Jeroen wrote:This is a handicap match, Bob. Without pawn f7 you are material down plus you are positionally worse. So white starts with an advantage of at least +1,25 at the beginning of the game and black needs white mistakes to draw or even win the game.
I understand that. But I have played plenty of "odds matches" against humans over the years, and I knew that I was a better player, and I was not going to let them lock things up and strangle me... the difference between a human and a computer is wide in that regard...
It does not prove superior positional undersanding of humans.
It is possible that even using the 31 piece tablebases is not going to help to avoid blocking the position because the slowest way to lose is going to allow the opponent to block the position.
Then you are simply not following the discussion. Have you ever played weaker players and given them material odds to help equalize? If so, have you _ever_ played such moves to totally lock up the pawn structure so that you can't use your superior experience and tactical ability to beat them? Computers fall into these stupid positions all the time, whether it be in odds matches or heads-up. And it clearly shows that humans have a better grasp of what is going on than the computers do...
Computers are basically not good at handicap games because they do not consider the fact that their opponent may miss something and inspite of this disadvantage Rybka won the match 2.5:1.5 when the GM could not win a single game.
Rybka proved to be superior even in closed positions when rybka saw that the GM can beat her but GM Dzindzichashvili did not see the right moves
Sorry but that is a _poor_ definition of "superior" when you see ways your opponent can beat you because you have played poorly in previous moves.
52.h5 could win the game for Dzindzi in game 3 and I understood that rybka3 could find it unlike GM Dzindzichashvili
Dzindzichashvili,R - Rybka 3, Dzindzi-Rybka handicap match 2008
[D]3rb3/4k1rp/2p1pNp1/2P1K1P1/1p1P1P1P/pP5R/P6R/8 w - - 0 1
The point is, it had played into a dead lost position. And, as is generally the case, the GM made an error. Roman is probably more likely to make errors like this than younger GM players, as age is a factor. But regardless of the opponent, even if it were Kasparov or Fischer in their prime, they still make mistakes... And the computer benefits from that more than its lack of positional understanding in these kinds of positions.
The point is that when you start without f7 pawn every move is losing so
it is not clear that rybka chose worse moves.
It is possible that not allowing the opponent to block the position could lead to faster loss.
You can say that it is better to play worse move because the opponent will not see something but at least today computers do not work like that and do not try to simulate the opponent to see what gives better practical chances.
Only a few super-GMs have a chance to win a game from time to time against the top engines. And they cost lots. In his book "Behind Deep Blue", Feng Hsiung Hsu states that one day of Gary Kasparov costs $30.000. Maybe the Anand , Kramnik and Carlsen would cost a bit less, as they're less known in the media, but we're ten years after Deep Blue. With inflation I bet they'd cost $30.000 still. Which engine writer can afford a super GM to "calibrate" his engine?
Maybe, but I think he started to mess up with 18.Kg2 already. 18.f4 Nc7 19.a4 Na6 20.Nf3 or 19...a6 20.Nf3 b5 21.Bd2 Qb6 22.a5.
I think 18.Kg2 was just a useless move at this point.
George Tsavdaris wrote:Rybka had black and didn't had the f-Pawn from the start, which is a HUGE handicap.
It is, only we can't tell how huge it is exactly. Therefore it is a complete waste of time to play handicap matches.
Human-engine matches should have the purpose of calibrating the engine rating lists. Handicap matches don't serve this purpose.
Very well said Gabor,exactly my understanding of this delicate issue
BTW,the GM has to earn some bucks sometimes even if he makes an ass out of himself....
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
Jeroen wrote:This is a handicap match, Bob. Without pawn f7 you are material down plus you are positionally worse. So white starts with an advantage of at least +1,25 at the beginning of the game and black needs white mistakes to draw or even win the game.
I understand that. But I have played plenty of "odds matches" against humans over the years, and I knew that I was a better player, and I was not going to let them lock things up and strangle me... the difference between a human and a computer is wide in that regard...
It does not prove superior positional undersanding of humans.
It is possible that even using the 31 piece tablebases is not going to help to avoid blocking the position because the slowest way to lose is going to allow the opponent to block the position.
Then you are simply not following the discussion. Have you ever played weaker players and given them material odds to help equalize? If so, have you _ever_ played such moves to totally lock up the pawn structure so that you can't use your superior experience and tactical ability to beat them? Computers fall into these stupid positions all the time, whether it be in odds matches or heads-up. And it clearly shows that humans have a better grasp of what is going on than the computers do...
Computers are basically not good at handicap games because they do not consider the fact that their opponent may miss something and inspite of this disadvantage Rybka won the match 2.5:1.5 when the GM could not win a single game.
Rybka proved to be superior even in closed positions when rybka saw that the GM can beat her but GM Dzindzichashvili did not see the right moves
Sorry but that is a _poor_ definition of "superior" when you see ways your opponent can beat you because you have played poorly in previous moves.
52.h5 could win the game for Dzindzi in game 3 and I understood that rybka3 could find it unlike GM Dzindzichashvili
Dzindzichashvili,R - Rybka 3, Dzindzi-Rybka handicap match 2008
[D]3rb3/4k1rp/2p1pNp1/2P1K1P1/1p1P1P1P/pP5R/P6R/8 w - - 0 1
The point is, it had played into a dead lost position. And, as is generally the case, the GM made an error. Roman is probably more likely to make errors like this than younger GM players, as age is a factor. But regardless of the opponent, even if it were Kasparov or Fischer in their prime, they still make mistakes... And the computer benefits from that more than its lack of positional understanding in these kinds of positions.
The point is that when you start without f7 pawn every move is losing so
it is not clear that rybka chose worse moves.
It is possible that not allowing the opponent to block the position could lead to faster loss.
You can say that it is better to play worse move because the opponent will not see something but at least today computers do not work like that and do not try to simulate the opponent to see what gives better practical chances.
Uri
If you believe that, then what is the purpose of playing the match or even accepting those conditions to play? There must be some idea that the handicap is not fatal given the program's skill level. And if so, the program should try to use that skill level effectively and not voluntarily withdraw into a locked up drawish position...
As far as your last comment goes, it is not true for _all_ of us. I have had code to try to avoid those kinds of positions for years, for the reason that I want tactical positions against humans, not passive/blocked ones. Other programs have done the same.
Maybe, but I think he started to mess up with 18.Kg2 already. 18.f4 Nc7 19.a4 Na6 20.Nf3 or 19...a6 20.Nf3 b5 21.Bd2 Qb6 22.a5.
I think 18.Kg2 was just a useless move at this point.
I think Kg2 was ok. If you want to play h4-h5, you'd like the rook on the h-file and that is one way to get it over there. There is no big rush here and white would like to have both rooks on the kingside when things start to happen...
fern wrote:Clearly this guy, no enough contented with the advantage given to him -a pawn and the whites all the time- consider neccesary to dug in deep.
What a shame!
Fernando
Not sure what you mean. He played exactly the kind of plan that computers fail to understand, even if he did overlook a winning move that was fairly deep. There is usually one key difference between the computer's side and the human's side in these kinds of positions... the human has an effective breaking move he can play when he chooses, and can do long-term preparation for that. If the computer can't see something in its search, it doesn't exist, so it doesn't know how to prepare for a single pawn break and lets it happen.
Perhaps we have some semantic differences. I see as digging-in what you call anti-computer strategy.
No that digging-in is bad un principle, but perhaps you could expect something else of a SGM.
The extreme case of it is our colombian friend here, that erects a wall and then waits. It works, but its is scarcely chess IMHO.
My best
Fernando
fern wrote:Perhaps we have some semantic differences. I see as digging-in what you call anti-computer strategy.
No that digging-in is bad un principle, but perhaps you could expect something else of a SGM.
The extreme case of it is our colombian friend here, that erects a wall and then waits. It works, but its is scarcely chess IMHO.
My best
Fernando
Actually, I thought Roman played exceedingly well. True, he blundered, and this cost him, but he obtained excellent positions and built on them. He didn't merely try to lock things up and draw, he positioned himself in the direction of the winning line. He merely prevented the program from playing its game, and that's a part of any good chess strategy.
Albert
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
fern wrote:Perhaps we have some semantic differences. I see as digging-in what you call anti-computer strategy.
No that digging-in is bad un principle, but perhaps you could expect something else of a SGM.
The extreme case of it is our colombian friend here, that erects a wall and then waits. It works, but its is scarcely chess IMHO.
My best
Fernando
Actually, I thought Roman played exceedingly well. True, he blundered, and this cost him, but he obtained excellent positions and built on them. He didn't merely try to lock things up and draw, he positioned himself in the direction of the winning line. He merely prevented the program from playing its game, and that's a part of any good chess strategy.
Albert
I agree. Problem with Roman is he is getting older and that leads to more and more tactical errors as it takes a _lot_ of concentration to play one of these tactical beasts... I've known him for many years now, and he's a good computer opponent. He understands them pretty well (partly my fault probably as we have had many conversations about how / why they do or don't do certain things.) But I noticed that over the past 10 years his games are less precise today than when I first met him on ICC in 1995...
That's just another advantage of computers, they are always "in their prime"...