|
< Earlier Kibitzing · PAGE 6 OF 6 ·
Later Kibitzing> |
| Nov-18-03 | | Benjamin Lau: Programmers tend to try to emulate positional chess, because computers cannot really play with a clear strategy in mind, but can play moves as if they do. For example, if a programmer wanted a program to favor piece activity and scope, he might assign a +0.02 evaluation value bias to each legal move available to each side. So if white has 50 available moves in total, he would get an extra +1 bonus while if black had only 25 available moves in total, he would get an extra 0.5 bonus, and the computer would thus favor white. But this is not a fool proof algorithm in real chess. The problem is that quality is as key as quantity. So the programmers might assign a higher evaluation bonus to based on other factors, like piece activity specifically in the center (which tends to be more important than when it's on the rim.) Maybe they assign a +0.05 bonus for each square in the center instead of +0.02. But then sometimes central piece activity can be ultimately irrelevant, like if there is some very strong play going on near the kingside and mate is threatened. It doesn't matter then how much mobility and control you have over the center if your'e down a king. So then yet again, the programmers have a problem, the computer does not realize the exceptions to the rules, and they must tweak the algorithm to realize higher truths. That's basically how programmers constantly revise programs- they know that it is impossible to show the computer how to play strategically because the computer is just a pile of metal and switches. Instead, the programmers each year try to find more and more accurate algorithms- they *won't* lead to positional chess, but they will emulate it so well in *most* positions that it would be considered as such. The above is not such a position. |
|
| Nov-21-03 | | Outcast Searcher: masig: <how do you programmers put value on gut feel and strategy?> You can put values on anything you can describe with an equation. The two chief "problems" are: 1). It is very difficult to accurately classify an entire position with equations. So, like Benjamin Lau described, typically many small values are assigned to lots of tiny features in the position. These values are all summed up to give an overall assessment of the position. These rules of thumb aren't consistently accurate - so the computer only plays "reasonably" where they are overall.
2. The goal is generally to have the best performance IN THE SHORT RUN. So far, the best way to do this has been to build a program that resembles a big hammer. It mostly just searchers TONS of positions very fast and uses tricks to eliminate as many bad positions from the search tree as quickly as possible. Then, the program that searches deepest tends to be the strongest. IMO, that's why programs haven't gotten stronger since Deep Thought won its match in 1997. Any gains in quality of positional understanding have been more than offset by running slower. (Deep thought could analyze about 200 million positions/second, whereas Fritz X3D only could handle 2 to 4 million, depending on which source you look at). Bigger hammers are FREE as time goes by (Moore's law -- computers get a lot faster).
For example, my first program (with me being completely ignorant of the chess programming literature and theory and only writing it for fun and creativity as a college project) I did something kind of unique for the time. I tried to use human-like judgement and only had the program look one full move deep and then analyze a LOT of things about those positions. So, for example, king safety could have a huge weight in a strong attack. The result was a program that would sac. pieces, even its queen. If it worked, it looked brilliant. If it didn't it looked like a total moron. BUT IT WAS FUN TO WATCH. It was fun to hear a group of people teamed up to play it swear a bloody streak as it chased their king across the board. Aside from the Chessmaster "Personality" emulation you see in the CM 6000 and above programs (i.e. part of the teaching mode), you just don't see that kind of stuff in chess programs, in my experience. Of course, if the goal is money or succss RIGHT NOW - that's what to expect, I guess. |
|
| Nov-21-03 | | damsel: TO BENJAMIN LAU OR ANYONE WITH ANANSWER If Gk offered Fritz a draw and Fritz was seeing a .01 advantage, would the programmers accept? ( if Gary had no time problems) |
|
| Nov-21-03 | | shadowmaster: Damsel, I believe the policy of the programmers was:
1) If the position showed Fritz down more than 2.0, they would consider resigning. In the third game, the computer showed pretty close to 0 but the programmers resigned because the commentators and spectators were deriding the play of the machine. 2) A draw would be taken only if it was obvious that the position was going to be repeated or if the material was inadequate. |
|
Nov-21-03
 | | Eggman: Incidentally, does anyone besides me think that Fritz's operators resigned a tad prematurely in game 3? Especially considering GK's blunder in the previous game, shouldn't they have played on? |
|
| Nov-21-03 | | PinkPanther: <Incidentally, does anyone besides me think that Fritz's operators resigned a tad prematurely in game 3? Especially considering GK's blunder in the previous game, shouldn't they have played on?> Honestly, I think they were tired of being embarassed by their creation. It made a number of rather comical moves becuase quite honestly it had no plan. And to make empty moves while Kasparov is lining up for the kill seems to me a good enough reason to go ahead and pull the plug (no pun intended). |
|
| Nov-21-03 | | drukenknight: No, I agree w/ Eggman, they pulled the plug way too early. But it becomes difficult to criticize these computer games, for a number of reasons. 1) moves that look silly may have to be played because of deep considerations; 2) some games appear to look staged or not for real, but that is just the nature of chess. We've seen that for a long time. From the games of Keres/Botwinnik, to the draws in Curacao, etc. Having said that, I still agree w/ Eggman. THe logic just doesnt stand to reason. How many moves had game 3 gone? almost 40? How many more perfectly sound moves would Kasp have had to make to win that game? ANother 25 or 30? Do the math. there's no guarantee that Kasp could find every single move. And we assume the computer will play flawlessly. Hence there is no logical reason not to keep playing. So again, I dunno. That's why it's dangerous to critique these games. Is there a tacit agreement between the parties, that if the man has an edge the machine will concede that, and not make him play it out? IT's possible. I would not think that too unfair either, since the man has to work hard. But again we dont know. even more suspicious was that draw a few months ago w/ Kasp/computer. After he had sacked the R. Why not play that out? |
|
| Nov-21-03 | | Benjamin Lau: <1) If the position showed Fritz down more than 2.0, they would consider resigning. In the third game, the computer showed pretty close to 0 but the programmers resigned because the commentators and spectators were deriding the play of the machine.> I think actually the advantage for white in tat game was 1.5 or so, not "close to 0." But the commentators were already giving lines to each other and hinting that the end was near. The programmers were tired of the humiliation, saw that the 1.5 was slowing getting bigger and closer to 2, and resigned. |
|
| Nov-21-03 | | Benjamin Lau: Btw you guys, wouldn't it make more sense to talk about the 3rd game on the page for the third game? |
|
| Dec-19-03 | | Chessnut Chris: Game 2 of the man V machine, Kasparov bludered against X3D Fritz and went down in flames. But was his position prior to his blunder winning? In the position that he bludered with 38...Rg7?, is there a winning chance in 38. Rc7!? My Junior 8 thinks that, after Fritz simplified the pawn structure on the queenside with 28.cxd6, black has all the play. The more I look at blacks attacking chances after 32...Rc7, instead of the fateful 32...Rg7, the better it looks for black. Is this winning for black? Probably not, but I'm really not qualified to say. But what about the line 32...Rc7 (grabbing the c-file) 33.Rc4 (otherwise black can gang-up on the weak d5 pawn with ... Qf7 and ...Rc5 etc.) Rxc4 34.Qxc4 h4 35.g3 h3 and black has a lasting advantage on the kingside and a better pawn structure. What do people think? |
|
| Jan-01-04 | | PapaiSabetudo: Why did Casparov resign in this game? Was he sick? I can't understand him. The game promised! |
|
| Jan-01-04 | | Helloween: <PapaiSabetudo> Kasparov had the black pieces in this game. |
|
| Jan-03-04 | | PapaiSabetudo: <Helloween>, I know that, but he resigned, didn't he? I can't see the winner way. It seems draw, doesn't it? |
|
| Jan-03-04 | | Prophylaxis: Look at the pages before, they contain analysis showing Kasparov's demise. |
|
| Jul-02-04 | | ArturoRivera: I think that kasparov resigned here beacause of the rule of the square, he cant stop two passed pawns after the exchange of queen and rooks he would have to stopp some very powerful pawns and the bishop would have been occupied on it, while he was also in minority of material in the kingside... |
|
| Sep-07-04 | | Knight13: Is X3D Fritz stronger than Fritz? |
|
May-28-05
 | | offramp: 32...Rg7 is a bad move - almost any other would draw. So what was the concept behind this move? |
|
| Jun-16-05 | | PinkPanther: Well, Kasparov was most likely playing for more than a draw. |
|
| Nov-21-05 | | Jmaal: (Pink Panther) I think that Kasparov should have thought out his move on 32... th draw was in his face! But that's ok. You learn from your experiences! |
|
| Mar-24-07 | | Brown: Give Kasparov credit for all the games where he outplayed the computer. We all blunder, but his conceptions, and general tactical and strategic awareness gave him many advantages in his games with computers. Unlike GM's after him, he played positions that were not "dry" but simply more positional. Of course, blunders happen, due to fatigue from playing such a way... |
|
| May-25-08 | | sfairat: I saw that 32. Rg7 was losing a mile away. I wish that made me better than Kaspy but it doesn't. |
|
| Aug-26-08 | | VaselineTopLove: Garry didn't learn from his loss to Deep Blue not to play the Spanish game against machines! |
|
| Sep-21-08 | | openingspecialist: [Event "Man-Machine World Chess Championship"]
[Site "New York Athletic Club, Manhattan"]
[Date "2003.11.13"]
[Round "2"]
[White "X3D Fritz (Computer)"]
[Black "Garry Kasparov"]
[Result "1-0"]
[WhiteElo "?"]
[BlackElo "?"]
[ECO "C65"]
[EventDate "?"]
[PlyCount "77"]
1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6 4.d3 d6 5.c3 g6 6.0-0 Bg7 7.Nbd2 0-0 8.Re1 Re8 9.d4 Bd7 10.d5 Ne7
11.Bxd7 Nxd7 12.a4 h6 13.a5 a6 14.b4 f5 15.c4 Nf6 16.Bb2 Qd7 17.Rb1 g5 18.exf5 Qxf5 19.Nf1 Qh7
20.N3d2 Nf5 21.Ne4 Nxe4 22.Rxe4 h5 23.Qd3 Rf8 24.Rbe1 Rf7 25.R1e2 g4 26.Qb3 Raf8 27.c5 Qg6 28.cxd6
cxd6 29.b5 axb5 30.Qxb5 Bh6 31.Qb6 Kh7 32.Qb4 Rg7 33.Rxe5 dxe5 34.Qxf8 Nd4 35.Bxd4 exd4 36.Re8 Rg8
37.Qe7+ Rg7 38.Qd8 Rg8
39.Qd7+ Rg7 40.Qc8 Rg8 41.Qxb7+ Rg7 42.Qc8 Rg8 43.Rxg8 Qxg8 44.Qxg8+ Kxg8
45.a6 d3 46.a7
1-0
|
|
| Oct-19-08 | | missingkasparov: THIS WAS FUN |
|
| Apr-13-09 | | WhiteRook48: o...k... |
|
 |
|
|
< Earlier Kibitzing · PAGE 6 OF 6 ·
Later Kibitzing> |
Kibitzing
