Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The King 3.33 Milan 1.7a - my final settings

Author: Andrew Walker

Date: 14:50:03 02/08/05

Go up one level in this thread


On February 08, 2005 at 02:10:28, Joshua Haglund wrote:

>On February 06, 2005 at 19:39:14, Andrew Walker wrote:
>
>>On February 05, 2005 at 03:29:03, Joshua Haglund wrote:
>>
>>>On February 04, 2005 at 22:52:42, Andrew Walker wrote:
>>>
>>>>On February 03, 2005 at 23:57:38, Joshua Haglund wrote:
>>>>
>>>>>On February 03, 2005 at 22:18:36, Andrew Walker wrote:
>>>>>
>>>>>>On February 03, 2005 at 22:10:15, Andrew Walker wrote:
>>>>>>
>>>>>>>On February 03, 2005 at 21:16:06, Jason Kent wrote:
>>>>>>>
>>>>>>>>>@ Playchess.com:
>>>>>>>>>
>>>>>>>>>Play 2649 more games with The King 3.23 SKR.
>>>>>>>>>Play 5000 games with The King 3.33 Milan 1.7a.
>>>>>>>>>Play 2483 more games with The King 3.33 R2
>>>>>>>>>
>>>>>>>>>Then these will be added to my list and go against:
>>>>>>>>>
>>>>>>>>>Engine            Elo Range Avg.   ELO RANGE
>>>>>>>>>The King 3.23 R1     2547          2544-2550
>>>>>>>>>The King 3.33        2535          2520-2550
>>>>>>>>>The King 3.23        2468          2459-2477
>>>>>>>>>
>>>>>>>>>Joshua Haglund
>>>>>>>>>toneewa@yahoo.com
>>>>>>>>
>>>>>>>>How do you know what setting is being used when it is played online?  I usually
>>>>>>>>call my 3.23 settings CM9 "name", and 3.33 settings CMX "personality name".  Is
>>>>>>>>the server able to tell exactly what is running?
>>>>>>>>
>>>>>>>>Jason
>>>>>>>
>>>>>>>Note those numbers don't take into account:
>>>>>>>- Computer speed (and if any other tasks are running)
>>>>>>>- Hash size
>>>>>>>- Learning accrued
>>>>>>>- The opening book used
>>>>>>>- If engine parameters have been changed
>>>>>>>- How many if any tablebases are in use
>>>>>>>Make your own judgement!
>>>>>>
>>>>>>Also, older engines will have a far greater proportion of games played on
>>>>>>slower machines. So his numbers are just hot air.
>>>>>
>>>>>That is why there is a ELO RANGE. Not just one rating on one machine, with
>>>>>certain settings, etc.
>>>>
>>>>Still doesn't make sense. I just saw one of your posts claiming +- 20 points
>>>>for Shredder8. The number of machines playing with this engine are going to
>>>>show a far greater range of ELO strength. You have not shown that you can use
>>>>this to compare the strength of two engines, as you are measuring different
>>>>quantities for each one.
>>>
>>>It's very simple. Shredder 8 has +/- 20. It is an ELO Average of all the users
>>>with Shredder 8 with a ELO RANGE +/-. The reason it doesn't have a such a wide
>>>range is because it's consistant between the ELO RANGE.
>>>
>>Does this mean you are taking the uncertainty for each individuals rating and
>>averaging those? Or something else? It would be clearer if you gave an example
>>from just a small number of figures.
>
>All right. I posted something similar to this before.
>
>If you have two people using the same engine, example: Shredder 9. Person 1 gets
>ELO 2575. Person 2 gets an ELO 2675. The ELO RANGE Average would be 2625 +/- 50.
>The ELO RANGE would be 2575-2675. This is only with 2 people. On playchess.com
>there are many more.
>
>(Example)
>Engine            Elo Range Avg.         ELO RANGE
>Shredder 9            2625      +/- 50   2575-2675
>Shredder 8            2605      +/- 20   2585-2625
>

Thanks Joshua, so ELO RANGE= min score to max score
                  +- figure= (max-min)/2

To improve your list I would suggest adding the number of ratings used for a
specific engine, to help the interpretation of these. Is it possible to get any
further data from the server? It would be nice for instance to know the range
and average of the CPU speeds used (duals etc may complicate this).

Andrew



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.