds match. But I don't think that would be as interesting, and anyway we have not yet made any effort to teach Rybka to avoid draws, so this is something for the future. Once Rybka has learned the difficult art of winning (especially with Black) against a strong opponent who plays for a draw, I think draw odds matches might then be quite interesting. But for now, we must give the human opponent enough advantage to allow him to at least try to win the game, if we want to have an interesting match.
     Please bear in mind that for the prize money we are offering for a match victory, namely $2,000, the Grandmaster must believe he has a good chance to win, so there is no point in proposing terms that don't offer him a serious chance of victory unless someone else is offering to sponsor the match.
Parent - - By Uri Blass (*****) Date 2007-06-16 02:38
Note that the opinion of GM artur kogan is that he believes that with some extra praparation he will have very good chances to draw and even to win the match!

Maybe he is going to change his mind after the next match and maybe not but
I think that as long as there are GM's who believe that they have good chances to win there is no need to increase the handicap relative to the handicap that you give to GM Ehlvest.

Result against Ehlvest is going to prove nothing and I remember that some years ago in the israeli league some 2200 player drew with 3 computers when GM's did worse against computers.

Uri
Parent - - By lkaufman (*****) Date 2007-06-16 04:44
     Because of his experience with the pawn odds match, Ehlvest is already more qualified than the average GM of his rating for this match. Certainly the result will not be conclusive, but it will be strong evidence of how others might fare. Sure, if we give fifty GMs a chance to play such a match, maybe one of them will figure out how to trick Rybka and perhaps win, but there is not unlimited money for these matches. We can't base invitations on how the GMs estimate their own chances; most of those who want to play believe they have a good chance, that's why they are so keen to play for a moderate guarantee. But I do hope we can have some match with GM Kogan, I believe he is sincere in his opinion.
    As far as I know, there is no GM in the US who really believes he has any decent chance to win a match against Rybka without a material handicap (I don't know what Nakamura really thinks, he would probably say he has chances). The fact that only four US grandmasters agreed to the challenge despite eleven thousand dollar prize fund shows this clearly. Maybe there are a few in the world (excluding the Elite) who do believe this, but not many.
    
Parent - - By Banned for Life (Gold) Date 2007-06-16 04:57
"Because of his experience with the pawn odds match, Ehlvest is already more qualified than the average GM of his rating for this match."

I'm not sure I follow this line of reasoning. Is there anything that happened during this match that another GM couldn't have reproduced by playing against Rybka in his basement? Seems like it should be pretty easy for a GM to assess his chances.

Regards,
Alan
Parent - - By lkaufman (*****) Date 2007-06-16 05:01
You learn a lot more from playing a game or a match than from watching one, believe me.
Parent - - By Banned for Life (Gold) Date 2007-06-16 05:07
I don't disagree with that, but I'm not sure what would prevent a GM from playing a game or a match against a very similar Rybka at any time. If he/she can get positive results against 2.3.2 under the conditions you are offering at home, they have reason to believe they could win a match. If they get slaughtered every time, they should probably look for other opportunities. No?
Parent - - By lkaufman (*****) Date 2007-06-16 13:22
Not many professionals would spend a couple days like this for free, to determine whether they should spend more days doing the same thing with the risk of only getting paid a little!
Parent - - By Banned for Life (Gold) Date 2007-06-16 21:14
I'll stipulate to your much more in-depth knowledge of high level chess. It would certainly seem reasonable to play at least one game against Rybka, with simulated match conditions, before accepting the LK challenge, but maybe most GMs wouldn't bother.

Its interesting to me that GMs don't consider playing against a strong engine as being helpful in developing their skills against other GMs.

Regards,
Alan
Parent - By lkaufman (*****) Date 2007-06-17 03:13
     Probably they do, but would not choose to play under unusual conditions like pawn odds or (almost) no opening book if they are practicing for other GMs. Also, they would not play really long games; it's hard for a professional to motivate himself to play multi-hour training games without pay. For training purposes, I would recommend grandmasters play Rybka normal chess at time odds like one hour to five minutes, but few could stand the humiliation of getting trounced under such conditions.
Parent - - By Gaмßito (****) Date 2007-06-15 22:57
Larry,

Indeed, GM Joël Benjamin is a very fine player. We are talking here about an experienced player in computer chess games and I think he could be very dangerous in a handicap match against Rybka. Of course, in my opinion I do not prefer a very strong GM's but without experience in these confrontations. I even prefer a non very strong GM's (per ELO Fide) but with a lot of experience. Anyway the invitations should continue open. GM Joël Benjamin is a good choise and I think the match against him will be quite interesting. Thanks LK. :-)  

Now, excuse me, per my insistence Larry, but could you please try to contact GM Dzindzi and give him an opportunity. I knew that you can easily contact him and I am sure that, if he is good enough and stable in health, he will play against Rybka with all the pleasure. Do not forget their wonderful abilities and his thousands of thousands of good games over the Internet against some of strongest engines, some years ago. Dzindzi is an artist and a great player with truly experience. He really knows how to play against the machines. Do not allow that a good opportunity like this, be vanished, plus if you can talk easily with him.

Regards,
Gambito.

FROM Wikipedia, some information about Joel Benjamin:

Joel BenjaminJoel Benjamin (born March 11, 1964) is a chess Grandmaster.

As of April 2007, his Elo rating was 2576, making him the # 12 player in the US and the 214th-highest rated player in the world.

Joel Benjamin is the top-rated active chess player who was born in the United States.

He is a native of Brooklyn, New York, and grew up in the Marine Park neighborhood there. He graduated Yale University in 1985.

At the age of 13 he broke Bobby Fischer’s record by becoming the youngest-ever U.S. Master. As a junior player he won the National Elementary title (1976), the National Junior High crown (1978), and the National High School title (1980-81).

Other triumphs have included the U.S. Junior Championship in 1980 and 1982, and the U.S. Open Championship in 1985. He earned the Grandmaster title a year later.

He was the US Chess Champion in 1987, 1997, and 2000.

He was hired as a consultant by IBM to help with the Deep Blue chess computer that defeated World Chess Champion Garry Kasparov in 1997.

As of the October 2006 rating supplement Benjamin had a United States Chess Federation rating of 2646, ranking him 14th best among American chess players.

He appeared in the movie “Searching for Bobby Fischer.”

In 1998 he was voted “Grandmaster of the Year” by the U.S. Chess Federation.
Parent - - By lkaufman (*****) Date 2007-06-16 04:29
I have no problem with asking Dzindzi to play one of these matches, but right now there is a "line" forming of GMs wanting to play. Maybe he'll have a chance before too long if he is interested.
Parent - - By Chandan (*) Date 2007-06-16 04:56
Hi Larry,

I feel that Anand can be persuaded for a match against Rybka if the Rybka team agrees to spend the sponsorship amount on his charity firm. He is known to play games for his charity. This way I feel chess and Rybka would have served the community and chess enthusiasts will have their dream game between the strongest machine and best human player today.
Parent - - By lkaufman (*****) Date 2007-06-16 04:59
Of course that would be great, but I believe Anand would expect sponsorship with at least an extra zero added to the end before he would consider such a match, and probably a lot more. Since Kramnik got half a million, Anand might feel that he cannot play for less, regardless of whether the money goes to him or to charity.
Parent - By Chandan (*) Date 2007-06-16 05:22
I still think that an attempt should be made to engage him in talks for such a match. If it does not work out , so be it. If it does it will be great for chess and charity
Parent - By Vasik Rajlich (Silver) Date 2007-06-16 10:25
Excellent -I'm looking forward :)

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Joel Benjamin Match

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill