Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Pondering ("think on opponent's time")

Author: Sune Fischer

Date: 13:17:09 11/11/02

Go up one level in this thread


On November 11, 2002 at 15:59:40, Sune Fischer wrote:
>>>
>>>The math here is wrong - it should be:
>>>0.6*0.9 + 0.4*0.1*0.3 = 0.54 + 0.012 = 0.0552
>>
>>The math is correct
>>You cannot change my calculation and decide that my math is wrong.
>>
>>I think that you assume that I assume that the program always use 90% of the
>>time for the move that it expects.
>>
>>I did not say it.
>>
>>There are 2 kind of moves:
>>
>>Moves type A:Moves that the program has the move that is going to be played in
>>the pv.
>>Moves type B:Moves that the program does not have the move in the pv.


Oh, I get it now, it's the utilization of the CPU of course.
In that case you need another percentage, like this:

>Moves type A:Moves that the program has the move that is going to be played in
>the pv and are correctly pridicted x percent of the time.
>Moves type B:Moves that the program does not have the move in the pv. (I don't know)

If we assume x is 50 like Bob's example, the first part of you equation would be
A% times 50% times 90% CPU, looks like that will be hard pressed to compete
against Bob's 50% times 100% CPU. Of course the assumption is that A=100, so you
only have 10% CPU to burn on the remaining 50% of the moves. You need 50%
efficiency out of those 10% to compete with Bob, and you have only moves that
failed low to search, I'd almost say that is impossible.

-S.

>>I assume that moves type A are 60% of the moves.
>>
>>I said that it is possible that the program use in average 90% of the time for
>>the played move in moves type A and 30% of the time for the played move in moves
>>type B.
>>
>>The point is that moves type A are not random moves and most of them are forced
>>positionally or tactically so the program can see after a short search that one
>>move is better than the other moves so it is using 90% of the time for the
>>expected move.
>>
>>The moves type B are not forced moves and in most cases there are 2 or 3 moves
>>that have the same score or almost the same score so the program can use 30% of
>>it's time for every one of these moves.
>>
>>Uri
>
>I think you are both wrong :)
>But I'm not sure you are trying to calculate here, shouldn't it be the expected
>ponder success rate?
>Looks like you are calculating 0.6*0.9+0.4*0.3=0.66>0.6, or:
> frequency of type A times percentage of search time + ....
>
>Whatever that is, it isn't the same.
>
>Bob's example was that in more than 50% of the cases the second move in the PV
>would be a correct guess, hence a ponder success rate of more than 50% (or
>0.50).
>This was only type A moves, ie we assume there is always a move to ponder.
>
>We could get the full success rate if we apply a ponder scheme to type B moves
>also, but I think the question was only about the optimal strategy if there is a
>move in the PV, otherwise it isn't so clear what to do.
>
>-S.



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.