Contemptuous

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Harvey Williamson, bob

carldaman
Posts: 1537
Joined: Sat Jun 02, 2012 12:13 am

Re: Contemptuous

Post by carldaman » Wed Dec 25, 2013 1:30 am

Top engines (or dedicated machines) losing to amateurs (and LT is a quite a bit stronger than your regular amateur) is nothing new.

Back in 2001, a book was published by the somewhat humorous name of "A Psychiatrist Matches Wits with Fritz", written by Ernest F. Pecci, MD.

It is a very detailed diagram-filled hard-cover book, with a foreword by none other than Garry Kasparov himself (!), and it covers the author's winning games and strategies against the top program of that time, Fritz. (I have the book, but it is hard to find and not very well-known. It sells for a lot of money on Amazon.)

About 10-15 years before that, when dedicated machines were already giving human opponents a hard time, an obscure A-player showed that it was possible to defeat the computers while giving them Queen odds (!!), by exploiting the machines' horizon effect weakness with slowly built-up kingside attacks that involved opening up the h-file later in the game. An article was written about this in Chess Life magazine (I think) and the moves played in the published games could be reproduced by others.

Anti-computer specialists have always been around. Before engines became really strong (back in the 90's), it was really possible to even defeat them tactically in open games, through various sacrifices that exploited poor king safety.

I believe Lyudmil is doing everyone a favor by publishing his games and related findings. It would be very wrong and detrimental to future progress to assume that exploitable weaknesses don't exist in today's top engines.

Regards,
CL

User avatar
M ANSARI
Posts: 3278
Joined: Thu Mar 16, 2006 6:10 pm

Re: Contemptuous

Post by M ANSARI » Wed Dec 25, 2013 8:12 am

carldaman wrote:Top engines (or dedicated machines) losing to amateurs (and LT is a quite a bit stronger than your regular amateur) is nothing new.

Back in 2001, a book was published by the somewhat humorous name of "A Psychiatrist Matches Wits with Fritz", written by Ernest F. Pecci, MD.

It is a very detailed diagram-filled hard-cover book, with a foreword by none other than Garry Kasparov himself (!), and it covers the author's winning games and strategies against the top program of that time, Fritz. (I have the book, but it is hard to find and not very well-known. It sells for a lot of money on Amazon.)

About 10-15 years before that, when dedicated machines were already giving human opponents a hard time, an obscure A-player showed that it was possible to defeat the computers while giving them Queen odds (!!), by exploiting the machines' horizon effect weakness with slowly built-up kingside attacks that involved opening up the h-file later in the game. An article was written about this in Chess Life magazine (I think) and the moves played in the published games could be reproduced by others.

Anti-computer specialists have always been around. Before engines became really strong (back in the 90's), it was really possible to even defeat them tactically in open games, through various sacrifices that exploited poor king safety.

I believe Lyudmil is doing everyone a favor by publishing his games and related findings. It would be very wrong and detrimental to future progress to assume that exploitable weaknesses don't exist in today's top engines.

Regards,
CL

I know about how anti computer specialists were able to take advantage of earlier engines, and I would consider myself one of them. I could score way better than much stronger players could. These type of weaknesses though simply DO NOT exist anymore with any of the top engines using very average desktop hardware. You cannot do what you could do many years ago, things have changed by such a huge margin that it is no longer possible to beat a well setup top engine unless you skew some things in the favor of the human ... eg. forcing the engine into an opening that is losing and the human has already memorized by rote which moves to play after the engine has a lost position. Sometimes you might be able to get a rare draw by closing the position and exchanging a lot of pieces so that the game ends in a draw ... but a very simple change in the engine software avoiding these type of positions makes it impossible for the human to play and win. If you look at a top human GM game and follow it with an engine, it is very clear that a human will always not be aware of all the tactical details of a position, especially in complicated positions. The human needs to translate the position in way where he can understand it and that usually is not the most accurate tactical move. And with the new engines one little mistake is all that is needed.

This is why I am very skeptical when someone mentions that he can regularly beat SF or Houdini and the reason is simple ... there are a lot stronger players that have tried to do the same (Anand, Kramnik, Carlsen, Aronian, Nakamura) and they publicly have stated that it is not even close. So I can believe an occasional draw is possible by a human of even average strength, but to continually be able to beat these engines ... sorry ... no way in hell unless some manipulation is done.

overlord
Posts: 198
Joined: Sun Jun 03, 2012 4:46 pm
Location: Trinec, Czech Republic

Re: Contemptuous

Post by overlord » Wed Dec 25, 2013 8:35 am

You cannot say that engines analysis of human games reveals that men doesn not have any chance against engines. Everything is wrong... The only point where I agree with you is king safety (it was improved very drastically). Engines still don´t undestand closed positions. Most of them enjoy pieces exchanges (small exception is Rybka/Houdini family). There are many openings where you have good drawing chances as white: classical queens gambit declined, Tarrash defence (not Schara-Hennig gambit), Nimzo indian (some lines in Capablanca system), generally French defence (the most difficult ofr engines), London system of Caro-Kann, main lines of Catalan systém and Grunfeld (in this case you need to have very good theoretical knowledge). If you want to win, the only one thing that you need is kings indian (even playing black). Computers play suprisingly well Samisch (although I play it against human with good results, I lost almost all games in this systém against different engines). Sharp bayonette systém is also not a proper choice (you need to have good opening knowledge, moreover positions are extremly complicated) - so it could be good choice for Radjabov or Shirov :) For us, normal human beings there is the best possibility classical kings defence where knight goes on e1 with minor attack. Without extensive book engines don´t know what to do (Houdini and Stockfish improved in this point well). There were funny positions where even Rybka 4 could play g4 and start attack against white king, but it played totally crazy h4 and then the fish just waited for its death. Moreover, engines try to prevent minor attack playing stupid b6. In reality, this move is something like christmas present for white :)

Lyudmil Tsvetkov
Posts: 6031
Joined: Tue Jun 12, 2012 10:41 am

Re: Contemptuous

Post by Lyudmil Tsvetkov » Wed Dec 25, 2013 10:27 am

carldaman wrote:Top engines (or dedicated machines) losing to amateurs (and LT is a quite a bit stronger than your regular amateur) is nothing new.

Back in 2001, a book was published by the somewhat humorous name of "A Psychiatrist Matches Wits with Fritz", written by Ernest F. Pecci, MD.

It is a very detailed diagram-filled hard-cover book, with a foreword by none other than Garry Kasparov himself (!), and it covers the author's winning games and strategies against the top program of that time, Fritz. (I have the book, but it is hard to find and not very well-known. It sells for a lot of money on Amazon.)

About 10-15 years before that, when dedicated machines were already giving human opponents a hard time, an obscure A-player showed that it was possible to defeat the computers while giving them Queen odds (!!), by exploiting the machines' horizon effect weakness with slowly built-up kingside attacks that involved opening up the h-file later in the game. An article was written about this in Chess Life magazine (I think) and the moves played in the published games could be reproduced by others.

Anti-computer specialists have always been around. Before engines became really strong (back in the 90's), it was really possible to even defeat them tactically in open games, through various sacrifices that exploited poor king safety.

I believe Lyudmil is doing everyone a favor by publishing his games and related findings. It would be very wrong and detrimental to future progress to assume that exploitable weaknesses don't exist in today's top engines.

Regards,
CL
When I think now, Carl, a very good suggestion to the Stockfish team: a bonus for opening the h file when the enemy king has castled short (or the a file when it castles long), maybe some 20-30cps, or the same penalty when the enemy attacks you along those files. Stockfish lost so many games because the h file was opened...

Lyudmil Tsvetkov
Posts: 6031
Joined: Tue Jun 12, 2012 10:41 am

Re: Contemptuous

Post by Lyudmil Tsvetkov » Wed Dec 25, 2013 10:29 am

M ANSARI wrote:
carldaman wrote:Top engines (or dedicated machines) losing to amateurs (and LT is a quite a bit stronger than your regular amateur) is nothing new.

Back in 2001, a book was published by the somewhat humorous name of "A Psychiatrist Matches Wits with Fritz", written by Ernest F. Pecci, MD.

It is a very detailed diagram-filled hard-cover book, with a foreword by none other than Garry Kasparov himself (!), and it covers the author's winning games and strategies against the top program of that time, Fritz. (I have the book, but it is hard to find and not very well-known. It sells for a lot of money on Amazon.)

About 10-15 years before that, when dedicated machines were already giving human opponents a hard time, an obscure A-player showed that it was possible to defeat the computers while giving them Queen odds (!!), by exploiting the machines' horizon effect weakness with slowly built-up kingside attacks that involved opening up the h-file later in the game. An article was written about this in Chess Life magazine (I think) and the moves played in the published games could be reproduced by others.

Anti-computer specialists have always been around. Before engines became really strong (back in the 90's), it was really possible to even defeat them tactically in open games, through various sacrifices that exploited poor king safety.

I believe Lyudmil is doing everyone a favor by publishing his games and related findings. It would be very wrong and detrimental to future progress to assume that exploitable weaknesses don't exist in today's top engines.

Regards,
CL

I know about how anti computer specialists were able to take advantage of earlier engines, and I would consider myself one of them. I could score way better than much stronger players could. These type of weaknesses though simply DO NOT exist anymore with any of the top engines using very average desktop hardware. You cannot do what you could do many years ago, things have changed by such a huge margin that it is no longer possible to beat a well setup top engine unless you skew some things in the favor of the human ... eg. forcing the engine into an opening that is losing and the human has already memorized by rote which moves to play after the engine has a lost position. Sometimes you might be able to get a rare draw by closing the position and exchanging a lot of pieces so that the game ends in a draw ... but a very simple change in the engine software avoiding these type of positions makes it impossible for the human to play and win. If you look at a top human GM game and follow it with an engine, it is very clear that a human will always not be aware of all the tactical details of a position, especially in complicated positions. The human needs to translate the position in way where he can understand it and that usually is not the most accurate tactical move. And with the new engines one little mistake is all that is needed.

This is why I am very skeptical when someone mentions that he can regularly beat SF or Houdini and the reason is simple ... there are a lot stronger players that have tried to do the same (Anand, Kramnik, Carlsen, Aronian, Nakamura) and they publicly have stated that it is not even close. So I can believe an occasional draw is possible by a human of even average strength, but to continually be able to beat these engines ... sorry ... no way in hell unless some manipulation is done.
You are very stubborn, Mr. Ansari, but you should open your eyes: Stockfish DD is a weak, ancient engine. :shock:

Post Reply