Author: Ricardo Gibert
Date: 11:30:32 01/21/03
Go up one level in this thread
On January 21, 2003 at 12:51:30, James T. Walker wrote: >On January 21, 2003 at 08:30:05, Ricardo Gibert wrote: > >>On January 21, 2003 at 07:48:51, James T. Walker wrote: >> >>>Here is the analysis by Fritz 8. The first few moves are perfect and I really >>>believe that Fritz 8 would win this ending easily. >>> >>>New game >>>8/1p1R2pk/5p2/7P/7P/5Kn1/4r3/8 b - - 0 1 >>> >>>Analysis by Fritz 8: >>> >>>1...b5 2.Rb7 Re5 3.Kxg3 Rxh5 4.Kg4 Kg6 5.Kf3 >>> -+ (-1.47) Depth: 7/14 00:00:00 28kN >>>1...Re8 2.Kxg3 Rb8 3.Rd4 b5 4.Rb4 Kh6 5.Kg4 >>> -+ (-1.88) Depth: 8/16 00:00:00 51kN >>>1...Re8 2.Kxg3 Rb8 3.Rd4 b5 4.Rb4 Kh6 5.Kg4 >>> -+ (-1.88) Depth: 8/16 00:00:00 51kN >>>1...Re8 2.Kxg3 Rb8 3.Rd4 b5 4.Rb4 Kh6 5.Kg4 >>> -+ (-1.88) Depth: 8/16 00:00:00 51kN >>>1...Re8 2.Kxg3 Rb8 3.Rd4 b5 4.Rb4 Kh6 5.Kg4 Rb7 >>> -+ (-1.81) Depth: 9/17 00:00:00 101kN >>>1...Re8 2.Kxg3 Rb8 3.Rd4 b5 4.Rb4 Kh6 5.Kg4 Rb7 6.Re4 >>> -+ (-1.97) Depth: 10/21 00:00:00 208kN >>>1...Re8! >>> -+ (-2.25) Depth: 11/20 00:00:00 409kN, tb=2 >>>1...Re8 2.Kxg3 Rb8 3.Rd4 b5 4.Rb4 Kh6 5.Kg4 >>> -+ (-2.19) Depth: 12/20 00:00:01 832kN, tb=17 >>>1...Re8-- >>> -+ (-1.91) Depth: 13/24 00:00:02 1749kN, tb=171 >>>1...Re8 2.Kxg3 Rb8 >>> -+ (-1.91) Depth: 13/25 00:00:03 1927kN, tb=174 >>>1...Re8! >>> -+ (-2.19) Depth: 14/25 00:00:06 4103kN, tb=595 >>>1...Re8-- >>> -+ (-1.91) Depth: 15/26 00:00:12 7285kN, tb=2014 >>>1...Re8 2.Kxg3 Rb8 >>> -+ (-1.91) Depth: 15/26 00:00:13 7914kN, tb=2061 >>>1...Re8! >>> -+ (-2.19) Depth: 16/28 00:00:26 17073kN, tb=5446 >>>1...Re8! >>> -+ (-2.47) Depth: 17/32 00:00:49 30631kN, tb=13041 >>>1...Re8 2.Kxg3 Rb8 3.Kf4 b5 4.Ke3 b4 5.Kd2 b3 6.Kc1 b2+ 7.Kb1 Kh6 8.Rd5 Rb7 >>>9.Rc5 Rb6 10.Rd5 Rb3 >>> -+ (-2.34) Depth: 18/33 00:01:34 63970kN, tb=32815 >>>1...Re8 2.Kxg3 Rb8 3.Kf4 b5 4.Ke3 b4 5.Kd2 b3 6.Kc1 b2+ 7.Kb1 Kh6 8.Rd5 Rb7 >>>9.Rc5 Rb6 10.Rd5 Rb8 >>> -+ (-2.41) Depth: 19/54 00:03:09 144284kN, tb=78986 >>> >>>(, 21.01.2003) >> >> >>This PV is garbage. All this shows is that F8 likes 61...Re8, because all it >>could find was a weak defense for White (63. Kf4), which does zip to resist the >>formation of connected passed pawns on the K-side for Black. Like I said, chess >>programs are clueless when it comes to Rook endings. Also, the fact that it does >>not switch to 61...Re8 until ply 8 is suggests that it likes 61...Re8 for the >>wrong reasons. > >I assume you are at least a Master and know what you are talking about. Since >I'm only a weak beginner I am tempted to take your word for it. What bothers me >about your post and I see this often in chess magazines, is that you seem to >think that your way is the only way and every thing else is wrong. The fact >that it finds the right move in "0" seconds says something for the program since >most all others don't find it at all. Doing 8 plies in about 0.1 seconds is >nothing to complain about. Since computers don't have human traits such as >intuition they have to gather some information before they can make the right >decision. As you yourself say below, just adding a bonus for getting the Rook >behind a passed pawn is not enough. So why would you want the computer to pick >Re8 without searching several plies first? Do you expect it to see the win to >the end of the game in 2 or 3 ply? My remark about 8 plys is directed against the idea it might be awarding a significant bonus for getting the Rook behind the passed pawn. It appears not to or it would have shifted to 61...Re8 after reaching about 3 plys. The question therefore is why it would like 61...Re8, so I look at the PV and it gives a ridiculous move 63.Kf4 for White. It appears then that it likes it, because it has furnished White with a weak defense rather than for a legitmate reason. I want F8 to play the move for a sensible reason and a sensible reason is not apparent. On the other hand, it could be it sees all the winning stuff and plays the weak 63.Kf4, only because nothing is satisfactory anyways. This is typical behavior when programs start getting a lot of hits in EGTB, though it can happen at other times too such as in the diagrammed position. > >> >>The basic problem with Rook endings for programs is that they are strategic in >>nature and to make up the difference, programs need to calculate to a depth that >>is beyond what they can do. Just adding a bonus for getting a Rook behind a >>passed Pawn is not enough, because it happens often enough that this is not even >>the right idea and that bonus will then work against the program instead of for >>it. > >I'm sure you are right but I still believe Fritz 8 would win this endgame vs >another program or probably even you. >Jim After 61...Re8 62.Kxg3 Rb8 I would agree (Nor after 61...b5 as it turns out. See below). It would for sure win this against me, but this is because the position is so gross, there is more than one way to win it and note I did give more than one way. I decided to have a look at it again with Ruffian and coerced it to play 61...Re8 62.Kxg3 Rb8. Instead of the 2 ideas I gave, it came up with yet a third way of winning. Playing the White side is just hopeless after 61...Re8. Next I looked at it with Ruffian after 61...b5. This too looked hopeless to me and now I got curious as to just how Crafty managed to draw this. I decided to take a closer look and the reason turned out to be simple. Searcherx preferred to defend the b-Pawn from the side rather than from the front (or rear). From the front, Black can advance the b-Pawn. This leaves White with 2 unsavory choices: 1) Move the King to the Q-side. This nets the b-Pwan, but then the Whites K-side Pawns will drop leaving Black connected passers that win. 2) Keep the King on the K-side, but then the b-Pawn reaches the seventh and the White King gets stranded on g2 and h2. Now the advance of Blacks f-pawn kills this. In short, Crafty did zip to earn its draw. Searcherx did all the work for it. It seems to have a terrible eval bug. BTW, it is worth noting that Ruffian does not make the mistake of preferring to defend the b-Pawn from the side after 61...b5, so I couldn't even hold this ending against many programs even when they *don't* play 61...Re8. In fact, it turns out that 61...b5 is just as good as 61...Re8! Thank you for your reply. I wouldn't have realized this otherwise. What I *can* do that not all other programs can do is win from the Black side against any program. Now *that's* a piece of cake.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.