Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: next deep blue

Author: Robert Hyatt

Date: 11:05:12 01/26/00

Go up one level in this thread


On January 26, 2000 at 13:03:09, Christophe Theron wrote:

>On January 26, 2000 at 10:27:38, Robert Hyatt wrote:
>
>>On January 26, 2000 at 01:38:07, Christophe Theron wrote:
>>
>>>
>>>
>>>I don't forget this.
>>>
>>>I'm just saying that the thing could have been much stronger if they had worked
>>>on implementing a good pruning scheme rather than this SE thing.
>>>
>>>
>>
>>
>>I definitely don't agree here.  I watched them play *socrates at an acm event,
>>where *socrates was searching considerably faster than they were in terms of
>>NPS, and in terms of depth of search reached.  They blew *socrates out on a
>>tactical level anyway...
>
>
>But you are only talking about ONE game, or maybe only a FEW games that you have
>witnessed yourself.
>
>You do not know how good or bad *Socrates pruning scheme was.
>
>I would not conclude from this that no pruning is better than pruning.
>
>On the other hand, years of study and practical experience have shown that using
>a good pruning scheme is way better than not using one.
>
>


Be careful.  You can state something in two ways, where they sound the same
but are not:

1.  their results show that their pruning is good.

2.  their results do not show that their pruning (or lack thereof) is bad.

the two statements are significantly different.  I can only conclude that
what they did is not bad...  not that it is good or "best"...




>
>
>
>>>Maybe or maybe not.
>>>
>>>My interest was rather in the fact that they used a probably suboptimal approach
>>>in a multimillion dollars project.
>>>
>>>Not sure anymore if you are interested in talking about this...
>>>
>>>
>>
>>
>>I am always interested.  I had said before that I would like to see how I
>>could play on hardware that would drive my search to a depth of 19 plies.
>
>
>You seem so keen on this idea that a pruning system is not useful at deeper
>levels that you should do the experiment yourself.

I'm not 'keen' on it at all.  Notice that I don't claim this is true.  I said
it "might" be true.




>
>Let Crafty without null move play games against the normal Crafty. Give the
>non-pruning one enough time to simulate a Crafty running at DB's speed.
>
>Your null-move version will beat the non pruning one, even with a huge time
>handicap.
>
>If it doesn't, then you have a point and this would be a new and interesting
>thing. You could write a paper on this, and we would better understand DB.
>
>You have done experiments on the "dimishing returns" idea. This is related to
>this idea, isn't it worth experimenting?
>
>


yes it is, except for the simulating DB part.  It would take a year to play
one game.  :)




>
>
>>But I am a long way from being ready to say "they probably used a suboptimal
>>approach in a multimillion dollars project" since I don't have the hardware to
>>play with.  But what if I had a null-move and non-null-move crafty, and on the
>>new hardware, the non-null-move program made fewer mistakes overall?  I'd go
>>with that in an instant.  Where on slower hardware the null-move depth gained
>>might be very worth-while (it obviously is as I use it today)...
>
>
>Do the experiment... It just requires that you let a PC run for enough time
>(maybe a month or two).
>
>
>
>
>>I simply can't say their way was right or wrong without testing a known program
>>to see.  If all I have to go on is their results, I have to say they did well.
>
>
>If you want to live on facts on this issue, you can easily do it. I'm sure you
>can find a computer for this experiment.
>
>
>
>    Christophe



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.