|
< Earlier Kibitzing · PAGE 4 OF 4 ·
Later Kibitzing> |
Feb-10-10
 | | playground player: How does the computer decide which opening to use? I mean, according to Opening Explorer, the Petroff is a bad choice for Black: the odds of White winning are roughly 2-1. So why would a computer ever choose this defense? |
|
Feb-10-10
 | | patzer2: <Morten: So are the rook moves on move 70 - 74 to be categorised as sacrifices or sham sacrifices?> I categorize them as stupid. Even if the Junior program recognizes the resulting endgame position as a win, it's ridiculous to give up that big an advantage. Certainly it's nothing for human players to model their play after, since most of us haven't memorized table base endings that deeply. |
|
Feb-10-10
| | patzer2: <playgroundplayer> I wouldn't base the probability of winning or losing with the Petroff or Russian Game on opening database results. Kramnik would have given up on this opening long ago if he really believed White has a 2 to 1 probability of winning. With strong play, it appears to be a good way for Black to equalize. |
|
| Feb-10-10 | | jmay: really a great game. very enjoyable for its formal qualities. Its chess, just not human chess. very, very interesting. In some post war holocaust, i imagine robots playing chess like this while they recharge batteries. |
|
| Feb-10-10 | | algete: ja ja ja
que chorrada |
|
Feb-10-10
 | | Once: <playground player: How does the computer decide which opening to use?> I suspect human intervention here! Left to their own devices, computers would presumably play the same white opening every time - ie the one that their databases or analysis told them was the strongest. But that would be pretty dull for a commercial program, cos we (the paying punters) would only get to play one part of our black repertoire. And it would be risky for competitive programs, as the opponents would know which openings to prepare for. So commercial programs have a random element built in. An off-the-shelf copy of Fritz or Rybka will switch between 1. e4 and 1. d4 on a fairly even basis, although it will rarely if ever deviate to crud like 1. h3 or f3. With competitive programs, the operators will sometimes choose the opening variation based on their own views of which is most likely to win. So Deep Fritz's use of the Petroff today is either random or based on a human decision. I don't think it means that Deep Fritz has analysed all the possible black responses to 1. e4 and decided that the Petroff is the strongest. |
|
| Feb-10-10 | | Yodaman: Wouldn't it have been faster for Deep Junior to win if he had thus thrown his rook to the left edge of the board and ignored it rather than trying to get black to take it? Was the computer not able to see that it could just ignore it? |
|
| Feb-10-10 | | lzromeu: <he had thus thrown his rook to the left edge of the board and ignored it> If DJ could do this, DF coud resign at 69. |
|
| Feb-10-10 | | kingsindian2006: what a relief to see kibitzing without alot of comments on lines from fritz or rybka and just chess..this computer seemed to make this endgame way harder than needed to be ...lol |
|
| Feb-10-10 | | SirChrislov: The never ending expanding universe of R + P endgame. |
|
| Feb-10-10 | | WhiteRook48: 33 Nb3 wastes time |
|
Feb-10-10
| | ajk68: If Deep Junior had a 6 piece tablebase, the game should have been over by move 85. At move 68 white simplified to a 6-piece ending. From there, it is a 17 move forced mate (with no sacrifice of a rook!). |
|
| Feb-10-10 | | chillowack: <patzer2: <Morten: So are the rook moves on move 70 - 74 to be categorised as sacrifices or sham sacrifices?> I categorize them as stupid.> I second that!
The only thing dumber than these moves are the explanations being offered for them by people in this thread. "Well, you see, the computer evaluates the resulting position as winning." So what are you saying, that the computer *doesn't* evaluate the position as winning *before* giving away the rook? Does it not know how to win with a +6 advantage? It has to pare it down to +1 before it can figure it out? If a human played this nonsense, it would mean he was mocking his opponent. When a computer does it, it's because "it evaluates the resulting ending as winning." Truly, computers are revolutionizing chess! |
|
Feb-11-10
| | Once: <chillowack: The only thing dumber than these moves are the explanations being offered for them by people in this thread.> A little harsh there, I think. At the risk of being accused of being dumb myself, let me try to explain what everyone has been saying. At the time that this game was played (2001) both computers had five man tablebases installed on their hard drives. In other words, when the position gets down to just five pieces on the board they knew exactly how to play it for a win or a draw. No calculation necessary - just follow the preprogrammed moves in the tablebases like following a recipe. Both computers would be programmed to exchange down into a winning tablebase position if they can. So if Deep Fritz had King, Queen and Rook against King and Queen, it would try its best to force the exchange of queens because it "knows" how to win with King and Rook versus King. And that is also what a decent human player would do. Around move 70, Deep Fritz has an overwhelming material advantage - two rooks and a pawn against a rook. It is winning by miles. But it can't yet calculate this position to a forced win, so the evaluation will be huge but not yet a forced mate. There are six pieces on the board, so its tablebases do not tell it how to play precisely to mate. By contrast, its tablebases can play five man positions perfectly. So Deep Fritz has a choice - keep both rooks for a huge evaluation or throw away one rook to get down to a five man ending which it can play perfectly to a forced mate. Huge advantage or guaranteed mate? No human would play this way, in part because we haven't memorised five piece tablebases like the machines have. So the computers' behaviour is weird, but perfectly rational when you know how they are programmed. And the kibbitzing is far from dumb. |
|
| Feb-11-10 | | lzromeu: Ok... Once. Perfect.
If DJ is programmed to see 15 moves, It can't see a forced mate in 17 with 2 rooks, then choose to simplify at tablebase and forced mate with 1 rook.
Very "conservative" choice.
Machines plays what we programm, not what we want. |
|
Feb-11-10
 | | kevin86: Why is white giving away his rook advantage. It makes no sense. But,I guess I'm a mortal human and not an advanced CONFUSOR. |
|
| Feb-11-10 | | ounos: <chillowack:
The only thing dumber than these moves are the explanations being offered for them by people in this thread. "Well, you see, the computer evaluates the resulting position as winning." So what are you saying, that the computer *doesn't* evaluate the position as winning *before* giving away the rook? Does it not know how to win with a +6 advantage? It has to pare it down to +1 before it can figure it out?> Please read my second to last post, above. I presume you find it trivial that "plus infinity" is considered greater than "+6" (or +10 as it really was), so I don't understand what remains unclear to you. The program is simply coded to choose the move which maximizes its evaluation, simple as that. |
|
| Feb-11-10 | | lzromeu: The pun is perfect, specially after dumb posts |
|
| Feb-19-10 | | Mulyahnto: <chillowack:
The only thing dumber than these moves are the explanations being offered for them by people in this thread. "Well, you see, the computer evaluates the resulting position as winning."?>please read the posts before criticizing them. It has been said that the computer evaluates the position as <a forced mate> aka <a tablebase win>. A table base win does not need to be calculated because it is pre calculated (hence table base), so the sacrifice makes sense. It takes less thinking time but more moves. Since more often than not in chess time is a limiting factor, and not move count. |
|
| Sep-27-10 | | meppi: here is one possible continuation taken from deep juniors database at move starting at move
73. Fritz = Kc3
74. Rb3+ Kc2
75. Rd2+ Kc1
76. Rbb2 and the king is trapped!
BUT this is a horrible blunder allowing Fritz to obtain a crazy rook
76. Fritz plays Rf8+
77. Kg3 Rg8+
78. Kf2 Rxg2+
And both future boxes have fried a chip fatal error do not compute1010101 |
|
| Sep-27-10 | | meppi: (from above) 78. Ke1. Rf2
and it looks very nice |
|
| Nov-17-10 | | BarcelonaFirenze: Why the computer decided to promote a rook? Does it make any sense? |
|
| Nov-27-11 | | asianwarrior: <BarcelonaFirenza> The king will capture it anyway. |
|
| Nov-28-11 | | King Sacrificer: I'm a little late to the discussion but i want to say something which can make people understand the moves 70-74 easier. Actually White doesn't think about giving the rook away at move 70 to reach the predefined mate. According to his calculations, this move is safe (Black can't take the Rook) because Black walks into mate if he takes the Rook. This fact makes 70.Rc6 the best move for White. |
|
| Dec-06-16 | | LA MAN: Unless I missed something on 30th if rxnp check he can win the knight for once he moves the king supports his rook. |
|
 |
|
|
< Earlier Kibitzing · PAGE 4 OF 4 ·
Later Kibitzing> |
Kibitzing
sac: 74.Rc5+