According to http://www.chessgames.com/perl/chessgame?gid=1472988&kpage=1
25. Rf1 Rc7 26. Bd1 Re7 27. Bb3 a5 28 a4!!
Zugzwang!!
Humans v Houdini chess engine (Elo 3300)
Sort:
browni:
I don't think that, no matter how much time you give a human, you can eliminate the human factor. Indeed, sometimes you will sit and think about the position for 30 minutes, totally content on every possibility, but you still miss something you never even considered. The problem? You looked a lot, but at the wrong lines. Even with as much time as they need, they will still get nervous; they will still get impatient; they will still experience blindness -- they can still experience anything humans always do in chess.
Jan 8, 2012
0
#84
Elubas wrote:
browni:
I don't think that, no matter how much time you give a human, you can eliminate the human factor. Indeed, sometimes you will sit and think about the position for 30 minutes, totally content on every possibility, but you still miss something you never even considered. The problem? You looked a lot, but at the wrong lines. Even with as much time as they need, they will still get nervous; they will still get impatient; they will still experience blindness -- they can still experience anything humans always do in chess.
Okay, I realize that no amount of time will eliminate human error. I'm trying to say that it would minimize it. You don't really think I mean 30 minutes do you? I'm talking about top players spending hours on each move. One way humans are better than computers is that humans are MUCH more efficient, which means that they will find the right lines with less positions considered. Tactical error is almost eliminated in CC, because the human can take almost as much time as needed to be very confident. They will not miss very much.
All of the misconceptions posters have about engines is mind-boggling, not you Elubas, but some of the statements have been ridiculous. Hopefully not too many of mine have been.
Remember that the pre-computer world chess correspondence champ games still have considerable error when checked by computers. No matter the time control humans without computer aid can't push past 70% or so #1 picks by 3000+ computer.
Maybe 1 of those moves the computer didn't pick was actually better. But most of them give away centi pawns that will never be gotten back.
Jan 8, 2012
0
#86
waffllemaster wrote:
Remember that the pre-computer world chess correspondence champ games still have considerable error when checked by computers. No matter the time control humans without computer aid can't push past 70% or so #1 picks by 3000+ computer.
Maybe 1 of those moves the computer didn't pick was actually better. But most of them give away centi pawns that will never be gotten back.
Do you know the percentage for modern centaurs? How often do they override the computer's #1 pick?
browni3141 wrote:
waffllemaster wrote:
Remember that the pre-computer world chess correspondence champ games still have considerable error when checked by computers. No matter the time control humans without computer aid can't push past 70% or so #1 picks by 3000+ computer.
Maybe 1 of those moves the computer didn't pick was actually better. But most of them give away centi pawns that will never be gotten back.
Do you know the percentage for modern centaurs? How often do they override the computer's #1 pick?
No, but that would be a very interesting statistic.
Because pre-computer match-up are relatively low, (and as pointed out above, the rarity of humans-know-better position is so rare as to make them famous) I just assumed these situations are negligible.
Jan 8, 2012
0
#88
waffllemaster wrote:
browni3141 wrote:
waffllemaster wrote:
Remember that the pre-computer world chess correspondence champ games still have considerable error when checked by computers. No matter the time control humans without computer aid can't push past 70% or so #1 picks by 3000+ computer.
Maybe 1 of those moves the computer didn't pick was actually better. But most of them give away centi pawns that will never be gotten back.
Do you know the percentage for modern centaurs? How often do they override the computer's #1 pick?
No, but that would be a very interesting statistic.
Because pre-computer match-up are relatively low, (and as pointed out above, the rarity of humans-know-better position is so rare as to make them famous) I just assumed these situations are negligible.
That statement of the rarity of humans-know-better positions doesn't seem right. There aren't very many positions where the average player can beat the engine, but what about the elite players?
Here are some things quoted from a correspondance GM Stephen Ham:
"strong players know from experience when to use their engine(s) and when to use it less, if at all"
"Dont' be influenced by the computer. It's common for my moves to not be top computer selections while my personal evaluations are always trusted over engine evalutions"
"top-level correspondence chess is not about who has the best computer hardware/software ... many chess engine fanatics with the fastest octal hardware and best programs fare poorly in serious correspondence chess."
He seems to think that humans often know what's up better than computers do.
browni3141 wrote:
Elubas wrote:
browni:
I don't think that, no matter how much time you give a human, you can eliminate the human factor. Indeed, sometimes you will sit and think about the position for 30 minutes, totally content on every possibility, but you still miss something you never even considered. The problem? You looked a lot, but at the wrong lines. Even with as much time as they need, they will still get nervous; they will still get impatient; they will still experience blindness -- they can still experience anything humans always do in chess.
Okay, I realize that no amount of time will eliminate human error. I'm trying to say that it would minimize it. You don't really think I mean 30 minutes do you? I'm talking about top players spending hours on each move. One way humans are better than computers is that humans are MUCH more efficient, which means that they will find the right lines with less positions considered. Tactical error is almost eliminated in CC, because the human can take almost as much time as needed to be very confident. They will not miss very much.
All of the misconceptions posters have about engines is mind-boggling, not you Elubas, but some of the statements have been ridiculous. Hopefully not too many of mine have been.
I think there comes a point where the human is just sick of calculating and wants to just play a move; their brain overflows with ideas, to the point where thinking of a certain amount of time might actually hurt their ability to see the big picture. Of course, I could understand putting in 20 minutes, then coming back to it later, but in any case, I think you are overestimating how much humans would benefit from the time given to them in correspondence. I am not a grandmaster, but if you are trying to think of this as a linear pattern, I could try experimenting by thinking for four hours on a move, seeing how that works. Even given that amount of time, I would not necessarily play like a grandmaster -- if my mind burns out at 5 moves ahead, I don't think more time will alleviate that.
If you tried to compensate a blunderous 1000 player by giving them ten hours for each move, do you really think they would stop blundering?
On the same topic, I visited a particular club once or twice a month were all the players were ~200 rating points better than me. Fantastically I would tend to break even most nights... but I was taking the games much more seriously (no blitz), pouring over every move, every possibility I could. It was literally physically tiring and I couldn't bring myself to visit more frequently than I did.
Which is to say, I do think extra time (read taking extra care) can increase your level of play. But like Elubas is saying, the brain only has so much stamina, and eventually extra thinking does no good.
browni3141 wrote:
That statement of the rarity of humans-know-better positions doesn't seem right. There aren't very many positions where the average player can beat the engine, but what about the elite players?
That statement of the rarity of humans-know-better positions doesn't seem right. There aren't very many positions where the average player can beat the engine, but what about the elite players?
Here are some things quoted from a correspondance GM Stephen Ham:
"strong players know from experience when to use their engine(s) and when to use it less, if at all"
"Dont' be influenced by the computer. It's common for my moves to not be top computer selections while my personal evaluations are always trusted over engine evalutions"
"top-level correspondence chess is not about who has the best computer hardware/software ... many chess engine fanatics with the fastest octal hardware and best programs fare poorly in serious correspondence chess."
He seems to think that humans often know what's up better than computers do.
I tend to believe this too. However I wonder how much different a game is when played against a computer from move one.
The famous quote:
I can see the combinations as well as Alekhine, but I cannot get into the same positions.
--- Spielmann
Seems to suggest that understanding isolated positions doesn't guarantee sucess in a full length game (which makes sense to me).
Jan 8, 2012
0
#92
Norton Jacobi won the Australian CC championship back in about 1979. He said that he was spending approximately 10 hours per move. He had books full of anaysis. His OTB strength would have been 1900~2000. Ahhh... the olden days.
Generalisation and opinion goes so far but . . .
My original question was - has a human beaten Houdini - or any of the top engines (Stockfish, Rybka) - recently in the 3000 + Elo era? If so, please post the score so we can all repeat the moves and win as well.
NB: OTB at any time control preferred to correspondence chess because ... well we all know that engines can beat other engines (although even a correspondence game would be better than nowt).
OR
Can someone give a specific example (position + machine) of a player successfully overriding an engine's choice in the middle game. The example given previously (Gusev v Averbakh) is a great example but it a positon like that would occur about once in a million games (if that).
Nothing less than the honour of the human race is at stake here!
Jan 9, 2012
0
#94
Computers are man made creations, so I don't think any honour has been lost.
Anyone that has doubts, and believe a patzer can be a super-GM, can simply register himself for CC chess at ICCF. He will find out that despite having a latest super-dooper machinery, he will be mercilessly beaten by skilled correspondence players.
It's true, to some extent, that the engines have changed the fundamentals of correspondence chess, because they have gained in strength. Some strong CC players, like GM Ulf Andersson (once upon a time #4 in the world in normal chess, as well as #1 in the ICCF rating lists) have withdrawed because of these new conditions. But STILL, the engines are used by all top correspondence players for the mere purpose of checking the tactical validity of their positional ideas. Every single player in the top-ten ICCF list has been in the past a strong OTB player. Why? Well, this is an easy question.
PEP can test his super-dooper machinery by simply applying for an ICCF tournament. I can safely bet he will be sorely dissapointed... 
When did computer assistance in correspondence chess start being officially sanctioned?
Look, as far as I can tell no one is disputing that a skilled player with a computer can beat a patzer with a computer. That was not the original question! ("Has a Grandmaster ever beaten it [Houdini] in any game?" doesn't imply computer use by the grandmaster.) So, to repeat, has any human (alone!) beaten Houdini?
PEP can test his super-dooper machinery by simply applying for an ICCF tournament. I can safely bet he will be sorely disappointed...
That may be the case but that has nothing to do with the original question. Take your crumbs of comfort where you find them, while you still can. The day has passed for men providing serious competition for machines.
A super-GM can still comfortably beat a machine, if he has enough time to check his lines for tactical flaws (say 2 days per move) - quite simply because he does understand chess, while the machine only understands ones and zeroes. But I guess we will never find out, because there is no motivation. If someone offers Magnus several thousand bucks to beat a machine, I do not have the slightest doubt about the result. The problem is that silicon fans will claim that the actual winner was not MC, but a centaur.
Then we are back in the realms of belief, not knowing.
How long can it take a computer to calculate a winning line for an endgame tablebase entry? How long would it take a human to do the same thing? If such a position arose in a game between human and computer, would the human stand a chance of finding all the winning moves?
The earlier position earlier with (1. Qe5 fe5 2. Rf1) is a great example.
Houdini 1.5 and Stockfish got stuck on Qd4, and Qa3 which they gave as about equal. Neither considered Qe5.
I also tried analysing after Rf1, This position is not 'obviously' won for white. In fact, after 500, 000, 000 nodes Houdini thought it was still advantageous for Black! It started at White = -5.4 and only slowly came down to White = -0.6.
In the game black played 2. . . Rc8 and resigned 11 moves later, so there was still a lot of play in the position. Houdini hadn't even considered 2. . . Rc8 but instead focused on 2. .. Rc7 and 2. ..a6.
Does anyone know the best moves for each side after 2. Rf1? How many moves is it before the position is obviously resignable?