Okay, I've given Houdini 40 minutes and it has completed depth 34. It seems to have no idea what's going on (1. Qxe5 fxe5 2. Rf1 and Black is zugzwanged). Nothing new there, it has always been possible to set up positions that computers don't get (especially zugzwang themes).
Another interesting position, not sure how that impacts on anything, really.
These positions are so rare in practice, that they become famous.
The position comes from a game played in 1951, Gusev vs. Averbakh, and it apparently stumped Rybka Cluster as well (see the post dated 2010-12-24 04:02 and followups).
It seems to me that it proves the point that sometimes people are able to see lines that computers won't.
The position comes from a game played in 1951, Gusev vs. Averbakh, and it apparently stumped Rybka Cluster as well (see the post dated 2010-12-24 04:02 and followups).
It seems to me that it proves the point that sometimes people are able to see lines that computers won't.
Sure, never thought any different.
However, when you play a whole game against an engine, it steals centipawns from you continually. You just don't get those sort of chances, you get crushed. That's why GM's are not falling over themselves to play the machines anymore.
p.s. Even if by pure luck the GM stumbles upon (or constructs) this sort of position (one where the machine suffers a lack of conceptual foresight), he's just going to be bullied in the next eleven games.
One swallow does not a summer make. 
I see, so you really meant "Humans have virtually no chance against good software/good hardware combinations, at any time control."
Cheers 
.
I see, so you really meant "Humans have virtually no chance against good software/good hardware combinations, at any time control."
Cheers .
Yes.
Well, at least it was an opportunity to show off a really cool position 
.
Here is the problem with the over-optimism in "humanly exclusive positional play":
1. Although a human might make a better plan than a computer, the slightest thing they miss will compromise its execution. You might play for a queenside attack but if you play one imprecise move the computer will find a tactical way to centralize his pieces and your attack is it. You would need too much precision to ever prevent the computer from escaping from a bad position. But then, it's hard enough to give it a bad position in the first place.
2. They calculate so well, that often they realize that the most promising sequence of moves is the precise execution of a plan humans would make. That's why you often find computers still playing for pawn breaks when the game is closed. This is a very important point, because it reflects that computers do fine with no real planning method.
3. Although there are probably certain positions that they would misevaluate badly (endgames/closed positions), the amount of these shouldn't be exaggerated. Yes, if a computer did make a completely impertinent move, I would expect it to be in a closed position, but it actually doesn't play those very often no matter the postition it's in. A computer will play a lot more closed positions well than closed positions poorly.
4. It is so hard for a human not to give in, at any point, with some sort of mistake or blunder. Even if it was a small, Super GM inaccuracy, that's the kind of thing that throws a win out the window when facing precise play.
The human would have to play unbelievably precisely, to execute any superior plan he came up with; he'd have to play at a 3000 level or something constantly, because at any point, the tiniest mistake will probably at least cost him any advantage he would hope to get.
In short: Their lack of positional play is just not enough of a weakness for humans to expect to beat them. Let's face it: we have hundreds of more weaknesses than they do! :)
Yes computers are very strong. What would be interesting would be to have matches where a draw would count as a win for the GM, and no opportunity to tweak the engine, nor the opening book after the match starts.
That would at least give interesting games, I think 
Yes computers are very strong. What would be interesting would be to have matches where a draw would count as a win for the GM.
In a similar vein, handicapping the engine by giving the human GM pawn odds has been tried a while ago, and Rybka has trounced Ehlvest quite convincingly...
[EDIT: here's a link to an article describing some handicapped games]
Yes, but I think draw odds is more interesting : pawn handicap positions are less familiar to the GM, so it reduces the advantage of the extra pawn.
GMs without the pressure to push for a win may be much more difficult to beat. They could also choose their opening variations accordingly, and the onus would be on the engine to unbalance the position.
Well, I'm just curious. But maybe it won't change anything. I don't know.
But I thought the whole thrill in setting up a match against a computer is the whole "Man vs Machine" hype -- who is the better chess player? But unfortunately, we know the answer to that question.
Here is the problem with the over-optimism in "humanly exclusive positional play":
1. Although a human might make a better plan than a computer, the slightest thing they miss will compromise its execution. You might play for a queenside attack but if you play one imprecise move the computer will find a tactical way to centralize his pieces and your attack is it. You would need too much precision to ever prevent the computer from escaping from a bad position. But then, it's hard enough to give it a bad position in the first place.
2. They calculate so well, that often they realize that the most promising sequence of moves is the precise execution of a plan humans would make. That's why you often find computers still playing for pawn breaks when the game is closed. This is a very important point, because it reflects that computers do fine with no real planning method.
3. Although there are probably certain positions that they would misevaluate badly (endgames/closed positions), the amount of these shouldn't be exaggerated. Yes, if a computer did make a completely impertinent move, I would expect it to be in a closed position, but it actually doesn't play those very often no matter the postition it's in. A computer will play a lot more closed positions well than closed positions poorly.
4. It is so hard for a human not to give in, at any point, with some sort of mistake or blunder. Even if it was a small, Super GM inaccuracy, that's the kind of thing that throws a win out the window when facing precise play.
The human would have to play unbelievably precisely, to execute any superior plan he came up with; he'd have to play at a 3000 level or something constantly, because at any point, the tiniest mistake will probably at least cost him any advantage he would hope to get.
In short: Their lack of positional play is just not enough of a weakness for humans to expect to beat them. Let's face it: we have hundreds of more weaknesses than they do! :)
All of those are reasons that humans get destroyed with standard time controls, but what happens when you play correspondance? The human gets days to contemplate a position, from the comfort of his own home, without psychological pressure involved in OTB play. It seems to me that an engine's greatest strengths over humans is their ability to play completely consistently, with near perfect tactics, and obviously without psychological pressures that humans have to deal with. In correspondance, the gap between humans and computers should be much smaller because the human's tactical error will be almost eliminated, and the advantages we have over computers should become more apparent.
I would be very interested to see a centaur match between pfren and Prawn, but it looks like it's not going to happen.
I would be very interested to see a centaur match between pfren and Prawn, but it looks like it's not going to happen.
To be fair about this, centaur play is extremely greedy of both time and resources. If it's not your thing then it's not your thing.
In tough positions, I leave the engine running for up to twelve hours. Given that I'm using a self-made, overclocked system running at 100%, I have a little horror every time I leave for work. I can't get it out of my head that I might come home to a burnt-down house.
I would be very interested to see a centaur match between pfren and Prawn, but it looks like it's not going to happen.
To be fair about this, centaur play is extremely greedy of both time and resources. If it's not your thing then it's not your thing.
In tough positions, I leave the engine running for up to twelve hours. Given that I'm using a self-made, overclocked system running at 100%, I have a little horror every time I leave for work. I can't get it out of my head that I might come home to a burnt-down house.
Yeah, I figured that he probably has better things to do than defend humanity. It would take real commitment to be able to win this match, maybe that's why you've had such an easy time with centaur chess on chess.com. Do you participate in ICCF? Surely you'd get some tough opponents there.
I would do it, but I'm sure I'd get destroyed. I don't have very good hardware, software or chess skills.
Do you participate in ICCF? Surely you'd get some tough opponents there.
The best players from this tournament, will shortly be playing in a team event on ICCF:
http://www.chess.com/tournaments/pairings.html?id=33993&round=1
Interested parties should note that we have a friendly, inclusive and reasonably active group on the site, devoted to advanced chess:
http://www.chess.com/groups/view/advanced-chess
Interested parties should note that we have a friendly, inclusive and reasonably active group on the site, devoted to advanced chess:
I dont know anything about this. You said you rarely or never overrule the computer, so is it mostly just who has the best hardware/software setup?
I can't imagine, at my level, outguessing the machine except perhaps in those fortress type positions.
Interested parties should note that we have a friendly, inclusive and reasonably active group on the site, devoted to advanced chess:
I dont know anything about this. You said you rarely or never overrule the computer, so is it mostly just who has the best hardware/software setup?
I can't imagine, at my level, outguessing the machine except perhaps in those fortress type positions.
The machine is often over-ruled in the opening, databases and cunning transpositions are the order of the day at that stage. Over-ruling a machine that sees 35 ply in a middle-game, is an altogether different kettle of fish.
Sometimes, if I have an idea, I'll show the engine my plan by playing forward a few moves. Then I let it have a long think. After that I hit the "back" button and ask it to reconsider the original position again. Once in a blue moon, human intuition causes the engine to change its mind. Quite a thrill when that happens, I can tell you.
Are you sure this position is right?
It is