Poll
Which handicap is best for the next Rybka vs human match? (Closed)
| Different pawn every game, Rybka always white | 2 | 6% | |
| Different pawn every game, alternating colors | 11 | 32% | |
| Different pawn every game, Rybka always black | 7 | 21% | |
| Rybka always black and without f7 pawn | 2 | 6% | |
| Human starts with an extra exchange | 3 | 9% | |
| Human plays white and with bishop pair | 0 | 0% | |
| Human makes first three moves | 4 | 12% | |
| Human gets 1. e4, 2. d4 and two more free moves | 3 | 9% | |
| Handicap decathalon - some combination of above | 2 | 6% |
![[hu]](/mwf/v2/flags/hu.png "Hungary")
Which of the above handicaps would be the most interesting for the next handicap match?
It's really a philosophical question. Do we keep it simple and classical (ie. pawn handicaps), or is it more interesting to try a variety of ideas.
Vas
It's really a philosophical question. Do we keep it simple and classical (ie. pawn handicaps), or is it more interesting to try a variety of ideas.
Vas
![[cz]](/mwf/v2/flags/cz.png "Czech Republic")
Different two pawn every game, alternating colors
![[cr]](/mwf/v2/flags/cr.png "Costa Rica")
Hi Vas,
I think the best options (in order) are :
1. Different pawn every game, Rybka always white, but with more time for both. It is better more than 1 hour.
2. Different pawn every game, alternating colors, but with more time for both. It is better more than 1 hour.
3. Rybka always black and without f7 pawn, with the same time rules of option 1 and 2.
The next options are maybe of too much advantage for the human at the moment.
Ragards,
Gambitto.
I think the best options (in order) are :
1. Different pawn every game, Rybka always white, but with more time for both. It is better more than 1 hour.
2. Different pawn every game, alternating colors, but with more time for both. It is better more than 1 hour.
3. Rybka always black and without f7 pawn, with the same time rules of option 1 and 2.
The next options are maybe of too much advantage for the human at the moment.
Ragards,
Gambitto.
Some comments about the votes (so far). The most popular single option, alternating colors pawn odds, is perhaps viable for the next match against a 2600 level opponent, though I think it would be tough for Rybka unless we succeed in making some non-trivial improvments. Playing all Black at pawn odds might be appropriate if our opponent is a slightly weaker GM, closer to 2500 level. Handicaps of the f7 pawn or the Exchange (white rook for black knight) are too large for now against any "true" GM (i.e. one who still has the requisite 2500 FIDE rating) unless you are talking about very fast games, but I do hope to see Rybka get there in a couple years or so.
I note that there was significant support for handicaps of N free moves, which at least has the advantage of being an unambiguous advantage (unless you believe White is in Zugzwang at the start!). This gives me the idea that we could play a match where the first game is played at (for example) three move odds, and the number of free moves is changed after each decisive game. If Rybka wins, the number of free moves is increased by one, if the GM wins, it is decreased. If he wins at 2 moves, the next game is normal chess where he just gets White.
I note that there was significant support for handicaps of N free moves, which at least has the advantage of being an unambiguous advantage (unless you believe White is in Zugzwang at the start!). This gives me the idea that we could play a match where the first game is played at (for example) three move odds, and the number of free moves is changed after each decisive game. If Rybka wins, the number of free moves is increased by one, if the GM wins, it is decreased. If he wins at 2 moves, the next game is normal chess where he just gets White.
![[is]](/mwf/v2/flags/is.png "Iceland")
>first game is played at (for example) three move odds, and the number of free moves is changed after each decisive game. If Rybka wins, the number of free moves is increased by one
What rules do you have in mind regarding the choice of moves to avoid an easily won game after the free moves? Is disallowing captures sufficient? Some rules are needed when the number of free moves increases to prevent checkmate or a huge material advantage.
The standard rule for multi-move handicap is that you must not go beyond the fourth rank. I think this would be sufficient.
Will you play this matches only, when you are sure that Rybka will win? That´s boring. I wish a match with Rybka under pressure; not always the humans! Engines are much better under pressure (they don´t feel it really) than humans. Therefore my rules for the match:
- alternating colors pawn odds (even better human always white)
- 90 minutes (+15 seconds) per side
- only two games per day and one rest day after four games
- human player is a top-100 player (FIDE 1.04.2007: Jaan (#74) or better Michal Krasenkow (#40) or most wanted Victor Kortschnoi (#100))
When the human will not win, then most likely the match-rules have failed. I´m abolutely sure, that the games will be not boring.
- alternating colors pawn odds (even better human always white)
- 90 minutes (+15 seconds) per side
- only two games per day and one rest day after four games
- human player is a top-100 player (FIDE 1.04.2007: Jaan (#74) or better Michal Krasenkow (#40) or most wanted Victor Kortschnoi (#100))
When the human will not win, then most likely the match-rules have failed. I´m abolutely sure, that the games will be not boring.
We were not at all certain that we would win against Ehlvest, we just thought we had the better chances. It turned out that he needed several games to understand how to handle the handicap and the clock. In general I agree that a close match with an uncertain result is desirable. I think that the conditions you suggest (with alternating colors) are pretty reasonable, though I would recommend a longer increment (the normal 30") and shorter basic time control (but at least one hour) to insure quality endgames. If we rematched with Ehlvest under those conditions with the same program, I think we would be the underdogs, but hopefully Rybka will keep getting stronger. As to the choice of opponent, that mostly comes down to who is willing to play for the money we offer. We were fortunate to get a player of the stature of Ehlvest given the size of the prize. There is also the factor that if I am to host the match, it needs to be a player who is already playing in the U.S. to keep costs down.
FWIW - it's my understanding that in Shogi and Go, when handicaps are used, they decrease the gap between the players but don't completely eliminate it. It seems logical that the handicap we pick should leave the human a slight underdog.
Anyway, if Rybka should lose one of these matches, sure, I'll break some glass and furniture and Larry will punch a few holes in the wall, but then we'll just get back to work :)
Vas
Anyway, if Rybka should lose one of these matches, sure, I'll break some glass and furniture and Larry will punch a few holes in the wall, but then we'll just get back to work :)
Vas
Okay, not so bad from you and Larry. But when Victor will lose, it´s not so easy. A atomic bomb explodes in Larry´s living room... And I´m quite sure that Larry wiil be pleased, when Victor will take the win. Emotions are unique.
Look at this: http://youtube.com/watch?v=TxeiGipoFSE
I don't want such a guy to play against Rybka. What he does is unfair and primitive.
I don't want such a guy to play against Rybka. What he does is unfair and primitive.
![[no]](/mwf/v2/flags/no.png "Norway")
He has always been known to be great fighter and a sore looser.... But I guess thats one of the reasons he is still as strong as he is.
![[br]](/mwf/v2/flags/br.png "Brazil")
Please, no handicap, let´s play chess. I think that you have to choose a GM that understand
computer chess.
Paulo Soares
computer chess.
Paulo Soares
By Michael Waesch ![[de]](/mwf/v2/flags/de.png "Germany")
Date 2007-04-03 07:14
I don´t get much from these handicapped games too and I missed the voting option for standard chess.
Mike
Mike
![[gb]](/mwf/v2/flags/gb.png "United Kingdom")
This "handicapped" chess is illegitimate. The likes of Adams, Anand, Kramnik and Kasparov all played under proper chess so why dilute the quality of the game?
By Michael Waesch
Date 2007-04-03 10:54
All in all I think it comes down to financial reasons, Team Rybka simply being unable to pay a high rated grandmaster who is willing to take on Rybka with no handicaps given to him. So we will never see such a match.
Mike
Mike
![[be]](/mwf/v2/flags/be.png "Belgium")
I totally agree with Nuff: All this handicap matches are (nearly) a show : not chess!
Vas, if you want something really interesting, it's needed (not easy at all, I know) to find a sponsor to pay a top GM (ELO >2675) for a long match (minimum 10 games) at classical slow time control, with one game every two days. The result would be highly significative and interesting... and probably fantastic for Rybka!!
With short time controls, the advantage for the computer is huge and don't show a lot. I'm sorry, this is sad.
BTW, it's 17h20 CET, and I can't see game 5 on "direct broadcast". What is the problem?
Vas, if you want something really interesting, it's needed (not easy at all, I know) to find a sponsor to pay a top GM (ELO >2675) for a long match (minimum 10 games) at classical slow time control, with one game every two days. The result would be highly significative and interesting... and probably fantastic for Rybka!!
With short time controls, the advantage for the computer is huge and don't show a lot. I'm sorry, this is sad.
BTW, it's 17h20 CET, and I can't see game 5 on "direct broadcast". What is the problem?
Handicap play does have some tradition in chess, going back to the last century. So, Paulo, you are too young :)
In other games, handicap play is a bit more prevalent. In fact, the top Shogi program is about to play a handicap match of its own against a top human player. The difference is that the human gives the handicap there.
Vas
In other games, handicap play is a bit more prevalent. In fact, the top Shogi program is about to play a handicap match of its own against a top human player. The difference is that the human gives the handicap there.
Vas
Slight correction--until recently the public games between the human shogi champ and the computer shogi champ have indeed been with the human giving a handicap (the last few games were bishop handicap, with mixed results). However the software ("Bonanza") has gone MP and been improved greatly, and so for the latest "match" (just one game, 2 hours per side) played just a few days ago on an octal system the computer played with no handicap. It was an extremely close and exciting game, but the human champ prevailed. Public predictions are that they will be truly equal within five years, I personally think it will be more like three. I'm on record as predicting that within ten years the computer will be able to give the human champ the minimal handicap (lance) successfully.
Wow, things are really flying forward in shogi as well.
Vas
Vas
Vas, this tradition exists, I agree, but it is for played between a strong player and a more
weak player. I believe that have GMs that understand the weakness of the chess programs.
And if not to exist, they can be prepared to play against any program.
You can see Edward and Pablo, they have not a rating >2300.
Paulo Soares
weak player. I believe that have GMs that understand the weakness of the chess programs.
And if not to exist, they can be prepared to play against any program.
You can see Edward and Pablo, they have not a rating >2300.
Paulo Soares
By Michael Waesch
Date 2007-04-03 19:08
LOL, so may be Pablo should play ...
Most likely he would do it for free, without handicaps and also most likely the result would be 7-1 in favor of Rybka ...
Mike
Most likely he would do it for free, without handicaps and also most likely the result would be 7-1 in favor of Rybka ...
Mike
result would be 8-0 for rybka.
By Michael Waesch
Date 2007-04-04 05:07
I am not sure on that. Sometimes his stonewall simply holds with the engine shuffling pieces around, just to no avail.
But one thing is for sure: It would be totally boring to watch.
Mike
But one thing is for sure: It would be totally boring to watch.
Mike
Paulo,
based on Larry's experiments, in a classical time control, a non-rook pawn is worth around 250 to 300 Elo. It means that Kasparov could give pawn odds to a 2550 player and expect an even match.
Vas
based on Larry's experiments, in a classical time control, a non-rook pawn is worth around 250 to 300 Elo. It means that Kasparov could give pawn odds to a 2550 player and expect an even match.
Vas
Vas, an interesting study from IM LK, I never saw something seemed with this.
But I would to ask to Rybka team, that if have a possibility of playing the traditional chess,
please, make it.
Paulo Soares
But I would to ask to Rybka team, that if have a possibility of playing the traditional chess,
please, make it.
Paulo Soares
Of course, this is also possible. We'd need a sponsor.
Vas
Vas
It seems that there is a clear consensus from the voters in favor of the pawn handicap format, all eight pawns one at a time, but without even the compensation of the White pieces for Rybka. So I guess that most likely this means that the next match will be with alternating colors, with the eventual goal of a match with all Black pieces. I don't think we are ready yet to defeat a 2640 FIDE player like Ehlvest this way (alternating colors), especially if we increase the increment, but hopefully we're not too far away from being competitive under such conditions. I note for the record that this is similar to the way Kasparov played a handicap match a few years ago against a billionaire who was said to be something like a 2200 level player or so; Kasparov gave him two pawns with alternating colors, but one was always the "a" pawn, so it was much less than twice the handicap we are talking about. Kasparov won by 2.5-1.5, I believe.
Larry,
there is a tendency for users of technology to "rush ahead", ie. Rybka beat Ehlvest with 8 whites, so let's move on. It's up to us to put the breaks on if Rybka is not quite there yet.
In my view, a perfectly reasonable next match is same starting positions as vs Ehlvest, but longer time control.
Vas
there is a tendency for users of technology to "rush ahead", ie. Rybka beat Ehlvest with 8 whites, so let's move on. It's up to us to put the breaks on if Rybka is not quite there yet.
In my view, a perfectly reasonable next match is same starting positions as vs Ehlvest, but longer time control.
Vas
I guess a lot depends on exactly whom we play next. If it's Ehlvest again, I quite agree with you; a longer increment together with his experience should be enough for a fair match. If we have to settle for a somewhat weaker GM, perhaps we can defer to the voter's wishes and try alternating colors.
![[ph]](/mwf/v2/flags/ph.png "Philippines")
I suggest a two-pawn-handicap. Probably the two Rook pawns. But the computer wins if the game is DRAW
Powered by mwForum 2.22.1 © 1999-2010 Markus Wichitill