Game Over: Kasparov and the Machine

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
User avatar
Cyberchess
Full Member
Posts: 658
Joined: Wed Jan 08, 2014 6:10 pm

Game Over: Kasparov and the Machine

Post by Cyberchess »

Greetings fellow computer schach enthusiasts. I recently watched this excellent documentary on the Garry Kasparov vs. IBM’s Deep Blue 1996 match in NYC.

[youtube]https://www.youtube.com/watch?v=FBzI7y8VNCA[/youtube]

Enjoy!
John
User avatar
Cyberchess
Full Member
Posts: 658
Joined: Wed Jan 08, 2014 6:10 pm

Post by Cyberchess »

Here is a direct link as the YouTube embed doesn't seem to be working here:

https://www.youtube.com/watch?v=FBzI7y8VNCA
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Funny how Kasparov made such a huge deal at the time regarding Deeper Blue's 37. Be4 in game 2. This was allegedly a "turning point" in the match from which Kasparov - by his own admission - never recovered. It is well documented in this video as it is in Daniel King's book that I own.

The thing is, both modestly priced commercial engines that I own (HIARCS 13 and Fritz 13), both prefer Be4 in that position and by a clear margin over all other moves. And on my Core i7 desktop, HIARCS 13 running on a single core comes to this conclusion in around 40 seconds.

It isn't a huge stretch of the imagination to conclude that if two mass market commercial engines of today can manage this even on an entry level laptop well within a classical time control, then a highly specialised piece of chess software painstakingly tweaked by world-calibre GMs and other specialists and running on hardware the size of a large refrigerator might have achieved something similar in the late 90s - without any direct, real-time human intervention (as was strongly intimated).

So as much as we can all admire Kasparov's legendary achievements, my bet has always been that everything was above board at all times and that Kasparov simply underestimated the ability of chess programmers to implement practical, strategic-like knowledge into their programs.

Maybe Kasparov was blinded by the early engines whose play was very clinical. I no longer have any engines earlier than Fritz 13 and HIARCS 13, but my bet is these two engines at least were favouring the superior Be4 long before the existence of my versions.
Chess is like painting the Mona Lisa whilst walking through a minefield.
User avatar
Cyberchess
Full Member
Posts: 658
Joined: Wed Jan 08, 2014 6:10 pm

Post by Cyberchess »

When Kasparov had won the first match in Philadelphia, he arrogantly asserted that he would never lose a match to a computer. I remember thinking that Kasparov had only displayed his ignorance of computer science advancement with this contemptuous proclamation. While I hadn’t expected it to occur so soon, I realized that it was only a matter of time before man’s superiority on the chessboard would simply cease.
Kasparov was so distraught over Deep Blue’s long-term sacrificial move, 37. Be4 in game #2, he may have overlooked subsequent drawing chances due to Deep Blue’s erroneous king play. There is an excellent article on this game here:

http://www.wired.com/playbook/2012/09/d ... puter-bug/

I had opportunity to speak with GM Joel Benjamin a few years after the match, and he informed me that the entire Deep Blue team had been dissolved immediately following the match. This is indicative of IBM’s unwillingness to allow a rematch, quite unlike the sporting Kasparov of ’96.

http://www.cartoonstock.com/directory/c ... master.asp
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Well when I went to the article to which the article in your first link refers, it stated that the "random failsafe" bug occurred in the first game (44. Rd1). That would make sense as it was a pointless move in an already lost position. Though I am no psychologist and don't know how Kasparov's mind works, if I had just crushed a computer program and it made a pointless move when it was already lost, I'd just put this down to the silly way in which every program I have ever played reacts when it is completely lost and does not have a "resign" function. It certainly would not have caused me even the slightest amount of mental trauma, let alone any suspicion whatsoever.

By the accounts I have read, it was the second game which caused all the grief, and that move was not caused by a bug - it was a genuine move.
Chess is like painting the Mona Lisa whilst walking through a minefield.
User avatar
fourthirty
Full Member
Posts: 763
Joined: Fri Dec 06, 2013 8:46 pm
Location: San Francisco

Post by fourthirty »

Deep Blue overview paper, published four (4) years after the 1997 Kasparov match:

http://sjeng.org/ftp/deepblue.pdf
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Nice find. Thanks for the link. So 12 ply full width capability in 3 minutes. Hugely Impressive for 1997. We were all in awe of the Super Constellation when it came out in 85. It would always complete 5 plies and be doing the sixth on a full board on tournament level. And that program beats me most of the time...
Chess is like painting the Mona Lisa whilst walking through a minefield.
User avatar
Cyberchess
Full Member
Posts: 658
Joined: Wed Jan 08, 2014 6:10 pm

Post by Cyberchess »

The Super Connie made her debut here in the states back in ’84. She was dubbed “The Blitz Monster,” but I think slow chess was her forte. Heck, she even outplayed Diane Savereide in this game:

http://www.schach-computer.info/wiki/in ... stellation

Miss Savereide is a 5 times US Women’s Champion, and was ranked among the world’s top 10 female players in the ‘80s. Here is a game in which she defeated GM Tony Miles:

http://www.edochess.ca/batgirl/Savereide.html

Getting back to Deep Blue vs. Kasparov, in retrospect, I think that Kasparov should have insisted that the computer be displayed in plain view of the audience, with no LAN cable, internet connection, WIFI, etc. Also, there should have been no tweaking of the custom search chip, program alterations, etc. while the match was in progress. In this way his mind would have been free from concern over possible human intervention, pruning and so forth. The TD wouldn’t even allow Kasparov to speak to his mom during the match, so there is no reason for the IBM team to have access to Deep Blue from behind locked doors in my opinion. Personally, I highly doubt that I.B.M. would risk the bad PR if someone on the team were to reveal any chicanery, but the point is that Kasparov was obviously distraught over the possibility of such.
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Cyberchess wrote:Getting back to Deep Blue vs. Kasparov, in retrospect, I think that Kasparov should have insisted that the computer be displayed in plain view of the audience, with no LAN cable, internet connection, WIFI, etc. Also, there should have been no tweaking of the custom search chip, program alterations, etc. while the match was in progress. In this way his mind would have been free from concern over possible human intervention, pruning and so forth. The TD wouldn’t even allow Kasparov to speak to his mom during the match, so there is no reason for the IBM team to have access to Deep Blue from behind locked doors in my opinion. Personally, I highly doubt that I.B.M. would risk the bad PR if someone on the team were to reveal any chicanery, but the point is that Kasparov was obviously distraught over the possibility of such.
Agreed. Apart from anything else, years later, when Kramnick played Fritz 10, he was given access to the program well in advance of the match and what he played in that match was exactly what he had on his PC during his preparation.

Kasparov was perfectly correct when he stated that in chess, preparing for specific opponents is an integral part of it all. It even has a profound effect at club level, where the ratio of sub-optimal moves and blunders compared to good moves is much higher than at Master level, let alone Super GM level.

There is no doubt in my mind that Kasparov was stronger than Deep Blue at the time. Apart from anything else, the match was too short to really mean anything statistically and the rest days were inadequate. I agree with Kasparov when he said that there need to be more rest days when playing computers.

Even in my own case, I played my brother's Cosmos over the weekend. Now that machine is significantly stronger than me - probably by a good 200 ELO or more - and my brain bleeds then blows a fuse when I try to play it on tournament level. You cannot make one single blunder when playing a machine like that. One tiny mistake and you are gone. I summoned all my inner strength just to manage a draw on Sunday. Four days later I am still mentally drained and can't calculate properly. I think it damaged some brain cells :wink:

And with IBM declining a re-match...it reminds me in some ways of the old USCF chess computer rating controversies. Such as Par Excellence. Take the result and run.
Chess is like painting the Mona Lisa whilst walking through a minefield.
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Cyberchess wrote:The Super Connie made her debut here in the states back in ’84. She was dubbed “The Blitz Monster,” but I think slow chess was her forte. Heck, she even outplayed Diane Savereide in this game:

http://www.schach-computer.info/wiki/in ... stellation
Don't forget though that back in 1984, strong players rightfully had a lot of contempt for micro computers. Either that or they simply had little to no experience and went the other way, showing too much respect. If you sat Diane down for a match against a Super Connie today, the machine would be very lucky to even manage a solitary draw.
Chess is like painting the Mona Lisa whilst walking through a minefield.
User avatar
Cyberchess
Full Member
Posts: 658
Joined: Wed Jan 08, 2014 6:10 pm

Post by Cyberchess »

Monsieur Plastique wrote:
Cyberchess wrote:The Super Connie made her debut here in the states back in ’84. She was dubbed “The Blitz Monster,” but I think slow chess was her forte. Heck, she even outplayed Diane Savereide in this game:

http://www.schach-computer.info/wiki/in ... stellation
Don't forget though that back in 1984, strong players rightfully had a lot of contempt for micro computers. Either that or they simply had little to no experience and went the other way, showing too much respect. If you sat Diane down for a match against a Super Connie today, the machine would be very lucky to even manage a solitary draw.
Funny thing is nobody really seems to know what Diane is doing these days. She gave up chess to pursue a career as a NASA computer programmer sometime in the latter part of the ‘80s, I believe. Perhaps she encountered Marty Hirsch while there.

http://chessprogramming.wikispaces.com/Marty+Hirsch
User avatar
Cyberchess
Full Member
Posts: 658
Joined: Wed Jan 08, 2014 6:10 pm

Post by Cyberchess »

fourthirty wrote:Deep Blue overview paper, published four (4) years after the 1997 Kasparov match:

http://sjeng.org/ftp/deepblue.pdf
Thank you for the technical abstract, Greg.

It turns out that Garry overlooked a draw by repetition in the second game of the 1997 rematch played in New York. Had he realized this at the time, the outcome of the match would have been a tie rather than a win for Deep Blue.

http://www.chesscorner.com/games/deepblue/dblue2.htm
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Cyberchess wrote:The Super Connie made her debut here in the states back in ’84. She was dubbed “The Blitz Monster,” but I think slow chess was her forte.
I think we tend to look at these old machines with rose-coloured glasses. Here is an example from a game I played between the Super Constellation program - level 7 tournament 40 in 2 - and my Nintendo only yesterday. As the young generation might characterise this in internet speak, "Kings Gambit Fail". The little Nintendo won the associated 6 game match 6 points to nil.

[Event "40 moves in 2 hours"]
[Site "?"]
[Date "2014.02.28"]
[Round "?"]
[White "Novag Super Constellation"]
[Black "Nintendo DS Fritz"]
[Result "0-1"]
[ECO "C30"]
[PlyCount "54"]

1. e4 e5 2. f4 Bc5 3. Nf3 Nc6 4. fxe5 d6 5. exd6 Qxd6 6. Nc3 Nf6 7. Bb5 O-O 8.
Bxc6 Qxc6 9. d3 Bg4 10. Bf4 Nh5 11. Bd2 Qd6 12. Nd5 Rae8 13. Be3 f5 14. Bxc5
Qxc5 15. Qe2 fxe4 16. dxe4 c6 17. b4 Qd6 18. Nc3 Qxb4 19. Qe3 Nf4 20. Kf2 Nxg2
21. Kxg2 Bxf3+ 22. Kh3 Qa5 23. Rhb1 Qh5+ 24. Kg3 Re5 25. Qxa7 Bh1 26. h3 Qg5+
27. Kh2 Qg2# 0-1
Chess is like painting the Mona Lisa whilst walking through a minefield.
User avatar
Mars
Full Member
Posts: 694
Joined: Tue Nov 27, 2007 9:48 pm
Location: Bonn, Germany

Post by Mars »

Monsieur Plastique wrote: I think we tend to look at these old machines with rose-coloured glasses. ... The little Nintendo won the associated 6 game match 6 points to nil.
Hmm, for a Fritz 9 or 10 on a 60-70MHz hardware this is by no means a weak opponent despite of possibly having the image to be just a 'toy console program'. Taking your own results from some time ago...
http://www.hiarcs.net/forums/viewtopic. ... &start=150
... this toy plays around 2000 SSDF Elo. Compared to the ~1800 of SuperConnie it is simply one class better. From top of my head this should give roughly 75/25 results on average. No rosy glasses on SuperConnie but rather an underestimted Nintendo opponent, me thinks.

No wimp regards,
Martin
User avatar
Monsieur Plastique
Senior Member
Posts: 1014
Joined: Thu Jul 03, 2008 9:53 am
Location: On top of a hill in eastern Australia

Post by Monsieur Plastique »

Well the Super Constellation is around 1633 ELO on the current SSDF list. I'm just completing the 100 game test of the Nintendo over the next week and it slots somewhere in strength between a GK2100 and associated clones and a Novag Sapphire - somewhere around low 1900s. The rating will always be unreliable because of the small 100 game sample and lack of sufficiently strong and varied dedicated machine opponents.

You also can't judge anything by the 66 MHz hardware. There is a lot of overhead due to the Nintendo OS and graphic load, etc, plus the machine plays close to twice as fast in the middle game as a typical dedicated machine at 40 in 2. Then there is the issue of whether or not it even has any components in native machine code, since Fritz was never written in ARM assembler so I understand. So you might only be looking at something in the end not a whole lot faster than a very early model 68000 based dedicated machine.

But my point is that the Nintendo might be strong against the average amateur player but it is obviously not strong in absolute terms. The Super Constellation was a landmark release that garnered almost universal respect even up to high expert and low master level back in the I d80s. But when I look at it's play these days with a critical eye, all I can think of is "coffee house". It get's itself into situations time and time again that are far too deep for it to understand. It only went well in years gone by because it created positions where a human was apt to go astray. But against lower level but relatively precise play (such as the Nintendo), this style of play is very suspect.
Chess is like painting the Mona Lisa whilst walking through a minefield.
Post Reply