< Earlier Kibitzing · PAGE 14 OF 14 ·
|Feb-28-13|| ||kingscrusher: Interesting - 50 move draw rule bug exploited|
|Feb-28-13|| ||paulalbert: <kingcrusher>
I was intrigued by your comment on the 50 move rule. I think there was a piece capture or pawn advance at the necessary and appropriate time throughout all phases of the game, and the computer rigorously avoids any 3 position repetition. Certainly 50 move issue no problem in the last phase of the game, 233... e1=B would have made move 283 end of 50 move line and Hikaru pulled off the mate before the limit.
Enjoy your videos!
|Feb-28-13|| ||paulalbert: Just noticed Hikaru gave up a B on 237 so 287 actually the limit.|
|Mar-03-13|| ||kingscrusher: Move 174 - c4 - I thought this may be an inferior move designed to avoid the 50 move draw rule|
|Mar-03-13|| ||AylerKupp: <kingscrusher> I agree. I came to a similar conclusion after the following game Aylerkupp / Rybka vs Kutztown46 / Stockfish, 2011, when Rybka 4.1 played 94.b4 in order to, I believe, avoid a draw by the 50-move rule. Rybka's eval of the position after Stockfish's 90...Rc8 was [+0.65] and after 94.b4 it was [+0.16], still better than the [0.00] that it would have evaluated the draw.|
There is an analysis of this game by GM Gserper in http://www.chess.com/article/view/c... that I just came across. He makes the same point about Rybka apparently going out of its way to avoid a draw by the 50-move rule.
So apparently this bug is still in the latest version of Rybka. I wonder if Stockfish (or Stockfish's developers) knew about it. :-) Vasich Rajlich, take note!
|Mar-04-13|| ||paulalbert: <kingcrusher> and <AylerKupp> Thanks for the clarification! Since I don't play against these machines, I thought you meant the bug was failure by the machine to count the 50 moves and claim a draw rather than making an inferior move to avoid it. I suppose at 174 c4 the machine possibly still gave itself a positive evaluation because of the material plus, so avoiding the 50 move rule by an inferior move might be a rational programming approach in order to still try for a win, but the evaluation routine must fail to see the long term consequences of turning a totally closed position into an open one.|
|Mar-31-13|| ||perfidious: Classic stuff.|
|Mar-31-13|| ||OhioChessFan: It's nice to see an engine look like a total idiot.|
|Apr-08-13|| ||kingscrusher: I have video annotated this game here :
|Aug-09-13|| ||GumboGambit: According to Rybka, Nakamura is a tool.|
|Sep-07-13|| ||Snehalshekatkar: Fake.. no comp would start from g4 and h3|
|Oct-23-13|| ||SicilianStar: This is fake. It is clearly a draw for a very long time but a draw is never excepted. In the near the end Rybka starts making blunders one after another
and somehow manages to lose. On move 194 white moves his rook to a square where a pawn can take it but black does not take. This would never have happened in a real game.|
|Oct-23-13|| ||AylerKupp: <SicilianStar> Remember, this was a blitz game. And I'm sure that Rybka was <very> short of time. And you obviously have never seen Rybka play when it doesn't have time to search more than one ply. Yes, it does do things like that. I've seen its play disintegrate when playing against Houdini when very short of time (and Rybka does tend to get into big time trouble on long games). |
And Rybka apparently will <NOT> accept a draw if it evaluates its position as superior. If you don't believe me, take a look at this game: Aylerkupp / Rybka vs Kutztown46 / Stockfish, 2011 and the commentary that follows it. This game is not a fake; I was (obviously) there and saw it happen in real time.
|Oct-23-13|| ||AylerKupp: <kingscrusher> I just saw your video of this game and heard your reference to Rybka's 50-move rule bug. As you can see if you look at my link above and Rybka's 91.b4 it is clear that the bug has not been removed since the version used in this game was the latest and current version of Rybka, 4.1. Maybe Vas Rajlich doesn't know about it ?!|
I wonder if there are any special "anti-Rybka" versions of other engines programmed to take advantage of this. :-)
|Oct-03-14|| ||Domdaniel: <Ohio> -- < It's nice to see an engine look like a total idiot.>|
They're *all* idiots. It's just that some of them play chess better than the rest of us idiots.
With exceptions, obviously.
|Nov-27-14|| ||welhelm1982: The pieses finished their school during this game and they get marrage and had achidren and and and|
|Feb-13-15|| ||manata27: Well i'am more impressed by Nakamuras coordination then by his (of course extremely tough) skills. I would just not be able to play 255 moves in three minutes.|
|Apr-08-15|| ||rahulgaikwad: lol, Hold your keyboards left key pressed for certain amount of time and then see the fun as how black rooks dances up and down and then bishops make a dance sideways and in the end the climax. I thought there was an error in the program. But to my amazement, so many bishops was a true story.|
|Apr-12-15|| ||Chichiboy: Perhaps Nakamura should enter the book of records for having had the most bishops on the board. Perhaps it's the game with the most bishops by an opponent(6).|
|Jul-25-15|| ||Ke2: this is like the crowning achievement of online chess|
|Oct-13-15|| ||FRED MALAWA: AMAZING GAME|
|Oct-19-15|| ||Eusebius: Rybka was programmed to avoid the draw...ridiculous. Games like this will mark the end of the chess game.|
|Aug-14-16|| ||esteger: Much was contributed to the theory of the king and six bishops versus lone king endgame.|
|May-05-17|| ||Dave12: On move 251 Nakamura missed a fine mate on the middle of the board with Bg5!!|
|Jul-18-17|| ||Chuckles: I don't really understand calling this a bug in Rybka. Draws are evaluated at 0.00, so if it evaluates avoiding the draw at any positive score, it will avoid the draw. What else should it do? I guess you could program it to always allow 50-move draws, on the basis that if there's been no progress in those 50 moves, then the position is probably a fortress like this.|
The basic issue is that computers often can't evaluate fortresses properly, but that's nothing to do with the 50-move rule specifically.
< Earlier Kibitzing · PAGE 14 OF 14 ·