Computer Chess Club Archives




Subject: Re: Branching factor, make me confuse more that ever.

Author: leonid

Date: 17:42:13 04/04/00

Go up one level in this thread

On April 04, 2000 at 19:29:18, Tom Kerrigan wrote:

>On April 04, 2000 at 17:44:03, leonid wrote:
>>On April 04, 2000 at 11:53:29, Tom Kerrigan wrote:
>>>On April 04, 2000 at 06:36:22, leonid wrote:
>>>>Simple things are not that simple that somebody would like to see. Recently
>>>>(trying to find how I must fix my branching factor) I put two counter in my game
>>>>(chess logic). This numbers are:
>>>>1) Number of all nodes that logic (chess logic) see inside of given position.
>>>>2) All the legal nodes (moves) that existe in all plies.
>>>>Deviding those two numbers I recieved 7%.
>>>>The same devision find already many months ago, and that worried me since, for
>>>>the plies starting with ply 6 and up was around 21%. Problem is that best games
>>>>represent proportion that is around 15%. I expect that I probably loose speed
>>>>between ply 2 and 10 in some 1000%.
>>>Possibly because you don't have a quiecence search, or extensions, or do
>>>iterative deepening.
>>Could be, even if I doubt so. Reason for this is that all those quiecence search
>>and extensions sound to me as the part of "partial search". If it is really so
>>all those new factors will only make all comparison more obscure that ever. This
>>is the primary reason why I tried to compare my logic on the ground on "brute
>>force" "fixed depth" and no extensions to lead me astray.
>Of course there will be differences between two chess programs. That's what
>makes comparisons interesting. If you compare two of the same thing, the outcome
>obviously won't be very interesting.
>People compare chess programs every day, and the programs all have some sort of
>selective ("partial") search. Sort of like racing cars. It's interesting to race
>a Mazda against a Honda to see how well a rotary engine stacks up. But the Mazda
>is still essentially the same as the Honda.
>Again, what you're trying to do is like racing a sailboat against a car. It's
>silly because they're totally different.

Tom, there are few very basic things that can say you the difference in
expectation of which racing car could be the best. That car that have more horse
power (raw power) in its engin. Later you can go about dynamics and all other
billion details that can obscure or make even more prominent this basic
advantage. In chess program it is the same.

When in my "mate solving logic" I found that my "quick logic" give me wrong
message, and my raw speed is not what it must be, I stayed still for two years
with it. I left working on mate solver only when I found that I came to the
place where I wanted to be. My intention is to do the same this time. Only now
it is not that simple. I don't even know where I am right now.


This page took 0.01 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.