"Full-width means never having to say you're sorry"

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mongrel

"Full-width means never having to say you're sorry"

Post by mongrel » Fri Aug 31, 2007 10:44 am

Toward the end, I'll explain why I bring up this question.

"Full-width means never having to say you're sorry".
(It might have been "brute-force" instead of "full-width", not being a programmer, I'm not sure that these are not the same.)
Who said that?
I tried Googling it.
I'm guessing it was Larry Kaufman, or Larry Kaufman quoting someone, in the thin white booklet, called, I think, Computer Chess Reports? This was from before I had ever seen the Internet. It was published by? ICD? the previous name of ChessUSA? our forum host?

I remember that it was a believable authority to me at that time, and, as I was toying with the idea of writing a program, I probably agreed with it: it would appeal to an attitude of extreme micro-control, or of wanting to be the total master over the objects being manipulated: in chess- pieces; in chess programming- representations of pieces.
I think I remember the context: It was during a time, before Richard Lang, probably, when there was still some debate about whether or not it would ever be possible to do major pruning, such as we have now. The conservative approach then was to look at everything. In testing against tactical problem sets, the selective programs were missing things.
Then Richard Lang's programs appeared (Mephisto, Genius), and the computer chess world was blown away by them, permanently.
(It is possible that the exact issue was whether or not a static exchange evaluation would ever be possible, but the principle is the same.)

Now I'll tell you why I mention this...
It is in relation to that other thread I started...
A revolution occurred in computerchess (I'll say "revolution" instead of "paradigm shift", to spread the fear and dread around a little.), and the quote in question was suddenly a derelict of a previous age, like the alleged "Let them eat cake.", or the alleged "Madam, -od Himself could not sink this ship".
The previous top programs (Fidelity? Novag?) were hurt, and had to adapt.

Well, the present top programs and programmers could suddenly find themselves naked too, when a person, or team, having sufficient gravity in this field (and the necessary chess and computerchess skills of course), suddenly realizes the principle of the new age in computerchess, and nails it to the door, so to speak. (In all of this I'm also relating to another thread, the one about an alleged crisis in this forum, or in computerchess generally.) And it could be, I'm suggesting that they could argue, that, given the collapse of the Computer versus Grandmaster epic drama, people are bored, and tired of being swindled by these stupid, successfully overcontrolling, calculating engines, to which they can't relate in some sense. And that this failure to relate, in the sense in which I mean it, or in whatever sense they choose to argue it, which was formerly too expensive to indulge or concede to, in that it would slow down the engine, is now not. I know that the following restatement of my argument is heresy. I clearly remember how much I would have despised it. But here it is: I'm talking about the inevitable grand entry of the feminine principle into this field. You know it is inevitable. And we should probably accept that it is ultimately desirable.

Before you dismiss this, and at the risk or boring you further, I'll tell you my qualifications and personal experiences in this matter.
I was serious about chess, but my education was in Math. Trying to keep this brief, this is what I observed: somewhere between undergraduate and graduate work in pure math, its nature changes; it becomes more like poetry. (I'm talking about writing "proofs".) Instead of, or speaking more accurately, in addition to, the microcontrolling and calculating (brute-force), it calls for language skills that I had always thought, since childhood, were antithetical to math, to the kind of person I imagined myself to be. What I didn't understand at that time, which I think I might now, was that "the feminine principle", whatever that is, was lacking in my whole approach to math, chess, and life, too. And it hurt me dearly.
I'm arguing that its absence is hurting computerchess in an analogous way, and that someday this will be understood.

User avatar
mclane
Posts: 18390
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: "Full-width means never having to say you're sorry&

Post by mclane » Fri Aug 31, 2007 10:59 am

it is difficult for me to answer your posting. i will try. somehow i do not understand everything you say.

richard lang did much pruning in OWN plies 1-3-5-7... and less pruning in opponent plies such as 2-4-6-8-...

this lead to a playing style that was DEFENSIVE because it often saw THREADS AGAINST him but not all BEST moves for the own plies.

i give an example, maybe you can replay it with your Lang program...

this is karpov-topalov
Image

1.Rxe6

old dedicated units of genius program were unable to find 1.Rxe6.

BECAUSE the move is not considered in the move-list.

now take the position 1 ply before

Image

here black (topalov) will play Rc8 but take a closer look in ply 2 and 4 of the computers main line. you will almost INSTANTLY see genius bringing
Rxe6 and Rxg6 into the main line. because NOW and HERE the key moves are in the opponents-ply 2-4-6-8-....

and those moves were computed not that much selective but more brute force.

this technic helped genius to dominate many many years in the computerchess market. it won almost any championship.

but with the invention of the NULL-MOVE search technics suddenly became that much effective that it was possible to do a deeper search than genius without overseeing moves like 1.Rxe6.

the last dedicated unit we had that was doing really brute force was the
B+P module and the MMII module of H+G for the mephisto machines.


in those times we had the following results (on a 400 mhz PC with 64MB Hash) for the above position:

Code: Select all

Fritz7.0.0.8               00:00
Hiarcs8                    00:05
CSTal2.03                  01:29
Shredder6 Paderborn        19:30
Genius3             not in 30:00
Genius4             not in 30:00
Genius 6.5          not in 30:00
Fritz was the null-move "monster".

mongrel

Re: "Full-width means never having to say you're sorry&

Post by mongrel » Fri Aug 31, 2007 11:21 am

mclane wrote: it is difficult for me to answer your posting. i will try. somehow i do not understand everything you say.
I'll explain. But first: Hello Thorsten. I think I remember you. Are you one of the people who defended (Chris Whittington?) and (CSTal?), arguing that certain conservatives here were too resistent to new ideas? Was that you, mainly? Well, my post is an attempt to extend now, what you were trying to say then. (I'm sorry if it is another person I'm remembering.) By the way, I was one of the conservatives. Do you forgive me?
What is your native language?
mclane wrote: richard lang did much pruning in OWN plies 1-3-5-7... and less pruning in opponent plies such as 2-4-6-8-...

this lead to a playing style that was DEFENSIVE because it often saw THREADS AGAINST him but not all BEST moves for the own plies.
Thanks. I remember that now. I was Karpov fan, and at that time it reminded me of him a little. And to me, that resemblance was a miracle, a great achievement. Edit: Also, it corresponded to the way I played, calculations focusing more on my opponents cheap-shots. I know why certain players seem to play stronger once their losing: they start calculating for themselves!
mclane wrote: i give an example, maybe you can replay it with your Lang program...
Sadly, I no longer have it. Your example is perfect.

User avatar
mclane
Posts: 18390
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: "Full-width means never having to say you're sorry&

Post by mclane » Fri Aug 31, 2007 12:11 pm

I'll explain. But first: Hello Thorsten. I think I remember you. Are you one of the people who defended (Chris Whittington?) and (CSTal?),

yes.
arguing that certain conservatives here were too resistent to new ideas? Was that you, mainly?
yes. but it is years ago.
10 years maybe. or even more.
in those years CSTAL had to fight mchess, genius, fritz and hiarcs.

they were the big ones.

Well, my post is an attempt to extend now, what you were trying to say then. (I'm sorry if it is another person I'm remembering.) By the way, I was one of the conservatives. Do you forgive me?
i could forgive you :-)

What is your native language?
german. but i give my best in english, only it is something not good enough.

Thanks. I remember that now. I was Karpov fan, and at that time it reminded me of him a little.
yes indeed. langs programs played little like karpov.
or petrosian comes to my mind.
And to me, that resemblance was a miracle, a great achievement. Edit: Also, it corresponded to the way I played, calculations focusing more on my opponents cheap-shots. I know why certain players seem to play stronger once their losing: they start calculating for themselves!
Sadly, I no longer have it. Your example is perfect.
yes but todays chess programs come MUCH MUCH deeper than they came
in the old days, many new pruning ideas have been tuned almost perfectly today.
it looked a while this could continue endlessly, when suddenly a program appeared that seemed to be DIFFERENT than others.

rybka.

before rybka the open-source made a big jump too when fruit appeared.

i guess fruit helped many programmers to increase the playing strength of their engines.

so much has happened.

there was btw. another weakness in genius, the static-exchange-evaluator.
often when genius is doing a mistake, it has to do with this algorithm.

mongrel

Re: "Full-width means never having to say you're sorry&

Post by mongrel » Fri Aug 31, 2007 12:56 pm

mclane wrote:
What is your native language?
german. but i give my best in english, only it is something not good enough.
The fault is mine. I must remember that the membership is international.
mclane wrote: so much has happened.
When I returned to this forum, I was happy to see some names here, of people who would remember the old issues.
mclane wrote: there was btw. another weakness in genius, the static-exchange-evaluator.
often when genius is doing a mistake, it has to do with this algorithm.
Yes.
It was a compromise.
It surrendered some total control, and permitted something more elusive: "behavior". And a program that has "behavior" could "misbehave". But the programmer, surrendering his need for the "total control" that I referred to, permits the birth of something higher. I didn't consciously plan for this to fit so neatly, but when I mentioned my admiration of Karpov, and my recognition of something like Karpov in Genius, and described this as a miracle, I was remembering a good illustration of the thesis of this thread.

No, it wasn't Mr. Lang who said "sorry".

bob
Posts: 20923
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: "Full-width means never having to say you're sorry&

Post by bob » Fri Aug 31, 2007 4:58 pm

mongrel wrote:Toward the end, I'll explain why I bring up this question.

"Full-width means never having to say you're sorry".
(It might have been "brute-force" instead of "full-width", not being a programmer, I'm not sure that these are not the same.)
Who said that?
I tried Googling it.
I'm guessing it was Larry Kaufman, or Larry Kaufman quoting someone, in the thin white booklet, called, I think, Computer Chess Reports? This was from before I had ever seen the Internet. It was published by? ICD? the previous name of ChessUSA? our forum host?

I remember that it was a believable authority to me at that time, and, as I was toying with the idea of writing a program, I probably agreed with it: it would appeal to an attitude of extreme micro-control, or of wanting to be the total master over the objects being manipulated: in chess- pieces; in chess programming- representations of pieces.
I think I remember the context: It was during a time, before Richard Lang, probably, when there was still some debate about whether or not it would ever be possible to do major pruning, such as we have now. The conservative approach then was to look at everything. In testing against tactical problem sets, the selective programs were missing things.
Then Richard Lang's programs appeared (Mephisto, Genius), and the computer chess world was blown away by them, permanently.
(It is possible that the exact issue was whether or not a static exchange evaluation would ever be possible, but the principle is the same.)

Now I'll tell you why I mention this...
It is in relation to that other thread I started...
A revolution occurred in computerchess (I'll say "revolution" instead of "paradigm shift", to spread the fear and dread around a little.), and the quote in question was suddenly a derelict of a previous age, like the alleged "Let them eat cake.", or the alleged "Madam, -od Himself could not sink this ship".
The previous top programs (Fidelity? Novag?) were hurt, and had to adapt.

Well, the present top programs and programmers could suddenly find themselves naked too, when a person, or team, having sufficient gravity in this field (and the necessary chess and computerchess skills of course), suddenly realizes the principle of the new age in computerchess, and nails it to the door, so to speak. (In all of this I'm also relating to another thread, the one about an alleged crisis in this forum, or in computerchess generally.) And it could be, I'm suggesting that they could argue, that, given the collapse of the Computer versus Grandmaster epic drama, people are bored, and tired of being swindled by these stupid, successfully overcontrolling, calculating engines, to which they can't relate in some sense. And that this failure to relate, in the sense in which I mean it, or in whatever sense they choose to argue it, which was formerly too expensive to indulge or concede to, in that it would slow down the engine, is now not. I know that the following restatement of my argument is heresy. I clearly remember how much I would have despised it. But here it is: I'm talking about the inevitable grand entry of the feminine principle into this field. You know it is inevitable. And we should probably accept that it is ultimately desirable.

Before you dismiss this, and at the risk or boring you further, I'll tell you my qualifications and personal experiences in this matter.
I was serious about chess, but my education was in Math. Trying to keep this brief, this is what I observed: somewhere between undergraduate and graduate work in pure math, its nature changes; it becomes more like poetry. (I'm talking about writing "proofs".) Instead of, or speaking more accurately, in addition to, the microcontrolling and calculating (brute-force), it calls for language skills that I had always thought, since childhood, were antithetical to math, to the kind of person I imagined myself to be. What I didn't understand at that time, which I think I might now, was that "the feminine principle", whatever that is, was lacking in my whole approach to math, chess, and life, too. And it hurt me dearly.
I'm arguing that its absence is hurting computerchess in an analogous way, and that someday this will be understood.
It sounds much more like something said by David Slate. They were the first to return to full-width in the middle-70's and used it to win every computer event they played in until Belle came along with far faster hardware.

User avatar
mclane
Posts: 18390
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: "Full-width means never having to say you're sorry&

Post by mclane » Fri Aug 31, 2007 5:19 pm

there was always a fight of ideologies when those programs fought each other.
the more brute-force kittinger versus the less-tactical spracklens versus the very very selective nitsche/henne programs versus the asymmetric richard lang programs.

it developed in waves. one year the brute-forces were better (super-constellation and kittingers with piece-square-tables)
and next year they were beaten by more selective programs.

but mephisto III with the 3-4 NPS on the 16-bit hardware was beaten by langs amsterdam on the same hardware.

h+g kicked nitsche/henne out of the competition.
novag always stand to dave kittinger.
scisys/saitek had kaplan and craig barnes or whoever they were called.
H+G took schroeder, frans morsch and lang in 3 different classes.
morsch for the low-cost, lang for the expensive hardware and schroeder for the masses.

fidelity relied on the spracklens.

the computerchess world of the dedicated machines was in a certain kind of ORDER.

then everything broke down.

similar scenes appeared in the software times. i remember endless debates with ossi weiner about the accurate genius method of computing.
in relation to the "inaccurate" and speculative method of Marty Hirsch's Mchess.

but one day genius did not "worked" anymore the way it should.

those things repeated again with shredder. and one day it will happen to rybka too, thats for sure.

one day the METHOD how somebody has done the trick/ the advantage will be found out and copied in other programs. and than a new paradigm is important and will appear.


i personally regret that Mephisto III was not continued. i still believe in this method.
if somebody is there to transfer the CDL-2 sources into something useful today, i would like to see Mephisto III beeing continued.
would be interesting to see how a program like this, with such a radical method, would play today.
and i do still believe that lang could have continued if he would have had the interest in doing so himself.
but maybe richard was in a crisis too.

it is said the new shredder will be 100 elo stronger. that would be good.
lets wait.

mongrel

Re: "Full-width means never having to say you're sorry&

Post by mongrel » Fri Aug 31, 2007 6:17 pm

bob wrote:It sounds much more like something said by David Slate. They were the first to return to full-width in the middle-70's and used it to win every computer event they played in until Belle came along with far faster hardware.
Well, that was so long ago that I must have read it in a retrospective piece.

It is interesting and fortunate that Thorsten and Dr. Hyatt have contributed to the thread.
First, as I mentioned above, Thorsten argued along these lines many years ago, prematurely, many said back then. And his mention of Karpov gave me the chance to mention this miraculous property of Genius- an "added value" made possible by *just the right* compromise, by Richard Lang. I spent a part of my life studying Karpov, and naturally felt a certain loyalty or attachment to him. A computer program that could play a little like him had an unexpected "added value", and to some extent my investment in Karpov could be realized, or exercised, in Genius. This was an inestimably higher value, than merely that it could also outcalculate me!

Now, it is pleasant surprise that Dr. Hyatt has also made a post. Because, back in the days of Fritz 5.32, it was Crafty (and also Hiarcs, I can't remember the number) that embarrassed Fritz, according to my chess values. I admired these lower rated programs because of their chess knowlege. (To what extent they were really more knowlegeable, or simply were products of decisions that led to their having a convincing *simulation* of greater knowlege, and style, and chess-judgment, I don't know. But that is beside the point.) I'll confess: watching Fritz-Crafty reminded me of my own games, wherein, if you don't mind my saying so, my more "correct" or "beautiful" games were spoiled by opponents' tactical swindles. (Yes I realize that I was simply not as good as I thought. But that is also beside the point here.) To conclude, my attitude was that whatever compromises made Crafty a little lower rated, were worth it in my opinion, and as far as I was concerned, the more the better. And now, though I'm not really immersed in this as before, I'm still looking at Crafty, Pro Deo, Spike, Glaurung, and others, looking for an appealing spark, even though I own Rybka (a great program of course).

That was my attitude then. I followed computer-chess, and I understood that there were compromises. Instead of repeating what I've already said, I'll just say that the time might be ripe for a reconsideration of the balance of these compromises. But anyone making serious compromises, because of what it might do in terms of ratings, which rule now, must *market* the idea discussed in this thread, and in the thread titled "New Paradigm desired, and we'll pay for it".
It seems obvious that it will happen some day. Or, it will be realized to some, believable extent, and then successfully marketed as having been realized.
Hmm. Who was the greatest at *marketing* his program?
It will happen someday, hopefully not after everyone has lost interest and gone home.

Terry McCracken

Re: "Full-width means never having to say you're sorry&

Post by Terry McCracken » Fri Aug 31, 2007 6:21 pm

Thorsten, David Kittenger used Selective Search for a longtime, it was the Spracklens that used Full-Width. BTW, the Spracklen's program faired better that Kittengers programs, albeit I feel Kittenger's programs played more human-like chess and more interesting middlegames, even pawn storms against the castled King, which Spracklen's program couldn't. In many ways Kittenger's programs played better middlegames as well, but faired poorly in the endgame, a source of many of its losses.

Terry

mongrel

Re: "Full-width means never having to say you're sorry&

Post by mongrel » Fri Aug 31, 2007 6:42 pm

mclane wrote:there was always a fight of ideologies when those programs fought each other.
the more brute-force kittinger versus the less-tactical spracklens versus the very very selective nitsche/henne programs versus the asymmetric richard lang programs.
Thanks, your post has reminded me of many things.
Would you agree that as better methods have evolved, the programs have necessarily converged, leaving a less-interesting "ideological" debate over the board? If so, then that would be another reason for some rethinking. This could be especially true for someone in a position to have less to lose. I can think of several such people.
mclane wrote:i personally regret that Mephisto III was not continued. i still believe in this method.
But I don't remember Mephisto III. (My first Mephisto was Amsterdam (which I could defeat!)
What was different about Mephisto III? What is CDL-2?
mclane wrote:it is said the new shredder will be 100 elo stronger. that would be good.
That's right, but you know I'm talking about abandoning, perhaps only with a configuration switch, the fight for more strength, as such, as versus humans I'm arguing that it is already superfluous, in return for chess-values of some sort, that have been neglected for the historical (and psychological) reasons I've offered.
Who has been the best at marketing?

Post Reply