Author: Roy Brunjes
Date: 08:41:37 01/01/03
I came across this endgame in a software course I am using for chess training. I gave the position to Hiarcs 8, Junior 7, and Chessmaster 9000 and they seem to miss the point. I am wondering what Fritz 8 and Shredder 7 have to say about it. [D] 5k2/7p/6p1/1p3p2/2r2P2/5P2/PP5P/5K1R b - - In this position, with Black to move, Rc2 is the stated best move. I have no argument with that and neither do any of the programs I tried. The rub comes after this: 1. ... Rc2 2. Rg1 Which gives this position: [D] 5k2/7p/6p1/1p3p2/5P2/5P2/PPr4P/5KR1 w - - Now White is attempting to activate his rook. Programs I have tried (listed above) all want to play 2. ... Rxb2 after searches running for 3-4 minutes (admiteddly on my slow PIII 500 MHz laptop). This leads to a much more difficult ending that Hiarcs 8 was unable to win against me (and I'm a Class C player) at a time control of 10 mins + 10 secs per move. That tells me that the move Rxb2 is probably not best. The training software says 2. ... Rxb2 is a mistake and instead recommends 2. ... Rxh2! Then, after 2. ... Rxh2 comes 3. Rg2 Rxg2 4. Kxg2. Now both sides bring Kings to the center and Black must use his passed pawn on the h file to distract the White King from his job of keeping Black's King away from his pawns on a2 and b2. Once the White King is forced to head toward the Black h pawn, the invasion of the Black King wins the game. My question is this: What programs will play 2. ... Rxh2? And 3. ... Rxg2 [after 3. Rg2 by White] instead of 3. ... Rh4, and after how much time? Chessmaster 9000 wants to play 2. ... Rxh2 for a bit but then opts for 2. ... Rxb2 instead after a search of 7/12 ply. It seems to stick with that for quite a while and then finally changes back to 2. ... Rxh2 at ply 10/15 (4 minutes 25 seconds on my old, slow PIII-500 laptop). It seems to me that the principles are pretty clear and that a good endgame playing program should not need huge amounts of CPU time to select the right plan. Thanks in advance for putting Fritz 8 and Shredder 7 to the test. Feel free to try it with other programs as well -- I have particular interest in those two however as they are newer and I have no access to them. So far, only CM9000 does find the right sequence of moves, though it takes quite a few ply to do so. Yes, of course my faster computers would get there quicker, but still, they will require just as many ply to see the win. I would have thought a program with lots of endgame knowledge could spot the win fairly quickly. Lastly, Happy New Year to all CCC readers! Roy
This page took 0.01 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.