Author: Christophe Theron
Date: 11:52:20 11/18/99
Go up one level in this thread
On November 18, 1999 at 03:18:19, blass uri wrote:
>On November 17, 1999 at 20:38:38, Thorsten Czub wrote:
>
>>On November 17, 1999 at 19:01:57, Christophe Theron wrote:
>>
>>>No Chris, you don't get it. If you prune too much on material criterias you
>>>completely destroy the positional skills of the program.
>>
>>Chris, ahem, or Bella , or whoever you are, and Christophe!
>>
>>Hi !
>>
>>
>>I don't think tiger plays fast-chess.
>>it plays very nice chess. positional alike.
>>
>>
>>>If I wanted to go deeper tactically, I could do it easily. I think I could go 2
>>>plies deeper. But in this case my program would be completely crushed because of
>>>incredible positional holes.
>>>
>>>Before the game goes on a tactical field, my program would have a totally lost
>>>position.
>>
>>>Each time I have tried to sacrifice the positional understanding to get deeper,
>>>it was a disaster. I have found that it works in the opposite direction: with
>>>better positional understanding the program goes deeper. Because it sees the
>>>right moves earlier, and spends less time analyzing sonense moves.
>>
>>Right. The more the program knows, the more it sees in the search.
>>Tiger has completely different evaluation than any other program
>>in most cases.
>>
>>If you want i can present lots of games and positions where tiger evaluates
>>DIFFERENT. often 2 or 3 pawns different than hiarcs/fritz/others.
>>
>>Why ?
>>Because it knows and comes also deeper.
>>And it seems it evaluates a special thing much higher than any other program.
>>i am sure i know what it is. but i will not say it. a secret so far.
>>
>>
>>
>>>You make it sound as if I had found a very specific way to shoot on a very
>>>specific weakness of a very specific subset of chess programs.
>>>
>>>You make it sound as if I was not programming chess.
>>
>>>You completely overlook the fact that several testers including our friend
>>>Thorsten have noticed that Chess Tiger has a large amplitude in its evaluation
>>>function. It's usual to see Tiger at +1.90 when the opponent is still close to
>>>0.00.
>>
>>Right. As i said :-))) it has a big evaluation function and very often has
>>completely different evaluation trend than the opponent.
>>and it wins. that makes it funny.
>>i like it because it is different. as chess system tal is.
>
>I do not think that the word bigger is right.
>It probably has a better evaluation function.
>
>The quality of the evaluation is relevant and not the size of the evaluation
>function.
>
>Uri
Quality and quantity in this case are related. If you have a lousy evaluation
function and do not trust it, you make it "small", that is for example always
returning values in the interval [-0.75;+0.75].
We all know many chess programs evaluating their position at -0.75 when they
have a dead lost position but material is still equal.
These chess program would happily accept the lost position if they can capture a
pawn in exchange.
With a "big" positional evaluation function, a program would never take the pawn
and save its position.
So if you have a good positional evaluation function f(P) (that does not take
material into account), you can use f(P)/10 in your score or use f(P)*2.
Both versions will play exactly the same moves in many positions. The difference
will be seen only when there is a choice like giving a pawn against a very good
position. In this case the "big" evaluation will coureagously accept the deal,
when the "small" evaluation version will stick to material balance only and play
boring.
You see what I mean. With a given quality positional evaluation, it's possible
to play on the scale of this positional evaluation and the scale makes a
difference.
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.