Deep Blue's cheating move

by Albert Silver
2/19/2015 – It is a sign of just how impactful the famous Deep Blue match against Kasparov was in 1997, that 18 years later, books come out citing it still, and magazines such as Time cast their eye on it even today. Here we take a close look at the most controversial move from game two, that prompted Kasparov to accuse the Deep Blue team of cheating. The results may surprise you.

ChessBase 14 Download ChessBase 14 Download

Everyone uses ChessBase, from the World Champion to the amateur next door. Start your personal success story with ChessBase 14 and enjoy your chess even more!


Along with the ChessBase 14 program you can access the Live Database of 8 million games, and receive three months of free ChesssBase Account Premium membership and all of our online apps! Have a look today!

More...

Nearly two decades later, the match still fascinates

This week Time Magazine ran a story on the famous series of matches between IBM's supercomputer and Garry Kasparov. The subject was a few of the moves that stood out for a variety of reasons, such as a bug in game one of the 1997 match, and a move in game two that Kasparov found so unbelievable that he accused the Deep Blue team of cheating. The allegation was that a grandmaster, presumably a top rival, had been behind the move.

In fact, grandmaster Yasser Seirawan told Wired in 2001, “It was an incredibly refined move,
of defending while ahead to cut out any hint of countermoves, and it sent Garry into a tizzy.”

The topic of analyzing some key moves by Deep Blue with a top engine was recently covered at ChessBase News in a series of three articles:

Komodo 8: Deep Blue revisited (part one)

12/26/2014 – When Garry Kasparov faced IBM's super computer Deep Blue, there was not a player alive who did not secretly dream of having their own private Deep Blue to consult on demand. That day has already come and passed as today's engines such as Komodo 8 can outperform the famous computer even on a smartphone. Here we take a look at the famous match and key moments.

Komodo 8: Deep Blue revisited (part two)

12/31/2014 – The first two games of the titanic match between the world's strongest player versus the world strongest chess computer showed significant changes. While the computer played moves that baffled humans, the number of moves a grandmaster would outright scoff at were dwindling and could only be judged as eccentric. Today's best engines are actually no different, just stronger.

Komodo 8: Deep Blue revisited (part three)

1/9/2015 – Nearly eighteen years later, the matches between Garry Kasparov and Deep Blue continue to stir the imagination and fascination of people around the world. It has inspired books, documentaries, and theatrical productions. Looking over the games with a top engine such as Komodo 8, and comparing with the computer logs, sheds new light, and even reveals a missed win by Kasparov.

However, the purpose had been to highlight refinements and revelations uncovered with modern hardware and chess engines, and since there had been no disagreement with the so-called cheating move, it was not mentioned. So the engines agree with Deep Blue's choice, is that all? Not quite. Let's take a look at the controversial moment first.

This was the position on the board after 35 moves. It is White to play.

Before trying to simply guess what was played, if you don't already know, it is important to remember that the year is 1997, and engines, and even supercomputers, played quite a bit differently than today's uber-engines. They were notoriously materialistic, and a tactic used by a player such as Kasparov, who would get deep into the head of his opponent, was to bait the machine with a 'poisoned pawn', allowing him to direct the game in a direction of his choosing.

The pawn on b5 is clearly weak, and it should not be hard to bring it crashing down. The problem is that the most direct method is far less than optimal. As noted by grandmaster John Nunn, who annotated the game then and whose notes can be found among the tens of thousands of commented games in MegaDatabase:


[Event "New York Man-Machine"] [Site "New York"] [Date "1997.05.04"] [Round "2"] [White "Deep Blue"] [Black "Kasparov, Garry"] [Result "1-0"] [ECO "C93"] [BlackElo "2785"] [Annotator "John Nunn"] [SetUp "1"] [FEN "r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - 0 36"] [PlyCount "6"] [EventDate "1997.05.??"] [EventType "match"] [EventRounds "6"] [EventCountry "USA"] [Source "ChessBase"] [SourceDate "1997.11.17"] {It would be wrong to play} 36. Qb6 Rd8 (36... Bc7 $2 37. Qe6+ Kh7 38. Qxe8 {wins}) 37. axb5 Rab8 38. Qxa6 e4 {when, at the cost of two pawns, Black has developed considerable counterplay based on ...Qe5 and possibly ...e3. However, this is again a remarkable move by Deep Blue; most computers (including Fritz and Hiarcs) go for the material without hesitation. Did Deep Blue reject this line based on very deep analysis, is there some subtle programming involved?} 1-0

Needless to say, Deep Blue did not follow the trail of bread crumbs Kasparov was carefully sprinkling in its path. It played the very strong 36.axb5! axb5 and 37.Be4! to which Nunn comments, " A cruel move. Black's only chance of counterplay is to activate his bishop by ...e4, but Deep Blue cuts the rope which might have saved the drowning Kasparov."

In fact, if you consult any of the top engines of today, whether it be Houdini 4, Stockfish 6, or Komodo 8, they all choose Deep Blue's move 36.axb5. For example:

Komodo 8: 36.axb5 axb5 37.Be4 Qd8 38.Kh2 Rcb8 39.Ra6 Kf8 40.R6a5 Kg8 41.R1a2 Rxa5 42.Rxa5 Bc7 43.Ra1 Bb6
+/- (0.84) Depth: 29 00:02:15 686MN, tb=57

The problem is that it does not condemn 36.Qb6 either, just the blind pawngrabbing that follows. That very same Be4 can be combined with it to great effect. So what does this all mean? It means that while 36.axb5 was indeed a very strong move, it was not the only one, and the move order was not actually essential so long as the ideas are maintained, and, as pointed out by Nunn, White avoids allowing Black the time to strike back with ...e4 himself.

Analysis with the top engine lines:

[Event "New York Man-Machine"] [Site "New York"] [Date "1997.05.04"] [Round "2"] [White "Comp Deep Blue"] [Black "Kasparov, Garry"] [Result "1-0"] [ECO "C93"] [BlackElo "2785"] [Annotator "John Nunn"] [SetUp "1"] [FEN "r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - 0 36"] [PlyCount "7"] [EventDate "1997.05.??"] [EventType "match"] [EventRounds "6"] [EventCountry "USA"] [Source "ChessBase"] [SourceDate "1997.11.17"] 36. axb5 $1 ({It would be wrong to play} 36. Qb6 Rd8 (36... Bc7 $2 37. Qe6+ Kh7 38. Qxe8 {wins gewinnt}) 37. axb5 ({Komodo 8 64-bit points out that after} 37. Be4 $1 {White will continue to maintain a strong advantage, even after Black's best chance with} a5 $1 ({The tempting} 37... Qh5 $2 {to try to get counterplay with the queen is insufficient and ultimately worse.} { Komodo 8 64-bit:} 38. axb5 Rab8 39. Qxa6 Qh4 40. Re2 Qg3 41. Qa3 Rxb5 42. Qc1 Bc7 43. Kh1 Rbb8 44. Qe1 Qxe1+ 45. Rexe1 Kf8 46. g3 Bb6 47. Ra6 Rd6 48. Rd1 Ke7 49. Kg2 Rd7 50. Rda1 Kd6 51. R1a5 Ke7 52. Ra2 Kd6 53. Ra1 Ke7 54. Rd1 Kd6 55. Kf3 Ke7 {1.80/31}) 38. axb5 axb4 39. Rxa8 Rxa8 40. Ra6 Rd8 41. cxb4 c3 42. Qc6 Qh5 43. Qxc3 Qe2 44. Qd3 Qe1+ 45. Kh2 Kh7 46. Qc4 h5 47. Qd3 h4 48. Qf3 Qxb4 49. Qg4 Bc5 50. d6 Qd4 51. Qxh4+ Kg8 52. Qe1 Rxd6 53. Ra8+ Rd8 54. Rxd8+ Qxd8 55. h4 Qd4 56. Qb1 Qc4 57. g3 Kf7 58. Kh3 Ke7 59. Bc6 Bd4 60. Qe4 Qf1+ 61. Kg4 Qd1+ 62. Qf3 {0.97/31}) 37... Rab8 38. Qxa6 $2 ({Komodo 8 64-bit quite agrees that Qxa6 opens the way for Garry, but sees} 38. Qc6 $1 {as quite winning for White.} axb5 39. Qxe8+ Rxe8 40. Ra6 Red8 41. Be4 Kf8 42. Kf2 Ke7 43. Kg3 Rb7 44. Kg4 Kf8 45. R1a5 Bb8 46. Re6 Kf7 47. Kf3 Kg8 48. Ra1 Kh7 49. Rc6 h5 50. h4 Kh6 51. Raa6 Ba7 52. Re6 Rbd7 53. g3 Bg1 54. Rxe5 {1.82/29}) 38... e4 {when, at the cost of two pawns, Black has developed considerable counterplay based on ...Qe5 and possibly ...e3. However, this is again a remarkable move by Deep Blue; most computers (including Fritz and Hiarcs) go for the material without hesitation. Did Deep Blue reject this line based on very deep analysis, is there some subtle programming involved?}) 36... axb5 37. Be4 {A cruel move. Black's only chance of counterplay is to activate his bishop by ...e4, but Deep Blue cuts the rope which might have saved the drowning Kasparov.} Rxa2 38. Qxa2 {This position is just lost. Black's bishop must stay on d6 to block the d5-pawn, but Black cannot both maintain this bishop and defend the weak pawn on b5. Black's total lack of counterplay is the deciding factor: White has plenty of time to slowly infiltrate with his queen and rook.} Qd7 39. Qa7 {and Deep Blue had a winning advantage.} 1-0

It is clear that while Deep Blue was far ahead of its time, by the time the controversial position arrived, it had more than one path to victory.

If you are seeking the best today, in the most advanced interface available, including a Premium subscription to Playchess, then Komodo 8 is the way to go.

Komodo Chess 8 includes:

  • The Komodo 8 engine, which can support up to 64 processor cores and 16 GB of hash memory
  • The new Deep Fritz 64-bit program interface (+ 32 bit program interface)
  • Online access to the world’s largest analysis database “Let’s Check”, with over 200 million extensively analyzed positions (free access to “Let’s Check” until December 31, 2016)
  • Access to ChessBase engine cloud
  • Premium membership to Playchess.com for six months
  • Database with over 1.5 million games

System requirements

Minimum: Pentium III 1 GHz, 2 GB RAM, Windows Vista, XP (Service Pack 3), 7/8, DirectX9, 256 MB graphics card, DVD-ROM drive, Windows Media Player 9 and Internet access for program activation, access to Playchess.com, Let’s Check and program updates.

Recommended: PC Intel i7 (Quadcore), 4 GB RAM, Windows 8.1, DirectX10, 512 MB graphics card, 100% DirectX10-compatible sound card, Windows Media Player 11, DVD-ROM drive and Internet access for program activation, access to Playchess.com, Let's Check and program updates.

Price: €79.90 (€67.14 without VAT for customers outside the EU; $86.62 without VAT). Languages: English, German. ISBN: 978-3-86681-442-4; EAN: 9783866814424. Delivery: Download, Post

Order Komodo Chess 8 in the ChessBase shop now!



Born in the US, he grew up in Paris, France, where he completed his Baccalaureat, and after college moved to Rio de Janeiro, Brazil. He had a peak rating of 2240 FIDE, and was a key designer of Chess Assistant 6. In 2010 he joined the ChessBase family as an editor and writer at ChessBase News. He is also a passionate photographer with work appearing in numerous publications.
Discussion and Feedback Join the public discussion or submit your feedback to the editors


Discuss

Rules for reader comments

 
 

Not registered yet? Register

yournumberisone yournumberisone 3/3/2015 05:40
Joel Benjamin 2/23/2015 07:16 "Some chessplayers see "suspicious" behavior from IBM that is really indifference. There was no rematch because IBM had nothing to gain from another match. The machine was dismantled because they were finished with chess and wanted to use the expensive machine for something else. They didn't think it was a big deal to publish the logs (though they eventually put them on their website, and gave me copies in 1997) because nobody cared other than a bunch of conspiracy theorists. Why should they have to prove a negative?"

For one thing, Kasparov states in the documentary "Game Over: Kasparov and the Machine" that he repeated asked for the logs, and IBM repeatedly agreed/refused to. So it was certainly a ‘big deal’ to Kasparov. Some may call IBM's behavior "psychological warfare". I call that underhanded and manipulative. Kasparov states he was able to analyze Deep Blues previous games before the 1996 match, but not for the 1997 match. Why? And what of IBM’s strong-armed tactics towards Jeff Kisselhof?

IBM had plenty to gain from a rematch. Let's start with respect. Giving Kasparov a rematch would not only be respectful for who most feel is the greatest chess player in history, but it would give IBM an opportunity to be transparent in how it conducted itself, because the unanswered questions surrounding the 1997 will simply not go away without IBM being more open and forward. This is something that should have been done with all matches of this caliber. Don't lock the machine away in some back room w/o even so much as a camera showing the room it’s in. As for why IBM should have to “prove a negative” as you put it, why would anyone – individual or corporation – not want to respond to negative comments or accusations in regard to their character? We’re not talking about criticism in regards to some product sold by IBM but rather one that could be viewed by some as borderline stock manipulation.

How exactly was the Deep Blue computer “used for something else” when it was dismantled immediately after the match (with half of it donated to a museum)?

It wasn't just a scientific exercise as IBM claimed it was, otherwise they would have been transparent from the beginning. IBM would like everyone to believe they simply improved on their 1996 Deep Blue computer and hoped it was good enough. But in fact what they did in a year's time was formulate a PLAN to beat Kasparov. Kasparov wasn't just playing a computer, he was playing a computer plus a team of grandmasters. Whether or not this 'dream team' of chess players had any influence on Deep Blue DURING the game (particularly game #2), the public may never know. Perhaps years from now, someone involved with IBM's team may come forward and offer a story that conflicts with the accepted history. Did IBM use similar 'tactics' in its matches with Watson and the Jeopardy champions? Again, who knows. All I know is, as long as one side remains guarded and hidden behind a 'curtain', and as long as one side can achieve a financial advantage from winning, the possibility of an unfair advantage will always remain.
Deep_Port Deep_Port 2/24/2015 02:56
"They didn't think it was a big deal to publish the logs"

What was their reasoning for thinking that? When you have the WCC accuse you of cheating, then you should everything to refute or address such an accusation. And they clearly failed to do that. Which explains why the questions still linger even today.

What they should have done was show convincingly that the machine did not receive human assistance. Instead, what you have are a bunch of unconvincing arguments or articles appearing that suggest that Deep Blue won because of a computer bug, or that GK was simply producing sour grapes, neither of which does anything to refute the cheating allegations.

It's also worth mentioning that the "unfairness" in the Deep Blue match (or what made the Deep Blue match unlike other top-level play Chess), is that Deep Blue had a deep understanding/knowledge of Kasparov's moves/games, but Kasparov had no information about Deep Blue's moves or games. And to make matters worse, any information he had built up in the previous games were erased once they tweaked the code. If IBM wanted to do this fairly they'd have given Kasparov at least as much data on Deep Blue as they had on Kasparov.
Deep_Port Deep_Port 2/24/2015 02:53
@Benjamin

There's no need to invoke conspiracy to explain IBM's behavior. Consider this: The day that Deep Blue beat Kasparov, IBM's stock rose by 15%. <-- That's not a conspiracy.

In addition, IBM's behavior immediately after the match was inconsistent with their stated agenda, which, if you take their word for it was - "To promote science".

"Behind the contest, however, was important computer science, pushing forward the ability of computers to handle the kinds of complex calculations needed to help discover new medical drugs; do the broad financial modeling needed to identify trends and do risk analysis; handle large database searches; and perform massive calculations needed in many fields of science." (http://www-03.ibm.com/ibm/history/ibm100/us/en/icons/deepblue/)

"Massively parallel, special-purpose computing like that found in Deep Blue could certainly be of great use to people if applied to finance, medicine, education, etc. Imagine an evaluative capability like Deep Blue's that could help an investor manage a portfolio, a huge retailer manage inventory, or a government deploy resources. These types of things justify the spending behind Deep Blue. Chess and Kasparov are merely ways of benchmarking progress." (https://www.research.ibm.com/deepblue/meet/html/d.3.3a.html)

However, when Kasparov was defeated, unconvincingly, there was no follow-through on the science. Instead, IBM immediately acted to remove all traces of evidence, while brushing aside GK's allegations and demands for a rematch, and implying that the science was completed.

As for indifference, there's no such thing in this case. Because once you have an agenda, where does indifference fit into the picture? You think they were indifferent about their public image or share price? Not possible.

"There was no rematch because IBM had nothing to gain from another match."

How sure are you about that? And what was it that they were trying to gain from both matches anyway? And what did they actually gain in reality? And what of the "sciecne"? All indications are that it was more about about beating Kasparov and improving their share price than about scientific advancement.

Also, it was not just about having another rematch. It was also about checking to see if the controversial moves could be found by the machine, thus protecting IBM's reputation.. But this never happened.

"The machine was dismantled because they were finished with chess and wanted to use the expensive machine for something else."

That's the official reason. But it does not add up. Someone could just as well argue that the machine was dismantled because they wanted to exclude the possibility of having to address the controversy that they brought upon themselves, and which they failed to address.

The other point is that the machine only cost around the 4-5 million... peanuts when you compare to IBM's net income for 1997, which was reported at over 7.6 billion USD.
Joel Benjamin Joel Benjamin 2/23/2015 07:16
Some chessplayers see "suspicious" behavior from IBM that is really indifference. There was no rematch because IBM had nothing to gain from another match. The machine was dismantled because they were finished with chess and wanted to use the expensive machine for something else. They didn't think it was a big deal to publish the logs (though they eventually put them on their website, and gave me copies in 1997) because nobody cared other than a bunch of conspiracy theorists. Why should they have to prove a negative?
Deep_Port Deep_Port 2/21/2015 11:28
@MvanVeen
"In essence there is no difference between chess programming then and chess programming now"

There are HUGE differences between the quality of the algorithms today compared to the 90's. More sophisticated techniques have been discovered and implemented, resulting in more refined evaluation function. In fact, most of the improvements in computer chess performance can be directly attributed to better software, not hardware, as some believe. Even Fritz 5.32 had better eval than Deep Blue. You give Fritz 5.32 the capacity to analyze 200million positions per second and it would easily have had Deep Blue for lunch.

"Though programmers can program more easily because of languages [such] as C++, and though chess knowledge is easily available these days, in essence the programmer must do the same as back in the 90s."

There are big differences today. For instance, a programmer is not required to do the same as what programmers did back in the 90's, because back in the 90's, techniques were far less refined than they are today. Nowadays a programmer does not have to start from scratch to make a good chess program (i.e., 3000+ elo).

Here is an excerpt from an interview with Robert Houdart, author of Houdini: (excerpt from AlgorithmicProgress.pdf)

"I started with this idea to build the best chess engine that I could—and I was helped a lot by the open culture that has come with the Internet. You know, two decades ago you had to invent every part of a chess engine from zero (and I’ve done my fair share of that), but today we’re in a situation where techniques, ideas and examples are readily available on the Internet. You can call it a coming of age of the computer chess scene—as an engine author you’re no longer obliged to sit in your corner reinventing the wheel. The computer chess Wikipedia, some strong open source engines, and discussions on Internet forums about chess programming techniques and ideas make the design and development of a strong engine a lot easier than, say, twenty years ago. . . "

See also...

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=26026
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=4059
http://ulysse.io/ComputerChess.pdf
http://intelligence.org/files/AlgorithmicProgress.pdf
Dinesh Panchamia Dinesh Panchamia 2/21/2015 11:04
The stock loss of IBM could have been recovered by only auctioning the historical machine today , if the Deep Blue machine was not dismantled!!.
but the question is why was it dismantled ?, in spite of raising doubts by GK , seems doubtful by all means ?
Deep_Port Deep_Port 2/21/2015 06:50
@JBejamin

"but they do show the computer had every move played as it's principal variation just before it played the move"

That doesn't prove anything. They could easily have added it before printing if they thought the stakes warranted it (share prices, public perception, etc). The only real way to prove conclusively that Deep Blue could find those moves on its own (using the exact same settings play against GK), was to have a legit demonstration. But they eliminated that possibility from occurring by immediately dismantling Deep Blue after that match.. presumably because the "science and research" was suddenly all completed at once. lol.

And why dismantle such an amazing machine? And why so hastily? IBM could easily have afforded to keep Deep Blue in storage, hold matches with it in future, and make it available for research, assuming research and science was what they were really interested in.
M van Veen M van Veen 2/21/2015 05:26
In essence there is no difference between chess programming then and chess programming now. Though programmers can program more easily because of languages as C++, and though chess knowledge is easily available these days, in essence the programmer must do the same as back in the 90s.
And that is that the programmer must transform knowledge into code. The computer then mimics the idea behind it. Simply put: if I want to make the computer behave like Mr. Bean, I can study Mr. Bean and put all the ideas into code. And if all goes well, I can make a computer program that acts like Mr. Bean.
This has to be done in recent programs, and it had to be done back in 1996. And that a move seems human can well be. It has to be a human in the first place who translates ideas and knowledge into code. Was Karpov involved in the development of Deep Blue? Did Bobby Fischer secretly give tips to the IBM team? Or Korchnoi? or Spassky? Who knows! It seems a fact to me that Deep Blue had some subroutines that reflected sophisticated knowledge, which would then emerge under certain conditions in the game. The good old "if-then" construction or the equivalent thereof. So, Kasparov's suspicions regarding a possible human rival in the IBM camp were not so strange... after all, humans are behind every computer program...
Frits Fritschy Frits Fritschy 2/20/2015 11:32
GM Benjamin
"Although they were slow in doing so, IBM released the logs years ago. None of you would understand them"
I believe you, but that's not a very strong defense. You could show the essential parts of the logs (if necessary, by a link) and explain them. When it's in the open, anyone could respond to them, including people who do understand them.
Frits Fritschy Frits Fritschy 2/20/2015 11:14
"Even when they didn't cheat (in my opinion, quite likely they didn't)", that's what I wrote myself. Why wasn't it likely, according to me? Not because anyone within IBM would consider that morally wrong. It's just about cost effectiveness and weighing risks against profits. They were already there, playing the world champion. Publicity guaranteed anyway. If they wouldn't succeed this year, they would succeed a following year; Kasparov would be more than happy to do his circus act again. Maybe IBM would even have preferred that...
Why should they run the risk of ruining that? Anybody involved could earn a little side money by revealing a scam. Moreover, chess players are an independent lot, and prone to put themselves in the spotlights - not very dependable when needed to cover up indecent behaviour.
Kasparov of course shouldn't have given in to easy money in the first place.
Joel Benjamin Joel Benjamin 2/20/2015 11:06
Also the comment about "IBM changing piece values on the fly while the game is going on" is complete nonsense. We made some changes between games, which was allowed by the rules (just as people are allowed to do). I don't think we actually changed piece values at all. That was a speculation from Kasparov, who didn't really understand the variables that went into Deep Blue's decision making.
Joel Benjamin Joel Benjamin 2/20/2015 10:59
Although they were slow in doing so, IBM released the logs years ago. None of you would understand them (there's only so much information there that's even useful to the programmers), but they do show the computer had every move played as it's principal variation just before it played the move. More importantly, KASPAROV saw the logs immediately after the match. He never claimed he didn't, but coyly remarked that "IBM must show the logs to the world."
It's a total non-issue, and a truly embarrassing episode for Kasparov. I'm still waiting for an apology.
Deep_Port Deep_Port 2/20/2015 10:15
@PaulPena

I think you hit the nail on the head with that comment.

Personally, I find it hard to understand why IBM did not simply give a public demo with DB to show that the machine was actually capable of finding the controversial moves. In addition, they didn't even give the Kasparov the courtesy of a rematch. Instead, they had the machine disassembled before anyone could learn anything about its play or its controversial moves. And that's interesting.

@KevinC & Karbuncle

"Modern computers show that Deep Blue's team was just ahead of its time programming-wise."
"even engines 10 years later picked the same moves, and the accusation is completely baseless."

While modern computers with far more sophisticated algorithms than DB find the moves easily, whether or not DB was capable of finding those controversial or "sophisticated" moves in 1997, using obsolete algorithms, is a whole other question entirely. Fact is, it was never validated because IBM had the machine dismantled and never gave a demonstration to prove that DB could in fact find and play the moves it supposedly found on its own. Hence there's still a big question mark that remains unanswered.

https://www.youtube.com/watch?v=EtMdMmrfipY (Kasparov vs Deep Blue doco)
valg valg 2/20/2015 10:02
I think IBM cheated as Kasparov implies ... and if they didn't they sure acted guilty
valg valg 2/20/2015 09:59
I very much doubt this comment will be posted in its entirety let alone posted at all, but I feel that this article is severely misleading, as it ommits crucial details.

First of all IBM themselves revealed that they can change DEEP BLUE piece values on the fly while the game is going on. During that time IBM called this a feature. We now call it cheating.

It is strongly suspected that a human grandmaster on the IBM team (like Joel Benjamin), substituted a couple of human consolidating moves in a messy position where Deep Blue had shown in similar positions (in previous games) it would rather just grab more material. Only the logs would tell for sure but IBM felt that their paper shredder had an urgent need for Deep Blue's logs immediately after the conclusion of the match.

Please keep in mind that Kasparov had already beaten Deep Blue in a previous match and their reputation and stock price was riding on their firmly stated convictions that the next match would be different. We’re talking a lot of money folks.

Here is the circumstantial evidence:
When Kasparov became suspicious he asked to see the logs. They told him to please finish the match and they would then provide the logs. After Kasparov finished the match they immediately disassembled the machine and destroyed the logs. Now remember that they said that this was all for AI research. Does that sound like research to you or just trying to get a bump in a stock price?
Frits Fritschy Frits Fritschy 2/20/2015 09:53
Paulpena.
Don't push things to far. "the combination of neural net technology and other AI artifices has given the appearance of 'human like' moves to modern software." Phew, you seem to know a lot more about hidden things in chess software than me. Maybe the NSA even is reading this forum!
There is a very good reason for IBM to not hand over the log files (ah, that was the word I was looking for). Why should they care about what that small chess community thinks? For the general audience, they have beaten chess itself. Even when they didn't cheat (in my opinion, quite likely they didn't), there is no reason to feed a discussion that won't do them any good.
PaulPena PaulPena 2/20/2015 08:13
Sorry Karbuncle's argument is completely invalid. 10 YEARS LATER you're going to ask what modern software would do? That's like asking what modern man would do 10,000 years after the caveman. Yes computer science has developed by leaps and bounds and the combination of neural net technology and other AI artifices has given the appearance of "human like" moves to modern software. But it does not explain how a computer driven by only brute force calculation would come up with the same move a human would. If IBM really wasn't cheating why not just release the logs like they said they would? All that does is serve to raise suspicions and cast a negative light on the company. Why would any company with a great PR department (as IBM does have) allow something like that...unless the alternative was worse.
mike warner mike warner 2/20/2015 06:25
IBM could have released deep blue's calculation logs for that game (as Kasparov requested) and put all speculation to rest. For some reason they didn't and here we are.
KevinC KevinC 2/20/2015 06:16
@Bertman, thank you. They did, indeed, update the article. I did go and look it up, which is why I thn posted my second comment.
Frits Fritschy Frits Fritschy 2/20/2015 05:16
Bertman,
I don't think much of Kasparov's allegation either; I was just saying that this article doesn't constitute any proof of the opposite. By the way, it is still possible to improve on computer evaluations as a human player. For instance, so called 'fortress draws' still can be a problem for engines: humans see it at once, engines sometimes don't have a clue. So yes, a computer can aid a player to beat a much stronger player, and the other way around: a strong player can aid an engine to beat a legend. Certainly back in 1997. GM Benjamin didn't have to handpick a move, as an operator he could see what the engine was thinking. When I use an engine to analyse something, I do the same: I use it for blunder check, but I also sometimes push it in a certain direction. Kasparov played a match against Topalov once where both could use a computer! They certainly didn't let the computer take all the decisions.
As far as I remember, Kasparov's main problem with the Deep Blue team was that they didn't want to hand over the records on what and how the engine evaluated (don't know the official English term for it). Does anyone know whether this issue has ever been resolved?
Bertman Bertman 2/20/2015 03:34
@Frits

If you stop to think about it for even a second, the allegations are obviously ridiculous. Cheating with a grandmaster? Kasparov was the best player in the world. Why would a grandmaster handpicking a move, know which move, and somehow be able to defeat Garry Kasparov in his prime? Frankly, the cheating allegations are absurd, but the fact he highlights that move as revolutionary for a chess playing machine is certainly worth looking at.

He certainly knew better than anyone what engines of his day could do, and clearly this move by Deep Blue was something special to his eyes.
Karbuncle Karbuncle 2/20/2015 03:04
And of course it was just pure sour grapes speculation. There wasn't one SHRED of proof any cheating took place. Add to that how even engines 10 years later picked the same moves, and the accusation is completely baseless.
Bertman Bertman 2/20/2015 02:39
@KevinC

The allegation was that a grandmaster, presumably a top rival, had been behind the move. It has been added near the beginning of the article.
KevinC KevinC 2/20/2015 02:29
I have to wonder if Kasparov would have a different opinion now, away from the heat of the battle, and knowing just how good computer programs have gotten since then.
KevinC KevinC 2/20/2015 02:26
I don't remember the back and forth on this, so it would be nice if they told us what Kasparov's specific cheating alegation was. Did the Deep Blue team force a move in, or otherwise alter the computer during the game? If not, it was not cheating.

I suspect that Kasparov just could not believe that a computer could make such a good set of moves, and ignore the prooffered material. Absent that proof that Deep Blue altered the machine, Kasparov, who I consider the greatest ever, comes off as whiney. Modern computers show that Deep Blue's team was just ahead of its time programming-wise.
Frits Fritschy Frits Fritschy 2/20/2015 12:34
The strange thing is that for probably most reasonably strong human players, the move Be4 is completely natural, as counterplay with e4 and Qe5 looks threatening. For engines the concept 'threatening' has far less value, as far as I know. Engines try to win as efficiently as possible, humans try to win as safely as possible. In fact, the first move I looked at was the immediate 36 Be4, but in a game, for the same safety-first reasons (the black counterplay with a5) I might well have opted for 36 axb5.
In the variation 36 Qb6 Rd8 37 axb5 Rab8 38 Qc6, you will have a hard time convincing me this position is winning for white (although he clearly is better); especially the black moves after 46 Kf3 are puzzling. I assume that GM Nunn did a bit more than just let the engine run for a while, but for me in a game, the unclarity might be enough to choose for 36 axb5.
As this 'forced win' by Komodo (in the variation 36 Qb6) was far beyond the scope of Deep Blue, this article of course doesn't constitute any proof that the Deep Blue team didn't cheat.

By the way, Karbuncle, the link you're providing isn't working.
Karbuncle Karbuncle 2/20/2015 11:05
I mentioned this same aspect of modern software picking the same moves back in 2007 with Fritz 10. It shows the move was in fact "siliconian" in nature and not GM Benjamin cheating. Here's a link to my discussion with GM Benjamin on this same subject:

http://www.uschess.org/content/view/6769/341/
1