|
< Earlier Kibitzing · PAGE 2 OF 2 ·
Later Kibitzing> |
| Aug-31-10 | | Marmot PFL: Rybka is always a step ahead of these clones. |
|
| Aug-31-10 | | crazybird: <Marmot PFL: Rybka is always a step ahead of these clones> Some say Rybka itself was a clone |
|
| Aug-31-10 | | Marmot PFL: The clone can never be better than the original, and Rybka is always #1. |
|
Aug-31-10
 | | OhioChessFan: Love the underpromotion to Rook. I didn't know computers could also say "It's time to resign". |
|
| Aug-31-10 | | kevin86: A crazy game;the underpromotion is a pixieish ending. The stockfish syndrome:when the losing computer identifies with the winner. lol |
|
| Aug-31-10 | | pers0n: don't computers try playing 'correct' chess? 'correct' chess would have meant a promotion to a queen, not rook. |
|
Aug-31-10
 | | Once: I think computers play whatever kind of chess their programmers tell them to play. Which means that some programmer has a wicked sense of humour - to program Rybka to play an underpromotion if it leads to the same result as a promotion to queen. |
|
| Aug-31-10 | | desiobu: f1=R. Brilliant. |
|
| Aug-31-10 | | hstevens129: The combo begins at 41...Nf3. Incredible depth of calculation. |
|
Aug-31-10
 | | Phony Benoni: Perhaps Rybka underpromoted because there wasn't another queen handy. |
|
| Aug-31-10 | | garrykasparov: Promotion to a Rook or Queen wouldn't really matter. |
|
| Aug-31-10 | | waustad: Underpromotion is a common feature if you look at the "preferred" lines from stockfish or rybka. If the piece will be taken immediately it usually picks a lower value one or if it doesn't matter, like here, it will usually pick the weaker one. Watching games in chessbomb or chessok one sees this often. |
|
| Aug-31-10 | | eykca: <Phony Benoni: Perhaps Rybka underpromoted because there wasn't another queen handy.> It's the simple humor in life... thank you Phony, I needed that. |
|
| Jan-07-12 | | patzer777: Well also you see a lot of cases wherein the lesser piece is needed in order to avoid a stalemate, but will still get the job done. The designers probably figure that it's better to be safe than sorry. |
|
Jan-07-12
 | | AylerKupp: There may have been a change in Rybka's evaluation function since 2009 when going from version 3 to version 4. Starting from the position after 77.Ka4, the current version (Rybka 4.1) promotes the f-pawn to a queen regardless of whether tablebases are enabled or not. Too bad, the game continuation made it more interesting. I should add that with tablebases disabled Rybka 4.1's PV#1 goes 77...f1=Q 78.Ka3 Qfa6# but its PV#2 goes 77...f1=R 78.Ka3 Ra1#, just like the game. I guess that, mate notwithstanding, Rybka's evaluation function shows an advantage in material at the end if it promotes to a queen rather than a rook and therefore chooses the queen promotion over the rook promotion! |
|
Jan-07-12
 | | Penguincw: Sigh. More underpromotion. Computers seem to like to do so. |
|
| May-20-13 | | iamsheaf: 41 .. Nf3 could very well be one of the best combinations ever played on a chessboard |
|
| May-20-13 | | PaulLovric: what about 50...Nxc2+, is that a mistake? 50...Nd3+ forking the queen seems like a better move at first glance??? |
|
| Nov-25-13 | | checkmateyourmove: <dortam888> brings up more beauty to this game. Not only does rybka play 41...nf3 but it doesn't do some moves likes 50...nd3 forking the queen.
Which moves jump out at you more , the ones being played by the silicon monsters or the ones they pass up on? |
|
| Nov-25-13 | | Everett: My guess is that 50.. nd3 wasn't played because Rybka likely "saw" that there was an easy k and p ending with ..Nc2. And I think Tal sees 41..Nf3 |
|
| Feb-15-14 | | PJs Studio: This game is a beauty. Positional accuracy on both sides (IMO) and then stunning pressure when the position cracks... These machines are so amazing. But don't beat yourselves up to much comparing yourselves to it, even the Kaparov like Usain Bolt can't beat a Ferrari in a foot race. Nor can a football player tackle one on the grid iron. Stockfish kicks the absolute poop out of me, daily. |
|
| Feb-15-14 | | whiteshark: As Tartakower once commented: <White missed excellent resigning chances!> |
|
Feb-15-14
| | AylerKupp: <<Penquincw> Sigh. More underpromotion. Computers seem to like to do so.> Well, of course, computers don't "like" to do anything. A more likely explanation was the order that the moves 77...f1=R and 77...f1=Q were loaded into the search tree. Since the evaluation of either move is the same (this would also apply to situations where the promoted piece is captured immediately), the computer picks either the first or the last move that it evaluated (depending on how the engine was programmed). I would like to believe other explanations like the computer picks the most "efficient" promotion that will get the job done, the computer has a sense of humor, etc. But I don't think so. And it often is a matter of chance. My post on Jan-07-12 was before I realized the extent of non-determinism in computer engines, particularly multi-core ones. If you run the analysis starting In the same position Rybka will sometimes pick 77...f1=R and sometimes 77...f1=Q. Rybka, like other engines, sometimes does play dice with the universe. |
|
| Dec-23-14 | | drdos7: <Marmot PFL: Rybka is always a step ahead of these clones.> Stockfish isn't really a clone, but it is a derivative of Glaurung and approved by Glaurung's original author Tord Romstad, as a matter of fact Tord has helped in the development of Stockfish. Also Stockfish has surpassed Rybka now by leaps and bounds, and would crush Rybka in a match now. |
|
Dec-23-14
| | AylerKupp: <drdos7> But, clone or no clone, Rybka was still a step ahead in 2009, but it wouldn't be for long. Houdini 1.0 was released in May-2010 and Houdini 1.5 was released in Dec-2010. When Houdini was finally entered in the CCRL engine competition in Mar-2011, it was ranked #1 and rated 68 Elo points higher than Rybka. Mortuus est autem rex, vivat rex. |
|
 |
|
< Earlier Kibitzing · PAGE 2 OF 2 ·
Later Kibitzing> |
Kibitzing
sac: 43...Qh2