Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Good Test Position, the "ply 16 rook sac"

Author: Uri Blass

Date: 12:57:26 09/21/00

Go up one level in this thread


On September 21, 2000 at 14:36:34, John Merlino wrote:

>On September 21, 2000 at 00:48:00, John Merlino wrote:
>
>>On September 20, 2000 at 14:31:25, Uri Blass wrote:
>>
>>>On September 20, 2000 at 14:12:25, John Merlino wrote:
>>>
>>>>On September 20, 2000 at 13:18:20, James wrote:
>>>>
>>>>>This position turns out to be an excellent comparitor ofd chessmaster versions.
>>>>>My puter is 733 Piii, not overclocked with 512 meg of ram. Tested cm55 and cm6k
>>>>>using cm default but with select 10, and ttables set to 30. results:
>>>>>cm55 found mate in 14 at 33 min 45 seconds < about 20 sec into ply 13> and at 20
>>>>>seconds into ply 13, cm6k still found m n 16 at 58 minutes and 45 seconds.
>>>>>  My play formula on cm55 found it in 31 min and 30 seconds. Am sure other
>>>>>programs could see it faster, if anyone does any comparitive tests on their
>>>>>system, would be interested in the results.  IF cm8000 is as efficient as cm55,
>>>>>and has 12 ply selectivity, it could maybe see it in 10 minutes or so i think.
>>>>>  This position seems to be another that begs the question : what is a ply : ?
>>>>
>>>>What is the position? I can test it on a Beta build of CM8000....
>>>>
>>>>jm
>>>
>>>[D]2r3kr/5pp1/7p/3PQRb1/ppq1B3/8/PPP3PP/1K1R4 w
>>>
>>>chessmaster6000(ss=10) found mate in 20 at depth 2/12,mate in 14 at depth 5/15
>>>and mate in 13 at depth 6/16
>>>
>>>Uri
>>
>>As promised....
>>
>>A beta build of Chessmaster 8000, running on my PIII-600, with only 128 MB RAM,
>>and only a 1MB hash table, BUT with SS=12, announced mate in 23 after 39 minutes
>>and 57 seconds. The announcement came at depth 13, after 258,350,432 positions
>>had been seen (just under 108,000 positions/second).
>>
>>Admittedly, this test is not terribly useful if you are trying to compare
>>results with YOUR testing, because of machine and some settings differences (and
>>also because I had to actually WORK and couldn't leave the machine idle for some
>>of the time).
>>
>>I'll try running another test overnight, with a 32MB hash table (the most I can
>>afford on my machine).
>>
>>However, there you are....
>>
>>jm
>
>The first thing I should mention is that my previous post was incorrect. CM8000
>did not announce mate in 23. It was, in fact, mate in 13. See the following
>explanation for why I made this error....
>
>So, I let it run overnight, with SS=12 and a 32MB hash table. CM8000 announced
>mate in 14 at depth 13 in 24 minutes and 49 seconds, at about 162 million
>positions.
>Depth 14 was reached at 35:46, and mate in 14 was still the eval, but the PV had
>changed (too long to bother with -- and the point was announcing mate).
>Depth 15, reached after 2:24:49, announced mate in 13.
>Depth 16, still showing mate in 13, was reached in 5:08:53 (2.1 billion
>positions, approximately).
>As of this writing, after about 13 1/2 hours, depth 17 has not been reached.
>
>Oddly enough, the EVAL of the PV at depth 13 was 99.73, which, in OLD versions
>of Chessmaster (7000 and previous) would mean mate in 27. This vexed me greatly,
>and begged the question why previous versions of The King announced a mating
>line with fewer moves in the above position. Thanks to the "Visual Thinking"
>Window, I have determined that the old Chessmaster method of scoring:
>
>(1 - eval) * 100 = mate in N

In chessmaster6000 it should be:
(100-eval)*100=N when the score is mate in N

example:(100-99.90)*100=10 implies that 99.90 is mate in 10
>
>no longer works. In CM8000:
>
>(1 - eval) * 100 = number of plies to mate

again I guess that you mean (100-eval)*100
>
>so the new math to determine mate in N would be
>
>(((1 - eval) * 100) + 1) / 2 = mate in N.

again 100-eval and not 1-eval.

Uri



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.