< Earlier Kibitzing · PAGE 2 OF 3 ·
|Feb-13-06|| ||jahhaj: <Bobak Zahmat> I think it's a clear case of the horizon effect. The explanation given by <Professeur Y> above is the correct one even tho' he seems to think it's rubbish.|
Happens much less these days since computers are much faster but it used to be really common for computers to make mistakes like this.
|Feb-13-06|| ||dakgootje: Strange Fritz made this blunder...
Though it could technically be a brilliant game, but to be honest...i wouldnt make it a GOTD...
|Feb-13-06|| ||CapablancaFan: Fritz is essentially a "numbers cruncher". "It" only saw that it was ahead in material in which it is programmed that it good to stay ahead. I have played fritz many times and noticed that fritz will only sacrafice if it "sees" a forced mate, exchange sacrafice, or a postion so superior that it is almost impossible to overcome. In this game I'm not surprised fritz made the exchanges that it did, but the problem was fritz has a more difficult time analyzing the postional considerations. Yes, materially fritz is ahead, but because of Ponomariov's active play and a file pawns the bishop's life will soon end and so will fritz's game.|
|Feb-13-06|| ||kevin86: The computer has a strange name:"on the Fritz" usually means that a machine is broken-in this case it looks that way. The human equivalent would be "on the Schlitz".|
|Feb-13-06|| ||patzer2: <wintep> <Simply 39...Qxg3+ 40. Kxg3 f5 is good for Black.> I agree! In fact it looks to be winning for Black. The move 39...Bc2? is a blunder that I can't understand the computer making it. Strong human players would likely play 39...Qxg3 even in blitz, and wouldn't ever consider a dud like 39...Bc2?|
P.S. In his victory over Fritz, Kasparov also took advantage of the computer program's occasional weakness in calculating the impact of a potential passed pawn.
|Feb-13-06|| ||Joshka: <Patzer2> I'm using Fritz 8 only, purchased in 2003, let it run at the highest level and after about 7 or 8 minutes of thought, it played 39...Qg3+!! And NEVER even considered 39...Bc2?? the losing move which Fritz 9 played!!.....is there a bug in Fritz 9?...in fact I run my Fritz 8 on and old Dell laptop running only Windows98!!|
|Feb-13-06|| ||belka: At the end of their lookahead, computers have to judge by positional elements. Positionally, Fritz gave up a passed but blockaded and isolated rook pawn, and got 2 pieces for a rook in compensation. In general, is that a good trade or not? That's a tough question to program a computer to answer. |
In general, if you couldn't see that your blockade would be broken, that would probably be a good trade, wouldn't it?
"It depends" is a better answer. It depends on whether or not White has an effective plan to break the blockade eventually. The computer can only see so far, and it can't anticipate its opponent's plan. So, it thinks its blockade is sufficient and makes the trade.
|Feb-13-06|| ||patzer2: <Joshka> I'm also running Fritz 8, and I also see no sign that it is inclined to play 39...Bc2? However, it took nearly five minutes to find 39...Qxg3+!, initially preferring 39...Nc4!? for some strange reason.|
|Feb-13-06|| ||patzer2: <...is there a bug in Fritz 9?> Don't know, but now I'm in no rush to upgrade my Fritz 8.|
|Feb-13-06|| ||AlexanderMorphy: interesting game...i followed this live|
|Feb-13-06|| ||Joshka: <Patzer2> Allright so that's your Fritz8 and mine, seems that this cannot be just a coincidence! Damn wonder if the chessbase folks even played Fritz 9 in a match against their 8 version to see if it WAS stronger?? Damn now I'd like to run my own test! Anyone who has Fritz9 wanna challenge me and my Fritz8 to a match? say best of 10 games? I'd want a refund, if I had purchased 9 and discovered that 8 could beat it in a head to head match!|
|Feb-13-06|| ||al wazir: <ReikiMaster: I think 34...Nbxd5 35.Bxd5 Qxh3 36.Bc6 Qxg4+ or 36...Nxg4 both look good for black.>|
What's the continuation after 37. Qg2 ?
|Feb-13-06|| ||Sigmund: Here, Ponomariov didnt won, Fritz gave him the game,and lose in a terrible blunder.|
|Feb-13-06|| ||jackmandoo: Facinating Captin. Anyway here is the "jackmandoo chess quote of the day".|
"We were like two knights protecting each other, we were both afraid to move because then we wouldnt be protecting each other. We were both powerful together in the middlegame of our relationship, but then the king marched to the other end of the board and we became useless."
Jackmandoo describing why he broke off his engagement to singer Tina Turner.
|Feb-13-06|| ||ajile: I believe this opening is a London System which is basically a Colle system with the Q-bishop outside the pawn chain on f4.|
Computers are generally bad in closed positions. That's why these Colle/London systems work so well. White provokes and gets Black to close the center with 13..e4. After that White outplays the computer strategically.
|Jul-14-06|| ||birnal: insted of this
39. ... Bc2 ??
39. ... Qxg3+ !! 40. Kxg3 f5 41. Rg1 Kf7 42. Kf2 f4 43. Rxg5 fxe3 44. Kxe3 Bc4 45.
Kd2 Rd3 46. Kc2 Bxb3 47. Nxb3 Nxa4 48. Na5 Rxd4 *
Panomariov would lost that
Toga II 1.2.1a 32 is better than Fritz and would'nt have made this mistake.
Panomariov wouldnt have win with Toga II 1.2.1a
|Jul-15-06|| ||lopezexchange: Obviously an example of the horizon effect or that the computer is programmed to move to quickly, thus not reaching a sufficient depth. |
The computer missed an easy win: 39...Nc4; 40.Qxh4,gxh4; 41.Bxc4,dxc4; 42.b5,axb5; 43.axb5,Ra3; and black won.
|Dec-01-06|| ||argishti: there is still hope for humanity!|
|Feb-28-08|| ||positionalgenius: Nice game...|
|Jan-12-11|| ||picard: The default settings on my chessmaster 2002 program suggested 39...Bc2 However, after i adjusted it to play more positionally it never even considered Bc2. It debated between Nc4 and Qxg3+ for about 5 seconds before settling on Qxg3+ Kxg3 f5. even if Fritz was short on time it should have never even considered Bc2, not even for a nanosecond. I think this version of fritz was programmed to be too materialistic.|
|Jan-19-12|| ||JustToClarify: This is a rather notable game because it was the very last game a human played against a top computer in a tournament without any odds and won (November 21, 2005)|
|Nov-21-12|| ||JointheArmy: ^^Almost seven years to the date!|
|Apr-21-15|| ||Penguincw: About 9 years and 5 months later, this game remains the last time a human beat a computer in a regular, classical time controlled game.|
|Jun-07-15|| ||apexin: Back when humans were able to beat a computer ...|
|Oct-18-15|| ||Kinghunt: We are now coming very close to the 10th anniversary of this game, and it is now certain that this will remain the last game humans were ever able to win against a top computer (and even here, it depends on a very unusual blunder by Fritz).|
As we approach the 10th anniversary of this historic game, I thought I would look at how far computers have come since then, both hardware and software.
First, hardware. This game was played on a Centrino 1.7GHz notebook (source: http://en.chessbase.com/post/man-vs...). The current season of TCEC is being played on 2 x 10 core Intel Xeon E5-2650v3 @ 2600 MHz. By looking at benchmarks for these processors, that is a roughly 120x increase in hardware. Using the rule of thumb of a 2x speed increase being worth 50 Elo, hardware alone has contributed 350 Elo to the increase in computer strength.
Second, software. This is much easier to quantify accurately. This game was played by Fritz 9, which is rated at 2743 by CCRL. On a single core, the current top engine, Komodo 9.2, is rated 3274. This is a 531 Elo gain from software.
Together, these suggest that on strong but reasonable modern hardware, the current top engines are ~880 Elo stronger than the program Ponomariov defeated here. (This is certainly true for engine vs engine, but it's possible that human ratings scale slightly differently.) So using a conservative estimate of ~500 Elo gained, this suggests that against humans, modern computers should be rated ~3300.
So we can be extremely confident that under tournament conditions, no human will ever beat a top program again.
< Earlier Kibitzing · PAGE 2 OF 3 ·