Author: Graham Laight
Date: 03:23:04 05/12/00
Go up one level in this thread
On May 12, 2000 at 05:36:40, Dann Corbit wrote:
>On May 12, 2000 at 05:22:45, Jouni Uski wrote:
>
>>I noticed, that in this position from correspondence challenge:
>>[D]r6k/7p/5pp1/7n/2bQN3/5q2/P1P4R/K2R4 w - - 0 0
>>
>>LG2000 reach 1113knps in AMD450Mhz and this with 8 MB hash. This means
>>2.47 kn/Mhz. Is this record?
>
>That's clearly the record for an AMD450MHz with 8 MB hash searching that
>position.
>
>That (and 50 cents) will get you a cup of coffee.
>
>Don't get me wrong -- LG is one of the top 5 freely available programs without a
>doubt. But if there was ever a more useless statistic than thousand nodes per
>megaherts, I would like to know what it is
I think it is interesting in terms of categorising programs as either brute
force or knowledge based (though just because you're slow, that doesn't
necessarily mean that you're clever).
>MCP8, Diep, CS-Tal and others will get several orders of magnitude less nodes
>than that. But they are equal or better at playing chess.
>
>Go figure.
What Dann seems to be saying is that a program with a high NPS per megahertz
could have a "search knowledge profile" something like this:
ply |-------------------------------------------------------------|
| $$$ # |
25 | $$$ # |
| # ## |
20 | ## ####### |
| ######## ############# |
15 | ############## #################### |
| ###################### ########################### |
10 |#############################################################|
|#############################################################|
5 |#############################################################|
|#############################################################|
|-------------------------------------------------------------|
Breadth of knowledge
Where the # represents good knowledge, and the $ represents areas where the
knowledge would be most valuable in this particular position.
However, in the same position, slower searching (but more knowledge
rich)programs that play equally well might have a profile more like this one:
ply |-------------------------------------------------------------|
| ~ ~ ~ $$~ ~ ~ ~ ~ ~ ~ |
25 | ~ ~ ~ ~ ~$$ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~|
|~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~|
20 | ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ |
|~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~|
15 | ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ ~~ |
|~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~~~ ~|
10 |~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
|~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
5 |#############################################################|
|#############################################################|
|-------------------------------------------------------------|
Breadth of knowledge
Where the ~ represents weaker knowledge. This shows a program's knowledge giving
it some degree of useful data about future positions by taking into account
positional factors which will only be of direct value later on.
In this case, the "high quality" knowledge represented by the # symbols doesn't
go as deep, but the extra knowledge gives a "thinner" covering to a greater
depth. Because the 2nd program hits the sweet spot where the $ symbols are in
this case, there is a chance that it might make a better move in this case.
-g
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.