Author: Robert Hyatt
Date: 20:54:08 11/10/99
Go up one level in this thread
On November 10, 1999 at 22:39:53, Will Singleton wrote:
>On November 10, 1999 at 20:47:40, Robert Hyatt wrote:
>
>>On November 10, 1999 at 20:32:58, Will Singleton wrote:
>>
>>>r2r2k1/pp3ppp/2n2b2/4p3/2N5/1PP2B2/1P3PPP/R2R2K1 b - -
>>>
>>>This position gives my program fits. Some stronger progs solve it easily. It
>>>appears to be a simple case of extending on captures, and then doing the qsearch
>>>right, which I apparently am not.
>>>
>>>Here's the main line which should be found:
>>>Rxd1 Rxd1 Rd8 Rxd8 Nxd8 Nd6 b6 Nc8 Bg5 Nxa7 Bc1 and gets the pawn back.
>>>
>>>It appears that you need 13 ply to see that the bishop can get the pawn back.
>>>Extending on a capture that brings material back to nominal, there are 2
>>>extensions in the above main line, so you now need 11 ply. The qsearch reduces
>>>this to 10. And that's what my program needs, 10 ply. Seems too long.
>>>
>>>I would love to find this sooner without blowing up the tree.
>>>
>>>(from a game on ICC today, Postmodernist vs Amateur)
>>
>>
>>There are two ways you have to look at such positions: (1) what is the right
>>move and why is it right? (b) what is the second-best move and why is it wrong?
>>
>>In this position, crafty isn't what I would call "happy". It plays Rxd1 after a
>>second or so on my xeon, at depth=7. But the score is roughly -.6 and it
>>doesn't get better even at depth=12... So it might be some sort of positional
>>mis-judgement you are fighting, rather than not seeing some tactic that looks
>>simple...
>
>Bob,
>
>My program has an eval similar to yours, but that's not what I was asking. Mine
>initially thinks black will lose the pawn (eval < 1) after Rxd1 Rxd1 Rd8 Rxd8,
>because it doesn't see the bishop going down to c1 to recover it. IOW, it
>doesn't want to play the initial RxR, which causes major problems.
>
>So it's a tactical problem for my program.
>
>Will
There is more going on... here is my output (PII/300 notebook) starting at
ply=1. I forced it to only search Rxd1 to see if the score changes at any
point significantly:
1 0.00 -0.03 1. ... Rxd1+ 2. Rxd1 Kf8
1-> 0.01 -0.03 1. ... Rxd1+ 2. Rxd1 Kf8
2 0.02 -0.07 1. ... Rxd1+ 2. Rxd1 Rd8 3. Nd6
2-> 0.03 -0.07 1. ... Rxd1+ 2. Rxd1 Rd8 3. Nd6
3 0.04 -0.15 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 b6
3-> 0.06 -0.15 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 b6
4 0.08 -0.21 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 b6 5. Bd5
4-> 0.30 -0.21 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 b6 5. Bd5
5 0.52 -- 1. ... Rxd1+
5 0.67 -0.88 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 b6 5. Nc8 Kf8 6. Nxa7
5 0.88 -0.49 1. ... a6 2. Rxd8+ Rxd8 3. b4 Ra8 4.
b5 Nd8 5. bxa6 Rxa6 6. Rxa6 bxa6
5-> 1.14 -0.49 1. ... a6 2. Rxd8+ Rxd8 3. b4 Ra8 4.
b5 Nd8 5. bxa6 Rxa6 6. Rxa6 bxa6
6 1.51 -0.71 1. ... a6 2. Rxd8+ Rxd8 3. Bxc6 bxc6
4. f3 c5 5. Rxa6 Kf8
6 2.45 -0.49 1. ... Rac8 2. Nd6 Rc7 3. Nxb7 Rxb7
4. Bxc6 Rxd1+ 5. Rxd1 Rxb3
6 3.52 -0.48 1. ... Rab8 2. Nd6 a6 3. Nxb7 Rxb7
4. Bxc6 Rxd1+ 5. Rxd1 Rxb3
6-> 3.65 -0.48 1. ... Rab8 2. Nd6 a6 3. Nxb7 Rxb7
4. Bxc6 Rxd1+ 5. Rxd1 Rxb3
7 4.75 -0.69 1. ... Rab8 2. Re1 Rdc8 3. Bxc6 bxc6
4. Rxa7 Rxb3 5. Nd6
7 5.57 -0.67 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 Bg5 5. Bxb7 Bc1 6. Bd5 Bxb2
7-> 6.59 -0.67 1. ... Rxd1+ 2. Rxd1 Rd8 3. Rxd8+ Nxd8
4. Nd6 Bg5 5. Bxb7 Bc1 6. Bd5 Bxb2
IE even at depth=1 my eval is about the same as it is at depth=7, and on to
depth=12...
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.