Website founded by
Milan Velimirović
in 2006
17:28 UTC
| |
MatPlus.Net  Forum General SPAM in MPR |
| |
|
|
|
|
You can only view this page!
| | Page: [Previous] [Next] 1 2 3 4 | | (21) Posted by Sarah Hornecker [Saturday, Mar 13, 2010 09:22] |
I was surprised not less than you are, Sergiy!
By the way, unrelated, I sent a mail about Harmony, do I get a reply?
| | | (22) Posted by Kevin Begley [Saturday, Mar 13, 2010 11:41] |
@Admin.,
This is not an article... this is numerous bullets, each essentially rehashing the conclusion.
Consider bullet 10.1 "'mining', whatever the skills involved, is not composing."
That's all we get -- no support, no critical thinking, not even a good definition for these terms.
In bullet #4, he essentially states that the entire nest of bullets (in which this bullet resides) are his position.
I doubt Gödel explored the logical absurdities of self-referencing statements to this degree!
In bullet #5, he actually declares himself impartial and ethical.
How can you publish this rambling nonsense???
And, it only gets worse...
I'm disappointed -- he should have rewrote his outline a few times, before drafting an article.
Instead, he only resorted to numerology.
| | | (23) Posted by Sarah Hornecker [Saturday, Mar 13, 2010 12:41] |
While we're at Roycroft (but unofficially I heard different things how the SOTY came to be this, making me concluding he was innocent here) the Study Of The Year 2008 is the greatest insult to study composers for a long time.
| | | (24) Posted by Sergiy Didukh [Sunday, Mar 14, 2010 16:10] |
I see there are no one here who could support openly Roycroft's claims that EGTB is a wonderful composing tool, a miraculous source of interesting positions for potential studies. Just order and the new study is served in a couple of minutes. Come on, you, scared sons of chess composition. You wanted a discussion but now you have nothing to say because you don't know the composing process and you have no idea what EGTB is for! Then, keep it quiet and don't spread your lying comments on printed pages.
Discussion closed!
| | | (25) Posted by Sarah Hornecker [Monday, Mar 15, 2010 00:21] |
Pointing out mistakes is one thing, but I don't like the direction to that your words are heading, implying Roycroft is lying. I would rather say he's mistaken since he does not use EGTBs very often.
QUOTE
Discussion closed!
Probably not.
| | | (26) Posted by Mihail Croitor [Monday, Mar 15, 2010 11:19] |
It depends. For Serghey, by example, discussion is closed, for Siegfried - just opened :)
| | | (27) Posted by Steven Dowd [Monday, Mar 15, 2010 22:18] |
I suspect there is a back story here that occurred in the original Russian article and perhaps in private discussions that led to all this acrimony.
Unless there is some way to extract only the positions you want, I suspect the EGTB don't amount to more than a tool; calling them an "oracle" is quite a misnomer, in my opinion. I've used the TB in a variety of ways: Once I looked for R+S vs. R+P because I had an idea that there would be certain positions that would be won for R+S, based on some analysis I did, I found three, all of which were already in the analysis I did.
I looked for 3 pawns versus a piece (B and S). I found one interesting Bishop versus three pawns that had an interesting twinning mechanism based on slight movements of the pawns and made for an acceptable "practical study" (Lehrspielstudie) but nothing more. I was interested in 3 pawns versus S not for a study of any kind, but because I have collected studies, games, and positions with the endgame S+P versus pawns for the last 6 years(I still have a dream of writing a small book on the topic someday), and wondered if there were any conversions where the P might have to sacrifice itself to secure a draw - this is of course outside the realm of the TB, but an intelligent interpreter might be able to see the Vorspiel that led to this position.
The S versus 3 pawns positions bored me to tears looking through them, although I had no way to extract only such positions, which could have led to a quicker search; I had to enter positions by hand that would seem to provide me with what I needed.
I suspect this lends some credence to Sergiy's theory about the tablebases, without the initial analysis I did on the positions, I would never have found the positions I needed as a start to the two small studies I made. Of course, my lack of knowledge in being able to extract positions wholesale doesn't mean someone has not figured out a way to do this.
I don't do a lot of studies, and I do mostly small ones that have some practical value to OTB players, mainly for Benko's column in Chess Life, so my experience may not be representative.
I don't see how the TB offer much of value to someone doing artistic studies, the kind that seem to me to belong in a FIDE Album, has a TB position ever made the Album? (not after the fact, of course, but studies composed only using the TB). I would guess the chances as low if not non-existent. Perhaps JR is discussing what may be possible in the future, but it is my understanding that we are light years away from a comprehensive 7 man TB, and that the 8 man databases will probably come about the day we invade the Delta Quadrant (Star Trek joke).
More interesting to me has been when I find a six man, older study,or one that converts to six man. Looking it up in the TB is an interesting learning experience for me, and I am often surprised by the depth of analysis some of these older studies showed.
I've followed Roycroft's discussions on TB in EG, they seem appropriate enough there; the inclusion here does seem ill-advised to a certain degree without more of what I refer to as the back story. Not having the initial article in English or the responses of the three individuals he mentions (including Sergiy) cripples my understanding of the debate.
Now in chess problems, we do have Poisson's many small problems that are cropping up since WInChloe can now solve mate problem from the tablebases (I've seen them, I think, in just about every magazine lately). In many cases, I don't see the point, they are simple mating continuations, but in many of the problems he shows problem themes such as cyclic play, etc. What is the consensus here?
Anyway, for good or bad, that is my contribution to the discussion.
| | | (28) Posted by [Tuesday, Mar 16, 2010 18:55] |
I thought I'd add some words to this, too...
After reading and re-reading the two pieces by John Roycroft, and the earlier postings, I've more
or less come to the conclusion that John sees oracle endgame databases
(odbs) as an undesirable component in endgame composing, and wants to see some
serious discussion on it.
The undesirability is associated with 'mining', which I take to refer to
a purely mechanical search for positions, in an odb, that fulfil certain
requirements, then taking one of the found position at random,
affixing a name to it, and presenting it as an artistical creation.
I use the word 'at random' deliberately to avoid introducing any aesthetic
evaluation of the positions found. Perhaps that is too restrictive, perhaps
such selection of the best of the positions found are included in the term
already.
Tools for this kind of searching already exist: CQL (Chess Query Language)
is one such platform on which such endgame searches can be performed.
I once had the idea to 'mine' the existing endgame databases for few-move
miniatures to see if there would be any that had not already been created.
(My interest here was entirely centered on the question if this might help
find 'blind spots' in human composing -- really!) That is 'mining' as I see
it -- totally mind- and soulless.
The point where I seem to get stuck is between judging the artistic content
of a study on the one hand, and judging the means by which a study was
composed (or perhaps 'produced') on the other. I feel a kind of
nagging doubt that a mined study can be as good as a composed one,
I must admit. But it also seems the contents is what matters: a gem, even
if found by a mudlarker, would still be a gem.
I don't see that the Codex is involved to any great extent, as some
previous poster seems to
want to invoke it -- it says nothing on what tournament judges may or may
not discuss. Footnote 3 is not clear enough to me to say that it is relevant
or not. The only point that seems to be involved is that 'A chess composition
is the result of an individual creative act of one or more authors', and
there is reasonable doubt that such mining as has been referred to is in
any way 'an indvidual creative act'. But as the Introduction makes it quite
clear that Part I of the Codex is descriptive, not prescriptive or normative,
that particular point probably wasn't intended to carry a great deal of
weight.
I'm not sure if I come to any conclusion: perhaps I'm wrong when I simplify it
to a question of 'contents' vs. 'means'. On the other hand, I do think
this *is* an entirely legitimate area of discussions for judges of studies.
There is, though, a usage of oracle databases that seem to be desirable from
the point of the composer, and it may be necessary to ensure it does not get
tangled up in 'mining' any further than it deserves.
In "The Chess Endgame Study" John Roycroft gives a chapter to the "Cook Hunter:
Friend or Fiend with Microsocope", where he notes that 'Ideally the composer
would like to have a tame cook hunter or tester as an assistant.' (p. 225)
I well believe it: I remember spending several interesting hours with a printed
collection of studies, Fritz 8 and a large set of EGTBs, and gradually coming
to realize that about 10% of the studies in that collection were formally
incorrect, no matter what scrutiny they had passed. And I once traced the very
convoluted story of E. Holm's study in La Strategie, 1912, which ultimately was proved
to have a second solution by Goetz -- after an examination of Puyg y Puyg
was proved to be faulty. An odb would be have been invaluable to
avoid submitting a faulty study -- or accepting it.
This knowledge of faults in studies is probably reflected a few lines further
down from the cited passage, where Walter Veitch is quoted
to the effect that he doesn't need not demolish studies, as they do it themselves.
An odb, then, can clearly be used as such 'tame cook hunter', and I
imagine that many composers would want to use it in this manner. I see nothing
wrong with this -- if such practice would be considered dubious practice, it
would be equally dubious to ask a human cook hunter for help. If carried to excess,
though, it ends up as database mining, and at that point we're back to the
central question again.
| | | (29) Posted by Sergiy Didukh [Tuesday, Mar 16, 2010 20:03] |
Mr.Roycroft’s case.
1. Persistent use of the term ‘mining’ instead of ‘checking’ .
2. The help from this ‘checking’ is exagerated enourmously. Chess programs calculates very quickly difficult positions. Some composers still sometimes don’t think necessary to check them with EGTB, because chess programs are seldom mistaken.
3. Inappropriate exhibition of his own credentials in the field of EGTB and composition (process), which have nothing to do with EGTB and composition (process).
I think it’s not a mistake, these are lies…
That’s why in the discussion that took place earlier in the Russian magazine “Zadachi i Etiudy” I called my response “Eclipse of mind” which means that there’s not any threat to the art of composition from EGTB, these are lies spread around by Roycroft and some of his followers. Sochniev (programist) and Akobia (the biggest ‘thief’ of interesting positions from EGTB in Roycroft’s opinion) in their articles tried to explain that it’s impossible to extract interesting positions from EGTB. I adopted the impolite tactics to accuse the man who did a lot to the development of study composition. It wasn’t an easy decision to take for such a polite man as I am :-)
Some info about the true ‘mining’. Positions with mutual zugzwangs have been mined and about a month ago John Beasly made available a PGN file with all 6-man zugzwangs. There are about 1 million positions, some of them are illegal. The mentionned tool CQL can be applied here in the search of specific positions (material, placement of pieces) but there are no moves, that’s why the search for underpromotions or other ideas cannot be done in the file.
| | | (30) Posted by Georgy Evseev [Wednesday, Mar 17, 2010 11:09] |
Well, I am one who considers it is necessary to seriously take into account EGTB use.
The main argument of Sergiy, Siegfried, Kevin is similar and quite simple. It can be formulated in the following way: "We know how the engames are created. There is no way to integrate the use of EGTB into the procedure. So, there are no real consequences of EGTB existence."
The arguments are quite correct, but the conclusion is wrong. This is because the "mining" procedure excellently described by Anders is quite possible.
It looks like there are no tools for such a mining now, but I think that such tools were simply not made public. But this is only a question of time.
There are two facts quite specific for computer science.
1. Anything created once will be with us forever. So EGTB is with us now and always will be with us.
2. The tools created by experts can be used by non-professionals with the same efficiency.
Concerning the second statement. There are two separate tasks: database creation and database search. Different tools are needed for both tasks. As I have said, there are still no well known tools for database search (using different criteria). But it is not really difficult to create such tools. Yes, I would agree that not all well-known study themes can be mined, but the studies with such traditional themes as mate, stalemate, pawn promotion (and underpromotion), positional draw, refusal of capture and probably others can be "mined" (including multiple instances of theme if the miner wants it).
Also, do not forget the possiblity to extract something with very nonstandard material. Does, for example, Queen and Rook always (100%) win against Knight and Pawn if a stronger side is to move? (I do not have tools to answer this question.) If not, it can be a study position))).
And for the finish I must say that once I tried to do some mining myself using the tools (quite poor in fact) that were available to me. My personal conclusion: yes, in the case of proper tools it is quite easy to extract studies or study-like positions from the database. My own best result is the following position:
 (= 2+3 )
White to win))
Solution:
1.Rd1 Sf3 2.Ke4 Sg5+ 3. Kf5 Sh7 4. Rg1 Sb3 5. Ke6 Sc5 6. Kf7 Sf8 7. Rg5 Scd7 8. Rg3 Se5 9. Kxf8 and white win (several more exact moves may be added in the end of this solution)
I intentionally do not comment on the solution, but I consider this would have been a worthy study if it have not been directly taken from the database.
| | | (31) Posted by [Wednesday, Mar 17, 2010 14:34]; edited by [10-03-17] |
Sergiy Didukh writes:
>The mentionned tool CQL can be applied here in the search of specific positions (material, placement of pieces) but
>there are no moves, that’s why the search for underpromotions or other ideas cannot be done in the file.
Well, since the positions are known, it is not particularly difficult to use them to extract the corresponding
move trees from existing EGTB files. Or -- why not -- modify CQL so that it works not with PGN but with
the EGTB files directly. It's probably not more than what is generally known in the business as a 'small
matter of programming'.
And if the miner doesn't want the bother to evaluate the findings, some time ago I noted an article on
a computer program that improved problems. It didn't do it too well, but that was probably more a question
of badly configured scoring functions (it didn't appreciate miniatures, for one thing). I can imagine
a similar program, with an endgame study specific scoring function, and a connection to EGTB files for deciding
if the new positions are acceptable as studies (the problem-improver had to solve the problem to make the same
decision). Just enter a few pieces, let the program chew on the position, and after some hours it will output
a position that it thinks is superior. And then it seems to be largely a question of tuning the scoring function.
And that could be done by software as well - collect a number of scoring attributes, set some weighting factors, and
use prize-winning studies as teaching material to recalculate the weights. It would be interesting
to know just how close pure brute force could get to identify good studies with such 'training'.
That the basic mechanics works was probably demonstrated by Samuels' checkers program.
That last point is probably the right place to start -- if it is possible to 'score' a study reasonably well by
a purely mechanical approach, the application of that to identifying them in odb won't be far away.
(Added: but perhaps I merely demonstrate my ignorance of evaluation of studies here?)
It would probably be enough if a professor in computing science (or perhaps AI, if that's still a politically
correct research area) to decide to look at game aesthetics, and have his undergraduates do work around it.
I'm reasonably convinced that mining with some kind of evaluation will be a reality soon enough. Might
even exist already -- it's not as if this is rocket science or requires some major break-through.
The main thing was the EGTBs, and those have been with us for quite a few years already. Any half-decent
programmer can work with those.
So the legitimacy of the suggested area of discussion remains valid: if it is not a fact today, it
may very well be tomorrow or the week after that.
The main question remains: would that be a problem to the study world?
| | | (32) Posted by Sergiy Didukh [Wednesday, Mar 17, 2010 19:08] |
EGTBs as well as the extracted PGN file of zugzwags don’t have moves. Everything is stored as positions. In order to get something that looks like a suitable line for studies (the line without duals) there exists another program ‘Wilhelm’ (public). It builds up lines comparing positions – ply by ply. That’s all I can say about this program because I don’t use it. Anyway, you have to download EGTBs on your computer so as to work with it. Maybe, the chief ‘miners’ Y.Konoval and M.Burzhutsky who generate EGTBs (every composer must be thankful for this precious tool that he uses sometimes to check lines) use some other programs too. Guess what ideas can be extracted using these proceedings. Mutual zugzwangs, underpromotions, stalemates, mates, dual-free lines … that’s all, or the rest is of no value to a composer. The perfect result of this work could be a PGN file with mutual zugzwangs.
And we have it at last! The problem is that there are 1 millions positions there, 99.9% of them cannot be used as good material for studies. The other 00.1% (if found) have to be developped with play that puts both sides in zugzwang, these positions have to be cleaned from incomprihensible lines and so on … I know only 5 or 6 people who can do such work. The knowledge and very good technique are required. This makes the chances of success for anyone who is not a good composer close to zero …
Anders suggests to add moves to this file and then use CQL (I had this idea too long ago). This could increase our chances of success. Unfortunately, the size of the file would be so big that it would have taken years to catch something belogning to this 00.1%, … plus the time to acquire excellent technique of composing …
In case of stalemates and mates the things look worse. These positions are much more than one million. I can set up 10 stalemates or more in one minute without any kind of tools. Why to ‘mine’ then?
The hunt for dual-free lines as the one that Georgy describes as his best result gives no prospects to a miner who intends to be a study composer. I don’t think this position is a study. Maybe it has some value for endgame theory, but a study requires more.
This is reality. It gives no reasons to worry for the future of study composition.
| | | (33) Posted by Marek Kwiatkowski [Thursday, Mar 18, 2010 07:23]; edited by Marek Kwiatkowski [10-03-18] |
By the way:
EGTBs (<= 6-men) are simply (with solutions) and directly accessible (online without download) from:
http://www.shredderchess.com/online-chess/online-databases/endgame-database.html
It is very strong tool for composers, especially study composers. I recommend it.
| | | (34) Posted by Frank Richter [Thursday, Mar 18, 2010 11:26] |
I think, the EGTB is only a tool like other software. The EGTB changes slightly the composing process, but this is similar to the changes of the composing process in all other genres, f.ex. long selfmates. So I don't understand the argumentation against the "EGTB studies".
Of course theoretically it is now possible for everybody to create 6-men positions with a unique solution. But the human, who find such positions, has to work on this position as in earlier times, he has to look for the aesthetic aspects, the economy of time, for predecessors etc. So he remain in his role as composer. Poisson's 5-men moremovers are very good examples, that a unique line of play for win or mate not in every case is a chess problem too ...
| | | (35) Posted by Gerhard Josten [Thursday, Mar 18, 2010 15:12] |
Okay, let's close our eyes and the EGTB disappear by themselves. Once also a Pope closed his eyes when Galileo had a new idea. But John Roycroft is right in my mind: We have to open our eyes and to organize a sensible use of EGTB.
Gerhard
| | | (36) Posted by Sergiy Didukh [Thursday, Mar 18, 2010 15:50] |
No one says to close one’s eyes on EGTB. On the contrary, its use is welcome. Personnaly I am happy that nowdays composers have such a powerful tool for checking their studies. Do you see any other use of EGTB by composers? Show me where I am wrong and then you can say that Mr.Liar is right.
| | | (37) Posted by Gerhard Josten [Thursday, Mar 18, 2010 18:40] |
If you explain that composers use EGTb for checking their studies, you are telling just one half of the truth, Sergiy. In the meantime there are tools which allow a special search for themes. So CQL was developed by Gady Costeff and Lewis Stiller. They wrote: „CQL specifies a small but powerful set of primitives to define chess themes. CQL can find much more complex themes than any other chess program. Users have searched for themes like stalemates with multiple pins; games in which the same position recurs but with the winning side missing just one piece; Nowotny and Grimshaw themes, games with some number of captures on a single square, games with a certain number of black and white passed pawns, and many more.“ I’m deeply convinced that this program is an often used searching tool for composers of endgame studies. They replace composing by searching.
In EG No. 151 -(Vol.X) of January 2004 Gady Costeff summarized: „CQL was invented to provide powerful search capabilities of chess positions, especially in conjunction with the Harold van der Heijden Study Database (HHDB). CQL allows researchers to match complex thematic requirements that cannot be accomplished using existing facilities. The target users of CQL are composers, researchers, tourney directors and judges.“
This trend will not be the end of the development. On the other hand, studies with some more than six pieces are increasing which after few captures are ending in EGTB. In these cases the only creative act has been to have produced an introduction. However, your theme is welcome because we need a friendly and unprejudiced discussion.
Gerhard
| | | (38) Posted by Sarah Hornecker [Thursday, Mar 18, 2010 19:09] |
Again: EGTB can be searched for a specific position, but not for multiple positions within a mating line. So themes like "The same position appears with one piece less" are not searchable in EGTB, to my knowledge.
| | | (39) Posted by Dan Meinking [Thursday, Mar 18, 2010 20:27]; edited by Dan Meinking [10-03-18] |
The whole question of 'mining' reminds me of the late Edgar Holladay. In his later years, Edgar's concentration and attention span waned, but he still wanted to be productive. And was he ever!
His typical 'composing' protocol was something like this:
- take a combination of material that have known mates, stalemates, etc.
- place them semi-randomly on a board, in Alybadix
- run Alybadix with, say, h#3 stipulation; then try h#4, etc.
- maybe try some twinning ideas: move a piece, shift the board, etc.
- repeat this process until a worthwhile problem emerged
In other words, Edgar (in his later years) did almost no 'composing' in the usual sense. But does that mean he should credit Alybadix for his findings? Absolutely not!
In the case of EGTB: understanding what makes a good endgame takes an 'artistic evaluation' that no program (yet) can conquer. Therefore, it is the composer -- not EGTB -- who has made the 'discovery'.
| | | (40) Posted by Kevin Begley [Friday, Mar 19, 2010 03:11]; edited by Kevin Begley [10-03-19] |
I agree, Dan.
The simple truth is, 'mining' is far more common in asymmetric problems, echo-mates-, sentinel-problems, and the like...
Using limited material, one can drop a few weak pieces, such that very few mates are possible...
Then, expecting to produce echoes, via twinning, shifting, and repositioning, let the computer find the optimum setting.
There are problems of these types where I often wonder if the human was completely out to lunch.
I concede, it is more involved than this -- but I do get that feeling.
And, other times, it's obvious that a computer was only used to speed up the convergence of phases.
Example:
K.Begley (original)
 (= 1+1+3N )
h#4 Circe Parrain (no promotions -- nP only promotes to nP)
2.1.1...
1.Kb8 axb5 2.Ka7(a4)! b6+ 3.Ka6 b4 4.axb3 e.p. b4(b5)#
1.b4 a5 2.Ka6 b5+ 3.Kxa5 b3(a6) 4.axb5 Kc5(b4)#
chameleon echo mates.
Did I create the echoes, or did I use exhaustive computer assistance to 'mine' them?
The truth lies somewhere in between... but why should it matter?
It's possible to twin dozens more echoes -- moving bPb5->b1 gives a third... and it's easy to find more optimal diagrams, too...
The question for the judge should be: did I come to win chloe with nothing more than a limited amount of weak material (to restrict mates), and an eagerness to type out (and later wade through) hundreds of twins?
Mr.Roycroft, on the other hand, might suggest that we ban such content.
I have yet to see an endgame study where I get the feeling that a composer came to the EGTB with less than a strong thematic idea.
I would submit to Mr.Roycroft that a good judge (of which FIDE has very few!) should be capable to distinguish.
To seek to exclude 6-men endgames (via Codex alteration) is to go after a phantom mosquito with a machine gun.
Just go back to the 'mining' analogy...
The value of gold depends upon purity & mass -- regardless whether it was found in the pan, or uncovered by industrial mining.
More to the point, the value of any commodity depends upon supply and demand.
If modern techniques should dramatically increase the supply of particular types of gems (as is the case with many types of echo problems, for example), then the value of all such problems (which might have been mined) is accordingly diminished.
What we don't do is devalue all types of diamonds, because *some* cannot be distinguished from cubic zirconium.
Nor do we do this for the benefit of those, in the business, who don't know what to look for...
Mr.Roycroft has provided no evidence that 6-men endgame studies are being heavily mined today.
And, he continues to ignore the fact that computer techniques are far more prevalent in other genres.
Delicate questions, like computer assistance in composing, require a much broader perspective than he has yet provided.
I'm disappointed such an article was published, in such a quality journal.
It reads to me as though Mr.Roycroft is more interested in pretending to be an authority, rather than advancing the discussion.
| |
Read more... | Page: [Previous] [Next] 1 2 3 4
MatPlus.Net Forum General SPAM in MPR |
|
|
|
ISC 2024
