Author: Albert Silver
Date: 14:18:47 08/17/04
Go up one level in this thread
On August 17, 2004 at 16:41:25, Ed Schröder wrote:
>On August 17, 2004 at 15:33:58, Joerg Oster wrote:
>
>>Hi Ed,
>>
>>would you be so kind to write a few words about the below listed parameters?
>>('Adjusted Selectivity' and 'Minimum Selectivity')
>>I've changed the 'Adjusted Selectivity' parameters, but couldn't see any changes
>>in the calculation process of Pro Deo 1.0.
>>My aim is to make Pro Deo reach higher depths in less time.
>>
>>Kind regards,
>>Joerg.
>
>>[Adjusted Selectivity MIDG ? 6] * Maximum selective depth (midgame)
>>[Adjusted Selectivity END0 ? 5] * Maximum selective depth (early endgame)
>>[Adjusted Selectivity END1 ? 4] * Maximum selective depth (endgame)
>>[Adjusted Selectivity END2 ? 3] * Maximum selective depth (simple endgame)
>>
>>[Minimum Selectivity MIDG ? 4] * Minimum brute force depth (midgame)
>>[Minimum Selectivity END0 ? 4] * Minimum brute force depth (early endgame)
>>[Minimum Selectivity END1 ? 4] * Minimum brute force depth (endgame)
>>[Minimum Selectivity END2 ? 4] * Minimum brute force depth (simple endgame)
>
>The goal of the parameters is to tune Pro Deo's selective search which is a
>mixture between nullnove and my own invented way of doing selective search. It
>would go much too far to start explaining what these parameters actually do. For
>that I refer to my Programmer Stuff page as listed on:
>http://members.home.nl/matador/chess840.htm#SELECTIVE%20SEARCH
>
>Regarding your second question, if you want a deeper search in less time
>consider the following settings:
>
>[Adjusted Selectivity MIDG = 8] * Maximum selective depth (midgame)
>[Adjusted Selectivity END0 = 7] * Maximum selective depth (early endgame)
>[Adjusted Selectivity END1 = 6] * Maximum selective depth (endgame)
>[Adjusted Selectivity END2 = 5] * Maximum selective depth (simple endgame)
>
>Don't expect these settings to give you an overall improvement, it's less. But
>the engine will fly through the plies.
>
>My best,
>
>Ed
Facsinating. I remember when you announced the development of those pages, but
hadn't been back in quite some time. Great job! I have a couple of questions:
First there's the note:
"It's very easy to make REBEL a full Null-Move program, all it needs to do is to
make S_DEPTH equal to iteration_depth, actually REBEL has an Interface driven
parameter for that: [Selective Search = 0] in the Personality Editor."
Presumably you tested the full Null-move and it yielded less than optimal
results. Do you remember what they were?
I must admit I find these two parameters slightly confusing also after your
explanation:
[Selective Search = 100]
[Search Technique = NULLMOVE]
If I reduce the Selective Search, it will presumably make it closer to a full
Null-move program. What happens if I increase it? What sort of effect can one
expect from increasing or decreasing this?
Also, the default Rebel setting has *none* of the parameters you just discussed,
regarding maximum and minimum selectivity. Are these still in testing? If one
were to enable any of them, should one adjust the Selective Search, or is this
something completely different?
Albert
Albert
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.