A program could certainly be tuned to play better in such handicapped games, but it would take some tuning to play optimally, which would be counter-productive when playing normal games.
I can see where you are coming from, but if done right I would disagree with this. Tuning to play well in this situation would improve understanding of exchange sacrifices and such, just like tuning for pawn handicaps improves understanding of pawn sacrifice positions.
I would imagine that these matches have been at least slightly helpful in improving Rybka strength in terms of supplying interesting data/ideas for improving handling of difficult concepts such as material imbalance and contempt (separate issues if done right).
A program could certainly be tuned to play better in such handicapped games, but it would take some tuning to play optimally, which would be counter-productive when playing normal games.
I can see where you are coming from, but if done right I would disagree with this. Tuning to play well in this situation would improve understanding of exchange sacrifices and such, just like tuning for pawn handicaps improves understanding of pawn sacrifice positions.
I would imagine that these matches have been at least slightly helpful in improving Rybka strength in terms of supplying interesting data/ideas for improving handling of difficult concepts such as material imbalance and contempt (separate issues if done right).
-Sam
It isn't quite the same. When trying to teach a program how to play exchange sacs, that is _optional_. But if it is a done deal before the game starts, I would evaluate things differently myself. A normal program will do exchange sacs relatively infrequent, because for the most part, they are bad. So you have to evaluate things in a way to prevent doing into such too frequently when it is wrong. But if you are forced to start there, then the evaluation would probably be a bit different. I'd have to think about how much different, but at least several things come to mind.
Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced. I guess all these grandmasters are just talking hyperbole. Garry kasparov said five years ago when you were saying computers play at 2400 that the top programs were in the 2700s. Kramnik said something similiar. After following you in this debate for years, i'm convinced that if God himself said computers were now 3000 you would disagree. Kasparov must have been dead right cause programs like Rebel were drawing Matches at standard time control with 2700 players 5 years ago.
bigo wrote:Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced. I guess all these grandmasters are just talking hyperbole. Garry kasparov said five years ago when you were saying computers play at 2400 that the top programs were in the 2700s. Kramnik said something similiar. After following you in this debate for years, i'm convinced that if God himself said computers were now 3000 you would disagree. Kasparov must have been dead right cause programs like Rebel were drawing Matches at standard time control with 2700 players 5 years ago.
Excuse me if I try to explain. I tend to agree with both sides. From a player's perspective all the title holders and GM are correct from their practical side, Bob could easily think similar, but he doesnt, and the reason is IMO that scientifically, statistically, logically such a verdict is wrong. At least we should differentiate between the difficulties to play these machines and the resulting performance in dependence of human GM players who seldom enough are playing reasonable chess variations against specifically a machine, and the sort of objective strength of these machine with also their crass weaknesses. If these weaknesses were systematically played against then the overal record would look different. You just cant publish Elo numbers after so few games, apart from the aspect that one shouldnt apply Elo to machines the way we know that for human chess. Human chess is something completely different with relative stable personalities over years. But in CC you can create different personalities for the morning and evening play sessions. Of course directed by human operators. Etc pp.
-Popper and Lakatos are good but I'm stuck on Leibowitz
Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced.
I am not sure who you are talking to. I personally don't really have an opinion about Rybka strength. If it was allowed in real tournaments and matches regularly we would know...just playing occasional exhibition matches its hard to tell. Computers certainly seem strong, but it would be pleasurable and informative to let one computer play serious chess for a few years and see how people adjust (preparing for it in tournaments like any other opponent). Without that its just darn hard to know how strong they are. If Rybka didn't have some humans on staff constantly updating its opening book, I would guess over the years its results would wane, so the strength of its helpers compared to the world's best must also be considered (particularly with grandmasters having access to Rybka to test how it would play against lines they are considering).
Anyway, folks around here that are experts in computer chess and master or near master at normal chess (which covers multiple people on this site and in this thread) seem like they are in a position to offer as valid an opinion on Rybka strength as the GMs you mention. Maybe not better, but maybe not worse.
Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced.
I am not sure who you are talking to. I personally don't really have an opinion about Rybka strength. If it was allowed in real tournaments and matches regularly we would know...just playing occasional exhibition matches its hard to tell. Computers certainly seem strong, but it would be pleasurable and informative to let one computer play serious chess for a few years and see how people adjust (preparing for it in tournaments like any other opponent). Without that its just darn hard to know how strong they are. If Rybka didn't have some humans on staff constantly updating its opening book, I would guess over the years its results would wane, so the strength of its helpers compared to the world's best must also be considered (particularly with grandmasters having access to Rybka to test how it would play against lines they are considering).
Anyway, folks around here that are experts in computer chess and master or near master at normal chess (which covers multiple people on this site and in this thread) seem like they are in a position to offer as valid an opinion on Rybka strength as the GMs you mention. Maybe not better, but maybe not worse.
-Sam
I agree with you Sam. It takes sometimes 10k games to tell
what some programs are better than others. Like you said
if we had one gm play 10k games vs. Rybka we may know
the ans.
bigo wrote:Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced. I guess all these grandmasters are just talking hyperbole. Garry kasparov said five years ago when you were saying computers play at 2400 that the top programs were in the 2700s. Kramnik said something similiar. After following you in this debate for years, i'm convinced that if God himself said computers were now 3000 you would disagree. Kasparov must have been dead right cause programs like Rebel were drawing Matches at standard time control with 2700 players 5 years ago.
Just read what I wrote. I said that a handicapped match does _not_ necessarily prove _any_ rating at all. It changes the dynamics of the game from an initial equality, to something different, where optimal play is different as well.
bigo wrote:Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced. I guess all these grandmasters are just talking hyperbole. Garry kasparov said five years ago when you were saying computers play at 2400 that the top programs were in the 2700s. Kramnik said something similiar. After following you in this debate for years, i'm convinced that if God himself said computers were now 3000 you would disagree. Kasparov must have been dead right cause programs like Rebel were drawing Matches at standard time control with 2700 players 5 years ago.
Excuse me if I try to explain. I tend to agree with both sides. From a player's perspective all the title holders and GM are correct from their practical side, Bob could easily think similar, but he doesnt, and the reason is IMO that scientifically, statistically, logically such a verdict is wrong. At least we should differentiate between the difficulties to play these machines and the resulting performance in dependence of human GM players who seldom enough are playing reasonable chess variations against specifically a machine, and the sort of objective strength of these machine with also their crass weaknesses. If these weaknesses were systematically played against then the overal record would look different. You just cant publish Elo numbers after so few games, apart from the aspect that one shouldnt apply Elo to machines the way we know that for human chess. Human chess is something completely different with relative stable personalities over years. But in CC you can create different personalities for the morning and evening play sessions. Of course directed by human operators. Etc pp.
Hi Rolf
Either I am just dumb or your command of the English language isn't very good. Because i have absolutely no ideal or understanding of what you are trying to say. I have read over what you wrote several times, ofcourse it might just be me, but I have a problem understanding nearly all your posts, please don't take it as an attack on you.
bigo wrote:Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced. I guess all these grandmasters are just talking hyperbole. Garry kasparov said five years ago when you were saying computers play at 2400 that the top programs were in the 2700s. Kramnik said something similiar. After following you in this debate for years, i'm convinced that if God himself said computers were now 3000 you would disagree. Kasparov must have been dead right cause programs like Rebel were drawing Matches at standard time control with 2700 players 5 years ago.
Excuse me if I try to explain. I tend to agree with both sides. From a player's perspective all the title holders and GM are correct from their practical side, Bob could easily think similar, but he doesnt, and the reason is IMO that scientifically, statistically, logically such a verdict is wrong. At least we should differentiate between the difficulties to play these machines and the resulting performance in dependence of human GM players who seldom enough are playing reasonable chess variations against specifically a machine, and the sort of objective strength of these machine with also their crass weaknesses. If these weaknesses were systematically played against then the overal record would look different. You just cant publish Elo numbers after so few games, apart from the aspect that one shouldnt apply Elo to machines the way we know that for human chess. Human chess is something completely different with relative stable personalities over years. But in CC you can create different personalities for the morning and evening play sessions. Of course directed by human operators. Etc pp.
Hi Rolf
Either I am just dumb or your command of the English language isn't very good. Because i have absolutely no ideal or understanding of what you are trying to say. I have read over what you wrote several times, ofcourse it might just be me, but I have a problem understanding nearly all your posts, please don't take it as an attack on you.
bigo wrote:Isn't larry in a better position to judge then you? After all what is your ratings 1700? Milov himself said on the rybka page that he believes rybka is playing at 3000 elo, also joel benjamin. And your friend Mr Anti Computer himself Roman said rybka plays stronger then any computer he has faced. I guess all these grandmasters are just talking hyperbole. Garry kasparov said five years ago when you were saying computers play at 2400 that the top programs were in the 2700s. Kramnik said something similiar. After following you in this debate for years, i'm convinced that if God himself said computers were now 3000 you would disagree. Kasparov must have been dead right cause programs like Rebel were drawing Matches at standard time control with 2700 players 5 years ago.
Just read what I wrote. I said that a handicapped match does _not_ necessarily prove _any_ rating at all. It changes the dynamics of the game from an initial equality, to something different, where optimal play is different as well.
Ok then I misunderstood but it seemed to me based on your response to Duncan that you were referring to non handicapped chess am I wrong?
duncan wrote:
regarding regular computer-human games, he said Hydra, earned a 3000 rating against humans in fourteen games.
is that correct ?
"Possible but meaningless. My first tournament, after playing 2 rounds, had me with a rating of over 2600. Because I beat two players rated just over 2200. But I was not a 2600 player in real life".
