Komodo 8: Deep Blue revisited (part two)

by Albert Silver
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.

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...

The swift punishment of Garry Kasparov's excessively cautious play in game two, and the computer's ability to properly use its space advantage to attack also revealed another important facet: playing against the very best PC programs in 1997 would be woefully insufficient to anticipate Deep Blue's huge advance.

Game three

The very first move revealed Kasparov's intentions to the T, and somewhat ironically emulates much of Magnus Carlsen's openings approach today.  The then World Champion played 1.d3 with the overt purpose of taking the machine out of its book, and to 'just play chess'. A few moves later, after 1.d3 e5 2.Nf3 Nc6 3.c4 Nf6 the position threatened to transpose to a fairly normal English Opening, so Garry insisted with 4.a3, ensuring fairly original channels. Ironically, Deep Blue was probably still in book, since as pointed out by John Nunn in his notes, even a few moves later there were significant grandmaster references such as Chernin-Morozevich (1993) or Panno-Bronstein (1989).

Although after the match, Kasparov was criticized for having played much weaker than his standard, game three was not one of the cases if one is to rely on today's PC software. In fact, although Komodo 8 certainly diverges from several of the moves played by either side, nor does it criticize them as mistakes, and the evaluation differences are usually minimal at best. The largest difference is actually at move twelve.

Position one

Kasparov immediately chose to repatriate the knight to h3 and then
f2, and while understandable, the machines thinks it is imprecise

John Nunn:

It is very hard for White to achieve anything while his knight occupies a vulnerable position on g5

Deep Blue:

In its computer logs, Deep Blue can be seen to have expected 12.Be3, and only after did it anticipate Nh3

hash guess bc1e3,Guessing Be3
9(6) #[h6](0)################################# 0 T=38
Ph7h6 ng5h3 Pa7a5 nc3b5 Nc6d8 pd3d4 Pc7c6 nb5c3 Nd8e6 pd4d5
10(6)<ch> 'nh3'
---------------------------------------
--> ng5h3 <--
---------------------------------------

The evalution shows it thinks the position would be equal after this.

Komodo 8:

Komodo also thinks 12.Be3 was more precise, so as to answer 12...Nd4 with 13.Bxd4 forcing Black to take with the e-pawn, freeing the f4 square. However, the difference is that it thinks White has a small but healthy plus in this move order.

12.Be3 Rab8 13.Qc2 Nd4 14.Bxd4 exd4 15.Ne2 c5 16.b4 Bg6 17.Nf4 Qc6 18.Bh3 Rfd8 19.b5 Qe8 20.Nxg6 hxg6 21.f4 Nd7
+/= (0.31) Depth: 27 00:04:18 1075MN

The reason this is worth noting is because after the immediate 12.Nh3, Komodo revises the line and evalution.

12...Nd4 13.Be3 c5 14.Nf2 h6 15.Rc1 Rab8 16.b4 Rfe8 17.h4 a5 18.Bh3 Qd8 19.Bxd4 exd4
= (-0.03) Depth: 26 00:03:20 849MN

The Deep Blue logs

IBM has several spaces dedicated to Deep Blue, one of its great successes, some recounting the project and history, and also the replayable games with complete notes. The site does not merely include the complete computer moves and analysis, but also all the live commentary made at the time by Maurice Ashley (who was not a GM yet), IM Mike Valvo and GM Yasser Seirawan among others. Sadly these transcripts only cover the early opening phase since after that the live commentary files are blank.

Link to the Deep Blue logs

Position two

Nunn explains that the general consensus was that White stood
clearly better here. What about the computers?

John Nunn:

The general opinion expressed by other commentators has been that this position clearly favours White, but in my opinion White only has a slight edge. True, Deep Blue has played the opening in a way which appears very odd to human eyes, because it doesn't fit any of the normal patterns for playing with Black against the English Opening. Ideally, White would like to set his kingside pawns in motion, but it isn't easy to achieve this. The f-pawn is pinned, and White cannot unpin it by moving his queen, because then f3 would be hanging. Moreover, there is no attractive square to put the queen (ironically, the move a2-a3 cuts out a possible Qd2). Of course, White could play 15 g4 Bg6 16 f4, but then 16...exf4 17 Bxf4 Rae8 is fine for Black; he intends ...b5 and the f2-knight is badly placed. White wants to play f4 without g4, so as to meet ...exf4 by gxf4. The manoeuvre which Kasparov plays later in the game (involving Bh3) is not possible here because 15 Bh3 loses a pawn to 15...Bxf3. Therefore Kasparov gives up any immediate idea of kingside expansion, and opens a second front on the queenside.

Deep Blue:

While Kasparov played 15.b4 here, Deep Blue had actually expected 15.g4

hash guess pg3g4,Guessing g4
10(6) #[Bg6](-8)########################################## -8 T=47
Bh5g6 pb2b4 Ph6h5 pg4g5 Nf6h7 nf2h3 Nd4e6 nc3d5 Nh7g5p nd5e7B Qd7e7n nh3g5N
11(6) #[Bg6](-13)########################################## -13 T=203
Bh5g6 pb2b4 Pb7b6 pf3f4 Pe5f4p be3f4P Ra8d8 ph2h4 Nd4e6 bf4g3
12(6)<ch> 'b4'
---------------------------------------
--> pb2b4 <--
---------------------------------------

Komodo 8:

In its analysis, Komodo is very much on the fence, and switches between Kasparov's 15.b4 or 15.h4 quite liberally, judging the position as roughly equal.

15.h4 Bd8 16.Rc1 Bb6 17.Bxd4 cxd4 18.Bh3 Qe8 19.Nd5 Nxd5 20.cxd5 a5 21.Qe2 Bg6 22.Rc2 a4 23.Bg4 Bc5 24.Qd2 Qe7 25.h5 Bh7 26.Rfc1 b5 27.f4 f5 28.exf5 Bxf5 29.Bxf5 Rxf5 30.Qe2 Qf7
= (0.00) Depth: 26 00:03:36 871MN

On the other hand, it doesn't think much of Deep Blue's idea of 15.g4, a move that Nunn himself says is dubious in his notes in Mega 2015. After 15.g4?! Komodo think Black gets the upper hand with:

15...Bg6 16.b4 a6 17.f4 exf4 18.Bxf4 b5 19.Kh1 Rad8 20.cxb5 axb5 21.Be3 Rfe8 22.a4
= (-0.30) Depth: 24 00:02:26 551MN

Position three

Deep Blue played 16...Kh8, deeply criticized as being planless

John Nunn:

The computer cannot find a constructive plan. A human might have played 16...Rab8, preparing for the possible opening of the b-file.

Deep Blue:

What stands out when seeing Deep Blue's analysis, is that it is true it was quite unsure how to proceed. Consider that each and every ply it is changing its mind on what to play. 16...Rad8, then 16...Rfe8, then 16...a5, and then 16...Kh8

6(5)[Rad8](7)[Rfb8](10)[Rfe8](11) 11 T=2
Rf8e8 pg3g4 Bh5g6 ph2h4 Pc5b4p
7(5) #[Rfe8](5)##################################[a5](12)########### 12 T=5
Pa7a5 pg3g4 Pa5b4p pa3b4P Bh5g6 pf3f4 Pe5f4p
8(6) #[a5](4)##################################[Kh8](9)########### 9 T=16
Kg8h8 pg3g4 Bh5g6 pf3f4 Pe5f4p be3f4P Pa7a5
9(6) #[Kh8](-1)#####[TIMEOUT] -1 T=166
Kg8h8 pg3g4 Bh5g6 ph2h4 Pc5b4p pa3b4P Ph6h5
---------------------------------------
--> 16. .. Kh8 <-- 24/83:55
---------------------------------------

The evaluation that it was planless is not unfair, but it should be noted that Komodo did not think this disturbed the balance, and in spite of planning to send the king back to g8 later in its analysis, still though the position was roughly equal. That said, 16...Kh8 is never one of its top choices.

Komodo 8:

Komodo quite agrees with the grandmaster that the game plan is to prepare the queenside operations as can be seen below:

16...a6 17.h4 Rfb8 18.Nd5 Nxd5 19.cxd5 Qe8 20.Bxd4 cxd4 21.Qd2 a5 22.Rfc1 axb4 23.axb4 b5
= (-0.12) Depth: 23 00:04:38 1140MN

Game four

Garry stuck to his guns with his anti-computer opening strategy and came extremely close to beating Deep Blue in the game. While grandmaster does not mince words about the machine's play, highlighting planless play on many occasions that ultimately led to the disintegration of its position, what is perhaps more remarkable is the number of times today's software seem unable to do better. One thing is clear, whether an accumulation of imprecisions, or outright mistakes, Deep Blue was on the ropes late in the game, and the only question is whether the world champion missed a forced win.

A rapt audience watches the match live

Although there are several moments when Nunn, with the help of the programs of the time, felt sure the world champion must be winning, Komodo 8 seems to debunk all the attempts, finding moves backed up by deep resources that hold everything together, even if just by a thread.

Position four

Deep Blue seemed to be flailing about here and played 22.a3

John Nunn:

An extremely odd move which has no visible purpose, other than the general one of avoiding potential back-rank problems.

Deep Blue:

10(6) #[a3](39) 39 T=301
pa2a3 Re8e6 ph4h5 Rf8e8 qe2d3 Kc8b8 re1e3 Pg6h5p qd3f5
---------------------------------------
--> 22. a3 <-- 18/72:35
---------------------------------------

Komodo 8:

It is interesting to note that Komodo is unable to really improve upon its great predecessor here.

22.a3 Re6 23.Ka2 Rfe8 24.Qd3 Qf8 25.Nc3 Kb8 26.b4 Nxe5 27.Nxe5 Rxe5 28.Rxe5 Rxe5 29.Re3 Qd6 30.Ne2 Qe6
= (0.13) Depth: 27 00:03:11 993MN

Position five

A move later, Deep Blue plays 23.Nc3, and again it is met with disapproval

John Nunn:

Deep Blue appears to be floundering. This move undoes all the good work that the earlier Nd1 had achieved. 23.Qe3 was still the best move. If White were a human being, one would say that he had 'lost the thread'. It seems odd that computers can have the same problem.

Deep Blue:

Deep Blue's analysis suggests it is feeling quite pleased with its position, as can be seen by its +0.65 evaluation upon playing 23.Nc3

9(6) #[Nc3](74)############################# 74 T=15
nd1c3 Nd7c5 pb2b4 Nc5d7 pb4b5 Ne6d4 nf3d4N Qf4d4n pb5c6P
10(6) #[Nc3](65)############################# 65 T=102
nd1c3 Nd7c5 pb2b4 Nc5a6 pb4b5 Na6c5 pb5c6P Pb7c6p kb1a2 Rf8f7
11(6)[TIMEOUT] 65 T=159
nd1c3
---------------------------------------
--> 23. Nc3 <-- 17/69:48
---------------------------------------

Komodo 8:

Although the mainline varies a bit, Komodo plays the same line with much the same idea. Who was right? The grandmaster or the program? Based on principles there is no doubt the former Top 10 player is on steady ground, but one thing that has changed is how many times these principles can be broken by newfound emphasis on other factors, or quite simply pure calculation.

23.Nc3 Kb8 24.Qd1 Qf7 25.Qd3 Nf4 26.Qd2 Ne6 27.Ka1 Qf4 28.Qd3 Qf7 29.b4 Nf4 30.Qd2 Qe7 31.Qe3 Ne6 32.Ne2 Nb6 33.Nd2 Qxh4 34.Nd4 Qe7 35.Nxe6 Qxe6 36.f4 Rh8 37.Qf2 Ref8
= (0.21) Depth: 27 00:05:26 1629MN

Position six

Consistent with its previous move of 23.Nc3, IBM's chess machine
plays 24.b4? earning gasps of dismay and disgust. Naturally, a top
program today will avoid such a disgraceful move.

John Nunn:

A very ugly move. This was clearly the point of White's previous move: to be able to expel the knight from c5 without it landing on e4. However, the time White gains is as nothing compared to the serious weakening of his king positions. Kasparov must have been mentally rubbing his hands with glee at the sight of this reckless advance.

Deep Blue:

Despite the harsh words, it is worth noting that Deep Blue spent its longest time on this move, using over nine minutes of its time.

9(6) #[b4](60)#######################[Ka1](62)########### 62 T=35
kb1a1 Pd5d4 nc3e4 Nc5e4n qe2e4N Qf4e4q re1e4Q Re8d8 re4e2 Pd4d3 pc2d3P Rd8d3p
10(6) #[Ka1](50)#####[b4](71)############################# 71 T=157
pb2b4 Nc5a6 pb4b5 Na6c5 pb5c6P Pb7c6p kb1a2 Rf8h8 nc3d5P Pc6d5n pg4g5 Rh8h4p
11(6) #[b4](56)################################## 56 T=709
pb2b4 Nc5a6 pb4b5 Na6c5 pb5c6P Pb7c6p kb1a2 Re8e7 re1b1 Re7b7 qe2f1 Pg6g5 rb1b7R
---------------------------------------
--> 24. b4 <-- 16/60:45
---------------------------------------

Komodo 8:

Needless to say, Komodo also plays 24.b4 with a modest plus for White. This should not come as a big surprise since it also chose 22.Nc3, so at least they are consistent. The length of its selective line is an example of just how deeply its selective search can reach. It is no less than 25 moves deep.

24.b4 Nd7 25.Qd3 a5 26.Ne2 Qf7 27.Ned4 axb4 28.axb4 Nb6 29.Nxe6 Rxe6 30.Nd4 Qxf2 31.Rgg1 Ree8 32.Rgf1 Qxh4 33.Nxc6 Qxg4 34.Rxf8 Rxf8 35.e6 bxc6 36.Qa6+ Kc7 37.Qa7+ Kc8 38.Qxb6 Qc4 39.e7 Re8 40.Re6 Kd7 41.Rxg6 Rxe7 42.Kb2 Qb5 43.Qd4 Ke8 44.Qc3 Rb7 45.Qe5+ Re7 46.Qd6 Re4 47.Re6+ Rxe6 48.Qxe6+ Kf8 49.Qc8+ Ke7 50.Qc7+
= (0.13) Depth: 30 00:12:40 4127MN

Position seven

Deep Blue follows up its 24.b4 strike with 26.b5

John Nunn:

Having played b4 to keep the knights out of c5, it is of course inconsistent to let them back in again. Now White has nothing positive to show for the advance of his b-pawn, while his king position becomes weaker and weaker. White should have just remained passive, although Black has various ways of improving his position, for example by ...Kb8 and ...Rc8, aiming for ...c5.

Deep Blue:

10(6) #[b5](58)########################################### 58 T=56
pb4b5 Nd7c5 qd3e3 Re8d8 re1d1 Kc8b8 pb5c6P Pb7c6p kb1a1 Kb8a8
11(6) #[b5](57)###############<ch> 'qf7'
[cont] 57 T=372
pb4b5 Nd7c5 qd3e3 Qf7f4 re1d1 Qf4e3q pf2e3Q Kc8c7
---------------------------------------
--> 26. b5 <-- 14/58:15
---------------------------------------

Komodo 8:

This is where Komodo no longer agrees with Deep Blue's line, and though it approved of Nc3 and b4, it thinks 26.b5 is not the best way to go, and prefer 26.Na4. Note that its mainline is actually consistent with Nunn's recommendation to stay passive and wait. One can see the queens of both colors going back and forth to little effect.

26.Na4 Kb8 27.Qd2 Nb6 28.Nxb6 axb6 29.Kb2 b5 30.Qc3 Qf4 31.Re2 Re7 32.Qe3 Qh6 33.c3 Qh7 34.Qd2 Qh8 35.Qc2 Qh6 36.Re3 g5 37.h5 Rf4 38.a4 bxa4 39.Qxa4 g6 40.hxg6 Qxg6 41.Rh3
+/= (0.46) Depth: 27 00:03:45 1091MN

Position eight

This position did not actually occur, and is merely to illustrate how
previous attempts to prove a win are being refuted by the power
and precision of today's world class engines.

In 1998, Nunn was convinced that Kasparov had missed a win here with 51...d4! and gave 52.Rc7 and 52.e6 as failed atempts to hold. Naturally he was also using the best software he had at the time, but Komodo is quick to show how to defend the position with 52.Rh6!

Komodo 8

52.Rh6 Re2 53.Re6 Re1+ 54.Kc2 Rh1 55.Rd6 Rxh4 56.Kd3 Rh3+ 57.Ke4 Re3+ 58.Kf5 Rxa3 59.Rd8 Rf3+ 60.Kg5 Re3 61.Kf6 d3 62.e6 Rf3+ 63.Ke5 Re3+ 64.Kf6
= (0.00) Depth: 33 00:04:33 2153MN

One point for progress? For PC programs perhaps, but it is particularly interesting to see that that not only had Deep Blue expected 51...d4, but it had also seen the very same defense proposed with 52.Rh6! See for yourself:

Deep Blue:

hash guess Pd5d4,Guessing d4

12(6) #[Rh6](-33)############## -33 T=137
rh7h6 Kc5d5 pa3a4 Kd5e5p rh6c6P Pd4d3 kb1c1 Rh2h4p kc1d2 Rh4h3
13(7)<ch> 'kc4'
---------------------------------------
--> Kc5c4 <--
---------------------------------------

So it is safe to assume that even if Kasparov had played the attempt with 51...d4, Deep Blue would have defended the position all the same. Remarkable.

To be continued in part three

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!



Topics: deep blue, komodo 8

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

Mark Stockwell Mark Stockwell 12/31/2014 03:14
Funny, JDMN's comment to the effect he's surprised computers can 'lose the thread'. Surely the surprise would be if computer could ever rally 'grasp' a 'thread' at all. It's surely harder for a human, whose mind works in terms of narrative thread, to lose that thread than it is for a computer, where the process is more about concrete positions and evaluations of them.
Jarman Jarman 12/31/2014 02:41
@Bertman: if indeed it was an attempt at irony on their part, it fell a bit flat. Moreover, diagrams four and five are currently identical, and in diagrams six & seven White still has to play the move they are referring to (diagram four looks like the only correct one as a3 has just been played).
Bertman Bertman 12/31/2014 01:18
@Jarman

It was not an oversight. It was irony.
Jarman Jarman 12/31/2014 11:03
Position six: "Consistent with its previous move of 23.Nc3, IBM's chess machine plays 24.b4? earning gasps of dismay and disgust. Naturally, a top program today will avoid such a disgraceful move."

Then:

"Needless to say, Komodo also plays 24.b4 with a modest plus for White."
AMD_Puma AMD_Puma 12/31/2014 10:22
How about dusting off Deep Blue and engage today's engines in a match? It would be the ultimate hardware vs software clash (since Hydra already appears to be extinct), or can also be billed as the RISC vs CISC match, since Deep Blue is a RISC. Or maybe even play a match with Magnus.
1