Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: the easy testset

Author: Vincent Diepeveen

Date: 04:44:33 10/18/01

Go up one level in this thread


On October 17, 2001 at 19:24:55, Miguel A. Ballicora wrote:

for me non-easy positions solution times are measured in minutes.

Grasz2002 for example the level most likely will be 3 minutes a move.

So i don't care whether some positions are found in 20 seconds or 30
seconds or 2 seconds or 5 seconds.

I do not see that difference.

All i see whether you make it (at new hardware) within 10 minutes,
as first moves out of book diep uses around that time a move,
and with filled hashtables 10 minutes is a good time to test at.

Of course you don't show up there with a laptop from 10 years ago!

>On October 17, 2001 at 17:47:04, Vincent Diepeveen wrote:
>
>>On October 17, 2001 at 10:47:19, Miguel A. Ballicora wrote:
>>
>>>On October 17, 2001 at 04:50:23, Vincent Diepeveen wrote:
>>>
>>>>On October 17, 2001 at 00:02:15, Miguel A. Ballicora wrote:
>>>>
>>>
>>>>You took away the hard positions and have left the easy positions
>>>>in the testset?
>>>
>>>No, I took away the incorrect ones. There are only 4 positions that could
>>>be more difficult but I have not tested the correctness yet. Those cannot
>>>be solved by YACE or Crafty in 1/2 hour in a K2-400 mhz.
>>>
>>>Obviously, these positions are not a challenge for commercial engines
>>>in the fastest hardware; but they are not meant to be. In fact, it is not easy
>>>to find positions where the solution is _purely_ combinatorial and it is
>>>challenge for top programs in top hardware. These positions are certainly useful
>>>for amateur programs.
>>>My idea is to get as many clean combinatorial positions as possible and later
>>>I might categorize them. For instance, this test suite is not useful for you
>>>but it is for me.
>>>Anyway, the idea was to have solutions where the time could be measured. i.e,
>>>more than one second but apparently they are easier than that.
>>>
>>>Can you please post the solution time of the ones that require more than a
>>>second?
>>
>>I don't care for solution times. It's plydepths i care for.
>
>But I do care, that is why I asked you.
>Of course, if you do not want to post them it is fine.
>
>>It's all 5 to 8 ply majority of positions
>>
>>You claimed that it was not an easy testset, but positions that solve
>>at 5 to 8 ply are pretty easy.
>
>I do not claim anything, I am sharing the positions that are correct that will
>be eventually useful for me.
>Easy is a very relative term. What I tried to do is to eliminate the positions
>that are "too" easy. Those are the positions that are solved in microseconds, as
>in WAC, when you cannot even get the time measured accurately.
>
>Of course, if you have a top engine in top hardware some positions might fall
>under a second.
>
>Miguel
>
>
>
>
>
>>
>>I got also 8 ply at world champs 1997.
>>
>>It's 4 years later now.
>>
>>>Regards,
>>>Miguel



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.