Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: ::sigh:: wac 17 problem

Author: Michael Henderson

Date: 13:45:46 10/16/04

Go up one level in this thread


On October 16, 2004 at 16:13:18, Rick Bischoff wrote:

>On October 16, 2004 at 13:34:23, Michael Henderson wrote:
>
>>Hmmm...your numbers do seem pretty large.  I can do 9 plies in < few hundred K
>>nodes.  Hopefully no bugs in extensions/depth parameters?  It seems like there
>>are some unwanted extra-depth searches.
>>
>>Michael
>
>In this output, I turned off every extension (including check) and disabled
>iterative deepening:
> 5      55      16      18853 1. Ne5 Rf6 2. Bg5 Qxb3 3. axb3
> 6     157      99     130437 1. Ne5 Qxb3 2. Nxd7+ Kc8 3. axb3 Kxd7 4. Rxa7
> 7     157     328     363396 1. Ne5 Qxb3 2. Nxd7+ Kc8 3. axb3 Kxd7 4. Rxa7
> 8     153    1284    1882658 1. Ne5 Qxb3 2. Nxd7+ Kc8 3. axb3 Kxd7 4. Rxa7 Nd5
> 9     153    6368    6609003 1. Ne5 Qxb3 2. Nxd7+ Kc8 3. axb3 Kxd7 4. Rxa7 Rb6
>5. c4
> 9     154   14384   17170667 1. b6 axb6 2. Ne5 Kc8 3. Nxg6 hxg6 4. f3 Nd5 5.
>Bg5
>10     138   40361   55631916 1. b6 axb6 2. Ne5 Kc8 3. Nxg6 hxg6 4. Qxe6 Bxe6 5.
>f3 Bd5
>
>So you can see something else is happening.. In this output, I disabled qsearch
>and just a static eval:
>
> 5     311      56      11935 1. Ne5 Kc8 2. Nxg6 Qxg6 3. Rxe4
> 6     -38     109      25782 1. Ne5 e3 2. Qxe6 exf2+ 3. Kxf2 Rxe6
> 7     288     329     128748 1. Ne5 Qxb3 2. axb3 Bxb5 3. Nxg6 Nxg6 4. Rxa7
> 8      36    2117     510910 1. Ne5 Qxb3 2. axb3 Bxb5 3. Nxg6 hxg6 4. h3 Rxh3
> 9     284    5035    2409281 1. Ne5 Qxb3 2. axb3 Bxb5 3. Nxg6 Nxg6 4. c4 Bc6 5.
>Rxa7
> 9     287   12681    4452561 1. b6 axb6 2. Ne5 Rd8 3. Ba3 Nc6 4. Nxg6 Qxg6
>
>I stopped it here, since there wasn't the dramatic improvement.  Could my poor
>evaluation function account for this type of explosion?

Evaluation shouldn't be a problem since your program can see the material gain.
What are your move ordering stats?

Michael



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.