chessgames.com
Members · Prefs · Laboratory · Collections · Openings · Endgames · Sacrifices · History · Search Kibitzing · Kibitzer's Café · Chessforums · Tournament Index · Players · Kibitzing
Fritz (Computer) vs Eduard Nemeth
Stufe=Blitz:30'. (2000) (rapid), ?
Spanish Game: Morphy Defense (C70)  ·  0-1

ANALYSIS [x]

FEN COPIED

Get this game explained with Decode Chess
explore this opening
find similar games 5 more Fritz/E Nemeth games
sac: 17...Qxg5 PGN: download | view | print Help: general | java-troubleshooting

TIP: You can learn a lot about this site (and chess in general) by reading the Chessgames Help Page. If you need help with premium features, please see the Premium Membership Help Page.

PGN Viewer:  What is this?
For help with this chess viewer, please see the Olga Chess Viewer Quickstart Guide.
PREMIUM MEMBERS CAN REQUEST COMPUTER ANALYSIS [more info]

Kibitzer's Corner
May-16-05  BadTemper: wtf.. totally weird
May-17-05
Premium Chessgames Member
  WannaBe: No, no, this is a typical Nemeth vs. computer game. He has an understanding of how computers are programmed. Therefore he exploits it. His game are focused on the H column attack. He has a Nemeth Gambit named after him.

In my opinion, he is quite incredible, to be able exploit the computer's programming and be able to have such an incredible record versus computer opponents, is absolutely fantastic.

Nov-11-06  2021: I think Nemeth is doing something to the computer to make it do bad.
Nov-11-06  macphearsome: yeah, beating it.
Nov-11-06  argishti: One thing Eduard Nemeth focuses on is the h and a files when fritz castles. He makes sure they are under attack because this was fritz's main weakness. You can check this same principle in his games against the same machine. Very nice technique to win over such a good computer.
Nov-12-06  ashin: woah this is a fancy game...and this is the first time im hearin about Eduard Nemeth...according to chessgames database he has all wins and one draw
Apr-18-10  D20: 2000? fritz wasn't exactly unbeatable back then.. it has fewer holes today. And if he just uploads his wins/ties... Get an FIDE rating Eduard. I bet you can't break 1600
Apr-19-10  Caissanist: Nemeth is, or was, a 2100 player. For a while in the early 2000s he was publishing a lot of wins (over 100 of them) against the most popular commercial chess programs, but he he stopped doing that about six years ago. Most likely all the weaknesses he exploited have been fixed, as D20 says.
May-01-14  GumboGambit: Well, from what I heard, Fritz went on the Fritz after Nemeth repeatedly slurred "I wanna kiss you".
May-02-14  mrbasso: Nothing was fixed. The programs simply search much deeper nowadays (much better hardware and improved search) and therefore this crazy approach does not work anymore.

This game was only a result of 'trial and error', not a real game.

May-02-14  DcGentle: If you want to see a real game, just look here: Karjakin vs Deep Junior, 2004
May-02-14
Premium Chessgames Member
  AylerKupp: <DcGentle> Nice cherry picking. But the Karjakin vs Deep Junior, 2004 game was the only one that the humans won against the computers at the 2004 Man vs. Machine World Team Championship. The machines scored 8.5 vs. 3.5 for the men, +6-1=5. And I presume that the other games were also "real". :-)

A similar event was held the following year, renamed "People vs. Computers", presumably in the interest of political correctness. Here the "people" did a little better; the computers scored 8.0 vs. 4.0, +5-1=6, with Fritz 9 the only loser (against Ponomariov). Presumably these games were "real" also. And the "people" had to scramble to do that well by drawing all 3 games against the "computers" in the final round. Or maybe the "computers" just got tired. :-)

Here is a game from this later event in which Junior outplays Khalifman <positionally> (shocking!) in the White side of the Ruy Lopez with an early 4.d3 until its positional advantage is sufficient to first win a pawn, then gives it up (yup!) to retain its positional advantage, and eventually translate that advantage into a winning B+N+2p vs. B+3P endgame. After the game Khalifman apparently compared Junior's play to Karpov at his very best (http://en.chessbase.com/post/wham-2...).

Presumably Junior learned a thing or two between 2004 and 2005. Or maybe it was a change in location to California. When one of its developer, Amir Ban, was asked why the move from the U.K. to California, his reply was "Because the weather is so good." (http://en.chessbase.com/post/3-0-sh...). I'll second that. :-)

May-02-14
Premium Chessgames Member
  AylerKupp: <<mrbasso> Nothing was fixed. The programs simply search much deeper nowadays.>

You couldn't be more wrong. There have been as many innovations in chess software since 2000 as there have been in hardware, probably more. For example, the reason that the programs search much deeper nowadays are improvements in heuristics for pruning the search tree, since without pruning the number of positions to be searched for each ply would increase far faster than the ability of newer and faster processors to evaluate them.

Example: High-end PCs available in 2000 were typically based on the AMD K6-2 with a clock rate of about 300 MHz and the Intel Pentium II with a clock rate of about 450 MHz, and both were single core. Today's AMD and Intel processors are multi-core (6 – 8) and run at clock rates in the order of 2.7 – 4.4 GHz, for first order theoretical speedups (clock rate x no. of cores) in the order of 54X to 78X. Other factors will contribute to the speed-up; memory size, bus speeds, number and size of processor caches, and architecture, but it is probably not unreasonable to assume a 100X effective speed-up between the home computers in 2000 and the home computers in 2014.

Contrast this speed improvement with the requirements for searching deeper without search tree pruning heuristics. Without search tree pruning the engine would have to look at <every> position arising from the legal moves from that position. Assuming a conservative (I think) average of 10 legal moves from each position, the 100X computer speed-up since 2000 would only allow searching 2 additional plies (10 x 10) beyond what was possible in 2000 in the same amount of time. Even a pessimistic reduction to an average of 5 legal moves from each position would only allow searching about 3 additional plies (5 x 5 x 5) beyond what was possible in 2000 in the same amount of time. And the branching factor per node is closer to 40 legal moves instead of 10.

The program HIARCS 14, although currently not among the 10 highest rated programs (ranked 14th in the Apr-26-14 CCRL tournament, so no slouch) has a useful (in this case) feature called Selectivity which allows you to control the extent of its pruning the search tree. At its highest setting, Selectivity = 7, it prunes its search tree aggressively and at its lowest setting, Selectivity = 0, it doesn't prune its search tree at all.

The differences are substantial. Starting with the initial position, using Selectivity = 0, HIARCS 14 required 01:23:01 to reach d=15, evaluating 7,536,030,558 nodes (positions). Using Selectivity =7, it required 00:00:02 to reach d=15, evaluating just 2,961,192 nodes. Even when going from Selectivity = 0 to Selectivity = 1 the differences were substantial; at Selectivity = 1 HIARCS 14 required only 00:01:01 to reach d=15, evaluating just 98,818,111. And, as you would expect, the number of Knodes/sec dropped from 1,507 for Selectivity = 0 to 1,163 for Selectivity =7, indicating that HIARCS was doing a lot more work per node, but the time spent was well worthwhile in terms of reducing the number of required position evaluations.

Even in 2004 all the engines did some search tree pruning, alpha-beta pruning at a minimum. But heuristics have improved tremendously since 2004, allowing the deeper search depths routinely achievable by today's top programs. And it is clear that much faster computers are not enough to explain the much deeper searches that are achievable nowadays.

May-03-14  DcGentle: <AylerKupp>:

<Here is a game from this later event in which Junior outplays Khalifman <positionally> (shocking!) in the White side of the Ruy Lopez with an early 4.d3 until its positional advantage is sufficient to first win a pawn, then gives it up (yup!) to retain its positional advantage, and eventually translate that advantage into a winning B+N+2p vs. B+3P endgame. After the game Khalifman apparently compared Junior's play to Karpov at his very best (http://en.chessbase.com/post/wham-2...)>.

Great, but I wouldn't be so proud on this achievement before you can assert to me that the engine played without an opening book. I went over this game and Khalifman more or less self-destructed, because he opened the game, granting the engine all tactical options it needs. The remark about Karpov was a nice excuse, but not fitting.

If engines are so good positionally as you claim they are, just tell any of them to find the win in Radjabov vs Carlsen, 2014. I am involved in a discussion there, because also to me it was not clear how to proceed at certain points, no wonder. Carlsen needed 101 moves in order to concede the draw and can you find anybody who doubts that this is correct? Engines of course also only show an about 0.0 eval, so this must be it. Draw, or not?

Tell current engines to offer a draw when appropriate, then we can talk.

;-)

May-03-14
Premium Chessgames Member
  AylerKupp: <DcGentle> Oh, I was partially pulling your leg, knowing your feelings about engines not being very good at playing positionally. I think that engines are capable of playing positionally <on occasion>, and this was one of them. And sometimes, maybe the majority of the time, they are not. But that's a long way from saying that they are not capable of playing positionally at all. It all depends on their evaluation functions, the positional factors included, and the relative weights assigned to them compared to the weights assigned to tactical factors. And these weights are typically a compromise and static; while some engines have different weights for the middlegame and the endgame, I don't know of any engines that recognize positions where tactics are premature and assign evaluations weights dynamically. It doesn't mean that there aren't any, just that I don't know of any.

As far as whether Junior used an opening book, I would assume that it did. But opening books typically address only the first 12 – 15 moves or so, therefore Junior had to find a way to exploit the position on its own. Opening Explorer has only 2 games through 15...Bg7, and this is the only one that continued 16.Be3. The 365Chess.com master's database (27,410 games with 1.e4 e5 2.Nf3 Nc6 3.Bb5) does not have any games beyond 14...g6, although it has 2 games with 15.Be3 (instead of 15.a3 as played by Junior) but then no games with 15...Qe7. The even larger Chess Tempo master level database (75,109 games with 1.e4 e5 2.Nf3 Nc6 3.Bb5) has 2 games with 16.Bb3 but none with 16...Qe7. So I think that it is reasonable to conclude that Junior was out of the book after 15...Bg7


click for larger view

And at this point Junior did not have much of an advantage. I don't have Junior but Komodo 6 at d=22 evaluates the position no higher than [+0.17], Houdini 1.6 at d= 24 evaluates the position no higher than [+0.10], and Critter 1.6a at d= 22 evaluates the position no higher than [+0.14]. So, even though these are low ply analyses, I think that it is also reasonable to conclude that Junior did not have a significant positional advantage when it played 16.Be3. Or at least it is safe to say that these engines did not evaluate White's position to be substantially better.

Khalifman did not try to open the game until 28...c4; previous game-opening attempts were by Junior by 22.bxc5 and 25.axb5 (and, of course, the preceding 23.a4), and by that time Junior had succeeded in opening and controlling the a-file as well as creating semi-open b- and d-files. I think that Junior deserves some credit for trying to open up the position rather than, as we've seen so often, engage in what I call "engine dithering" by moving its pieces back and forth starting with move 22. After all, one of the aspects of successful positional play, starting from a closed position, is recognizing when it is advantageous to open up the position.

And in the position after 28.Qxa1 Houdini at d=26 considered 28...c4 as being Black's best move, evaluating the position at [+0.35]. And Komodo also considered 28...c4 as being Black's best move, evaluating the position at [+0.42]. Good for White but hardly winning. So neither Houdini nor Komodo, at least at low ply, considered Khalifman's opening the position with 28...c4 to be a case of self-destruction. That perhaps came later. But 28...c4 did make the position assume a more tactical character which could not be to Khalifman's advantage. So maybe that was his real error.

But I don't consider this one game as "proof" of anything. Each of us can find examples and counterexamples of games in which the engines played well positionally and games in which the engines fell flat on their faces (assuming, of course, that they have faces). All I'm saying is that engines <sometimes> can play good positional chess, even though they are not (yet?) up to par with their tactical prowess. Maybe your engine will do better in this respect.

As far as Radjabov vs Carlsen, 2014, I had not seen your posts. I'll take a look at them in more detail and see if I can find anything concrete, with engine help, of course. But, frankly, I doubt that they will find a win for Black.

Finally, as far as engines offering a draw when appropriate, in this game, with the tournament rules and the tiebreak situation, the appropriate time for Black to offer a draw was after 41...Ne8, since that would have guaranteed Carlsen the tournament win. So I guess that Carlsen didn't know the appropriate time to offer a draw either. :-) (more leg pulling).

May-03-14  DcGentle: <AylerKupp>: *smile* Oh well. I don't claim that engines don't play any positional moves at all. Of course they do, because they can deduce them when detecting tactics which needs these moves as a preparation. In the Khalifman game it was already not good that Black played <14... g6>, because this doesn't grant Black any real success, as you can see from the Opening Explorer. And Junior answered with a novelty even, which might have been in its book. So how much positional engine performance we can admire in this game, remains questionable.

The genuine problem with the application of positional knowledge in current algorithms is the uncertainty, how important any positional criterion like piece mobility (for example) is. If you have a cramped position, it can be very important, even decisive for the game, but if it's only a temporary restriction of a single piece, it may be not so relevant. This is but one real dilemma current programmers face with positional play, there are others.

Anyways, this and other issues also show up in Radjabov vs Carlsen, 2014, and if Chessbase and other commentators say this is a draw because the engines don't find anything, they might be mistaken. On the game page I said that further analysis is required, and this is true. I guess I achieved a break-through there today, and now I have to verify the result by examining some sidelines.

Generally these endgames are also hard for engines, because the resulting lines can be so long. Greetings from the horizon problem.

All the best,
<DC>

NOTE: Create an account today to post replies and access other powerful features which are available only to registered users. Becoming a member is free, anonymous, and takes less than 1 minute! If you already have a username, then simply login login under your username now to join the discussion.

Please observe our posting guidelines:

  1. No obscene, racist, sexist, or profane language.
  2. No spamming, advertising, duplicate, or gibberish posts.
  3. No vitriolic or systematic personal attacks against other members.
  4. Nothing in violation of United States law.
  5. No cyberstalking or malicious posting of negative or private information (doxing/doxxing) of members.
  6. No trolling.
  7. The use of "sock puppet" accounts to circumvent disciplinary action taken by moderators, create a false impression of consensus or support, or stage conversations, is prohibited.

Please try to maintain a semblance of civility at all times.

Blow the Whistle

See something that violates our rules? Blow the whistle and inform a moderator.


NOTE: Please keep all discussion on-topic. This forum is for this specific game only. To discuss chess or this site in general, visit the Kibitzer's Café.

Messages posted by Chessgames members do not necessarily represent the views of Chessgames.com, its employees, or sponsors.
All moderator actions taken are ultimately at the sole discretion of the administration.

This game is type: RAPID. Please report incorrect or missing information by submitting a correction slip to help us improve the quality of our content.

Featured in the Following Game Collections[what is this?]
dlawtho's favorite games
by dlawtho
Clue #23
from Holiday Present Hunt Solutions: 2014 by Penguincw

Home | About | Login | Logout | F.A.Q. | Profile | Preferences | Premium Membership | Kibitzer's Café | Biographer's Bistro | New Kibitzing | Chessforums | Tournament Index | Player Directory | Notable Games | World Chess Championships | Opening Explorer | Guess the Move | Game Collections | ChessBookie Game | Chessgames Challenge | Store | Privacy Notice | Contact Us

Copyright 2001-2021, Chessgames Services LLC