Author: Christophe Theron
Date: 17:47:22 11/08/00
Go up one level in this thread
On November 08, 2000 at 09:28:14, Joe Besogn wrote:
>On November 08, 2000 at 07:39:24, Thorsten Czub wrote:
>
>>On November 08, 2000 at 06:28:54, Joe Besogn wrote:
>>
>>>>Just replace amerindian's bows and arrows with nuclear bombs, leave them alone
>>>>for a while, then come back and ask them what they think about the efficiency of
>>>>this new tool.
>>>>
>>>>Oops... There is nobody left to answer. They had the choice to use the bombs to
>>>>hunt (and nuke themselves), or to die from starvation.
>>>>
>>>>Just a remark about the different ways to view things. :)
>>
>>if Gorbatschow would not have changed paradigms,
>>(the americans would never have done this ! they still live in
>>a paradigm where Russians are their main danger and communism
>>is the devil ! - for americans the paradigm changes have not been happened,
>>but here in europe, the paradigm change in politics has been done succesfully.)
>>we would still live in a cold war, with 2 berlins, 2 germanys and
>>the rockets on the left and on the right side.
>>in the moment the americans give up to boycott cuba, they show that
>>they got an idea about paradigm change. so far - no progress.
>>
>>
>>>Quite so. To draw on Kuhn's ideas ....
>>
>>>Fernando was doing 'normal science' - and was arguing on the more-is-better,
>>>bigger-is-better basis (nukes more effective than bows and arrows).
>>
>>right.
>>
>>>Christophe feels the revolutionary space is growing, so he fills it a little
>>>more.
>>
>>:-)
>>
>>>Just drawing parallels.
>>
>>IMO many people don't understand that gambit-tiger is different
>>than other programs, like cstal was different in its time.
>
>Yes. They are in a trap set by the old paradigm. Results trap. They only notice
>a thing if it has more ELO points than another thing.
>
>You were different, you looked closely at the thing itself.
>
>They looked at results-of-the-thing, you looked at the thing. Plus you thought
>in another framework in any case.
>
>So you saw very fast, they had to wait five years.
>
>>
>>it trusts its evaluations, and the very efficient search gambit-tiger
>>has, makes its IDEAS so powerful that it can easily win
>>to country-championships en passant.
>
>Yes. The IDEA. If they wait for the results-of-the-idea then they are blinded by
>themselves.
>
>The idea was with Botwinnik, but he never made a program, only a program on
>paper; then with CSTal, the idea was there for all to see, but they only
>looked-at-results and said the programmer never offered anything constructive;
>then with Tiger - the anomoly to disprove the old paradigm was there all along,
>but it had to jump up and bite them before they noticed. Tiger bites !!!
>
>If they looked at CSTal, if they listened, they'ld have seen that ALL moves at
>the root were scored, scores broken into components, all to two places of
>decimals. The entire evaluation algorithm was there for all to see. Only they
>didn't look. It was there for five years, only they didn't look. And said the
>programmer 'never offered anything constructive'. One of them, who never let
>anyone ever see his program ever, said "he never offered anything constructive".
>Astonishing how those in another paradigm see the 'truth'.
>
>>
>>christophe is very clever in combining new ideas.
>>old tiger was very strong, and aggressive. that was the reason the
>>old tiger was a beast when it first appeared in the scene.
>>but by tuning and tuning, old tiger (that is now version 13)
>>lost some strength. christophe wanted to make it more accurate,
>>and this completely stopped tiger from playing its own game.
>
>Yes, this is an effect of the new paradigm. More 'accuracy' doesn't help.
>
>>
>>but christophe, and that is his main strength IMO, is capable to
>>say: i maybe made a mistake, i have to reconsider. if i don't get
>>good results anymore, my way is maybe wrong. i have to try out
>>the new way. i have often realized how pragmatic christophe
>>is when it comes to changes in tiger. this is rare IMO.
>
>Yes. Partly. All programmers are very pragmatic. Pragmatism is the driving force
>in normal science. It is pragmatism that causes them to hill-climb.
>
>Christophe's strength was that he was prepared to ask the how and why and to try
>another hill. To his surprise it worked, as it will work for the others. His
>strength was his ability to reassess what he was doing. To ask the questions the
>others didn't think of.
>
>
>>most chess programmers are afraid to change their child.
>>especially when it is very strong. they only tune.
>>little by little.
>
>Yes, exactly. So always they stay on the SAME hill.
>
>>christophe made a new engine, and combined all his experience so far
>>into it, and IMO he also learned from cstal-way and that you have
>>to CONTROL the game, with initiative.
>
>We don't know what were Christophe's sources. He may never have even had a copy
>of CSTal.
I have a copy of CSTal that I have bought in 1997 after the Paris Championship.
Just after the championship, I ran to the nearest "FNAC" store in Paris and
bought it.
I installed on my sister's computer (she lives in France) but did not try it
seriously (I had to go back to Guadeloupe a few days after and had other things
to do).
Since then, I must admit to my own shame, that I have never installed it. I
thought I would install it with the purpose of test-stressing my king safety
code, but I never worked seriously on king safety since 1997. And anyway my KS
code was so stupid that even Genius5 or Fritz2 was stressing it badly.
My motivation for the change I have done (which is the ground of Gambit Tiger)
is that Thorsten Czub, Enrique Irazoqui, Marcus Kästner and Ed have been
repeatedly insisting since a long time Tiger
* was playing passively
* had no idea of what was happening when the opponent was building a strong king
attack against it (especially human players)
I have started to rewrite TOTALLY the part of Tiger's evaluation that deals with
king attacks one month before the release of Rebel-Tiger II. My hope was to
improve Tiger on this matter before the release, but I thought deep inside that
it was impossible to do such a dramatic change in such a short time.
Just one week before the deadline, I had a version that was able to play wild
games and caused a lot of trouble to good old Genius5 and others. It was time
for the release of the second beta of the product, so I decided that I would
provide this new unfinished engine "just for fun" as an additional engine. I
named it Gambit Tiger 0.95. At the same time, a little bit of the new knowledge
of Gambit Tiger has been added to the classic Chess Tiger (which was version
12.95).
Almost immediately (a few hours after the release to the beta test team), the
beta testers started to report that they liked a lot this new engine and that it
was getting good results.
I was expecting the good "emotional" reaction (because I like myself a lot the
Gambit playing style), but not the good results.
The efficiency of Gambit Tiger is a total surprise for me. Call me lucky. But
now I must convert the luck to deserved luck, so I try hard to understand what's
going on, and I have to reconsider many things I considered as "truth".
It is not the first time I have to change my mind deeply. It happened to me on a
regular basis since I work on chess programs. It was hard at the beginning, but
now I'm used to it. So I can stand it.
> But that wasn't important, he ony had to be open to the ideas behind
>it. And these ideas were 'open-source', spoken about many times. Likely he
>didn't even consciously think of CSTal, but, when his surprising results came
>from his act of daring, he already had the language, the words, to help form his
>ideas.
I have nothing yet. I have discovered an anomaly, I was not prepared to it, and
I'm now slowly rebuilding my "knowledge net" to include this anomaly.
I have several hypothesis to explain why it worked when I thought it wouldn't,
but I still don't know which one is correct, or if several are correct, or if
none will help.
This is not exceptional. I have noticed during the years a number of other
"anomalies" which I'm still trying to explain.
Chess Tiger and Gambit Tiger use a number of basic principles. Some of them are
logical and indeed work well. Some of them are not logical and I don't know why
they work. I wonder if I'll know one day. Some of them do not work, but I'm not
aware of it yet and I keep on using them while I shouldn't.
Most of the time I'm not able to say in which category a given principle
belongs. Just to give you an idea of the strength of my convictions about chess
programming, and the level of doubt and uncertainty I am used to work with.
When you have to work under these conditions, then you are more than happy to be
able to count some beans. Sometimes it's the only way to make the right choices.
I believe all the other good chess programmers have to work this way, and are as
capable as me to manage thru the fog.
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.