Author: Christophe Theron
Date: 03:03:58 01/04/98
Go up one level in this thread
On January 03, 1998 at 13:13:11, Thorsten Czub wrote:
>Lets compromise.
>The search depth GENIUS shows is NOT the full-width. It is
>full-width + extensions and what is the other search-depth-indicator ?
>
>If Genius shows e.g. 6/18
>6 cannot be full-width. Otherwise it would have seen Rxe6 because the
>plot of this position is NOT that deep.
>So - the first number is NOT full-width. it is full-width and maybe the
>extensions are asymmetric.
>I can compromise to this.
Well, I suppose that 6/18 means that the first 6 plies are searched
full-width, or near-full-width. In fact maybe some optimizations are
done in the "full width" plies, like null-move, but it normally sees
everything a real full width search would see. Or almost everything,
99%.
But maybe I'm wrong. If you want to prove that the "first" plies are
searched asymetrically, I suggest that you turn off the selective part
(selective search=0), to simplify everything. I think a lot of Genius
magic is in the selective part, so it would be interesting to see what
is left when you remove this part.
Could you try to see if Rxe6 is found or not, and at which depth? Maybe
something can get clear with experiments like this.
>Maximal peak of extensions ?
>I don't think so.
>What is S18 ??
>6+12 = 18 ??
>
>
>I still believe the old dedicated-machine search depth indicator's were
>always misunderstood.
Unfortunately I cannot say anything about this, because I only have
Genius for PC. But maybe you can see if MxSy (on the dedicated machines)
is the same as x/y on Genius. Just give the same position, stop the
search as soon as the key move is found, and compare the lines.
From what I see when I use Genius, I think the first number is
"full-width depth", and the second is "selective depth (peak)".
Sometimes Genius spends too much time in the selective part, and cannot
do enough full-width depth. When this happens, Tiger often finds a line
Genius didn't see, and may win something.
If the selective depth can be modified on the dedicated machines, maybe
you can learn something by setting it to zero and look at what happens.
>What if M was not the search-depth but the number of odd-plies
>(2,4,6,8,9) and the second number , the S-term is maybe not what we
>guess it is.
>I think Richars search (or richards extensions way) is so different that
>anybody THOUGHT these indicators are the same like in HIS OWN normal
>programs.But it wasn't.
>His search indicators (of the old dedicated machines) made only a sense
>when
>you have a different explanation from my point of view.
>As I said, the M stuff in the dedicated machines was very very low.
>Too low !!
>
>Often you saw e.g. Roma 68000 only in M1 and M2 was often the maximum in
>a 40/120 game. Sometimes you saw M3 !!
>
>But not often.
I suppose you are right. Maybe M means "Move" and not "ply". When you
see M1, the machine has computed its move and your reply "full width",
and *some* other moves after in a selective way.
So the ply number you see in Genius is twice the "M" number. M2 means 4
plies + selective part?
>But I am sure, when I would show him my old dedicated Lang Roma 68000
>machine and we would let it compute 1 second each move and we would get
>the log-file (we would have do write it down manually) he would see that
>there are some secrets with this 64-KB program.
What do you call the "log file"?
I agree about the fact that there are "secrets". Maybe just simple ideas
used in a smart way...
>It is not coming deep enough (with the M-part) and has big senseful
>main-lines and deep S-parts.
The S part is also what I would like to understand.
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.