MatPlus.Net

 Website founded by
Milan Velimirović
in 2006

10:27 UTC
ISC 2024
 
  Forum*
 
 
 
 

Username:

Password:

Remember me

 
Forgot your
password?
Click here!
SIGN IN
to create your account if you don't already have one.
CHESS
SOLVING

Tournaments
Rating lists
1-Jan-2024

B P C F





 
 
MatPlus.Net Forum Internet and Computing Databases and accuracy
 
You can only view this page!
(1) Posted by Michael McDowell [Wednesday, Jul 30, 2008 20:23]

Databases and accuracy


On the Chess Problem Net site Siegfried Hornecker posted a "Paul Morphy" mate in 8 which has been known to be a case of mistaken identity for at least 80 years (it is by E.B.Cook).

Siegfried found the problem in WinChloe. This raises the question of what safeguards there are in problem databases to check the spread of misinformation. I'm not familiar with WinChloe. What happens with a case like the above? I would imagine that if the "Morphy" problem was removed from the database it would simply be added again at some point, since it has been republished so many times. Is there a means of annotating problems, so that the Cook would have a note like "This problem has often been misattributed to Morphy"? Another case is the supposed Davidson anticipation of Nissl, discussed on this site recently. How will WinChloe ensure that it doesn’t help to perpetuate this hoax?

Other questions come to mind, such as how improved versions or corrections are recorded. For example, the settings of many of Frank Healey’s problems differ between the 1866 and 1908 collections of his work. Would both settings be recorded in WinChloe? Would the later setting be given under the original source details?

Databases are very useful in many ways, but their potential for contributing to the rewriting of problem history is unfortunately also huge.
 
(Read Only)pid=2545
(2) Posted by Steven Dowd [Wednesday, Jul 30, 2008 21:17]

Is that the one Michael where the mistake is either because it is dedicated to Morphy or because it sits next to a diagram of one of Morphy's games? This is from shaky memory....
 
 
(Read Only)pid=2546
(3) Posted by Paz Einat [Wednesday, Jul 30, 2008 22:54]

WinChloe has an option to detect duplicates. I don't know if "duplicate" means just diagram or also all other details (author etc.).

In principle it must be easy to detect not only duplicates but also mirrored positions. These can be either marked, with link to the relevant problem, or disallowed for entry into the database. We can ask Christian Poisson on this.
 
 
(Read Only)pid=2547
(4) Posted by Juraj Lörinc [Wednesday, Jul 30, 2008 22:56]

There is possibility to annotate problems as needed in the field "Commentaires". Many problems are already remarked with anticipation remark.

Also, versions of the same problem might be all included. There is possibility to indicate both original source as well as source of version, in the "Source" field, with reproduction source noted in the "Reproductions" field.
 
(Read Only)pid=2548
(5) Posted by Sarah Hornecker [Wednesday, Jul 30, 2008 23:00]; edited by Sarah Hornecker [08-07-31]

In WinChloe, it's possible to give the original author and a comment that it is wrongly attributed to someone else (that happened to a famous Platov study of which CP thought, it's by Vukcevich).

Steven, I gave both. However, it is not unusual that an ending of a game is analysed and published as study. Same happened to a game of Ahues where Maiselis made a study and an Israel Championship (PS: May have been Hungarian) pawn ending game in 1958 where Farago was judging and immediately saw how to win, thus made a study (hhdbiii 39341).
 
   
(Read Only)pid=2549
(6) Posted by Steven Dowd [Thursday, Jul 31, 2008 01:25]

I've found a few of those positions where not-so-well known games were converted into studies, usually just because I am also studying the relevant ending in ChessBase. I have a few of those myself but I am hesitant to publish them as I know I am always disappointed when I recognize them... or should I? In any case, I would think the players of the game deserve some credit.
 
   
(Read Only)pid=2550
(7) Posted by Juraj Lörinc [Thursday, Jul 31, 2008 10:29]

By the way, wouldn't this topic fit better into the group Internet and Computing? Perhaps admin would like to move it there.
 
   
(Read Only)pid=2551
(8) Posted by Harry Fougiaxis [Thursday, Jul 31, 2008 11:55]

 QUOTE 
WinChloe has an option to detect duplicates. I don't know if "duplicate" means just diagram or also all other details (author etc.)

I don't know exact details, but based on my experience, I think it works like that : as soon as you insert a new entry in the database, the program checks the position and the stipulation (and, perhaps, the fairy conditions -- but I'm not sure about that). If they match, it gives you an error message about "index violation" (can't recall the message in French right now). This means that such a problem already exists (by the same or some other composer, it does not matter), so what you have to do is update that entry's comments field.

 QUOTE 
In principle it must be easy to detect not only duplicates but also mirrored positions.

At the moment, it does not, so you can have the same problem in mirrored positions, for instance. Of course, when carrying out a query searching for a specific pattern, it can find all possible rotations, displacements, etc.
 
   
(Read Only)pid=2552
(9) Posted by Harry Fougiaxis [Thursday, Jul 31, 2008 12:39]

 QUOTE 
Other questions come to mind, such as how improved versions or corrections are recorded. For example, the settings of many of Frank Healey’s problems differ between the 1866 and 1908 collections of his work. Would both settings be recorded in WinChloe? Would the later setting be given under the original source details?

Versions and corrections are annotated with (v) [you simply check the "version" field] and tagged accordingly in the "dedication" field. What I usually do is to keep the source field the same and add the details (where and when the version was published, of which ID this version is, etc.) in the comments field. Others prefer to mention in the "reproductions" field where the version/correction appeared.

 QUOTE 
Databases are very useful in many ways, but their potential for contributing to the rewriting of problem history is unfortunately also huge.

True. Would you be interested and have the time in helping with old compositions? You are an expert in this field and such projects need the support of competent persons. Someone could extract the problems of a specific period and by specific composers, create a Word document and you (or anyone else interested) can check them.
 
 
(Read Only)pid=2553
(10) Posted by Juraj Lörinc [Thursday, Jul 31, 2008 14:31]

My two cents regarding Harry's description of the duplicate checking process. As soon as you insert a new entry in the database, the program checks the position AND the stipulation AND the fairy conditions (this part I am sure about) AND assignment of fairy pieces to symbols (not so sure, but I mistily remember that from the past).

Of course, for orthodox problems or rather for problems not featuring any fairy conditions or fairy pieces, the first two parts (positions and stipulation) are sufficient.
 
   
(Read Only)pid=2554
(11) Posted by Harry Fougiaxis [Thursday, Jul 31, 2008 15:19]

Thanks, Juraj, for the clarifications. Just for fun, let me quote two particular examples where despite all these checks, the program fails to "understand" that the problems are the same (reasonable for a computer, a bit strange for a human).

1. Same positions, one with the fairy pieces facing to the left, the other with pieces to the right, are considered as being different. One particular example is one problem of yours in my birthday tourney (can't recall which one right now, but it is easy to spot it). Perhaps I wrote to Christian about it and he may have meanwhile replaced one of them (by the way, for those who don't know, you cannot remove a "dead" entry, you can only replace it).

2. Once I found a proofgame which was recorded twice. Same position, same number of moves, no conditions, no fairy units. I could not understand why the program made the distinction. Well, it had a very good reason. In one of them, the "aim" (but in French) sub-field was empty (as it should), in the other "mate" was selected, so the stipulation fields were different, as a whole. I remember it took me some time to figure out what was going on...
 
 
(Read Only)pid=2555
(12) Posted by [Thursday, Jul 31, 2008 19:28]

Michael McDowell notes:

>Databases are very useful in many ways, but their potential for contributing to the rewriting of problem history is unfortunately also huge.

'Database rot' is a well-known phenomenon -- and when database users realize that the considerable time and effort spent in getting one 'good' problem entry into a database may be wiped out the next time they import a problem collection file because there is no discrimination between high-quality entries and average entries, and no way to protect the former from contamination of the latter, I would expect some kind of protest.

Are there any 'content providers'? People who do their damnedest to transcribe and perhaps even edit chess problem information from original sources (rather than later collections), and sell or otherwise make them available as 'collections'? If there are, such people are probably those who could set the appropriate requirements for chess databases: say, that such collections must be possible to protect if the user so decides, that it must be possible to detect that entries in such collections have changed or even relations between entries, or perhaps allow for separating the original entries (and their relations) from the edits and additions the user has made on his own, and so on and so forth.

Harold van der Heijden is the only one I know -- but his HHDB III is rather standaloneish. If it was available in WinChloe format or any similar problem database format, the problem of 'rot' would perhaps been a more well-known phenomenon by now.

Just ensuring that database entries come with information about the source used (say, White's Loyd-book), and the referenced source (New York Albion, this year or that) helps enormously in evaluating the quality of an entry: if the source is doubtful, a second more reliable source may be needed before that source is used. (For example, there's a russian anthology of 19th c. miniatures that I am somewhat wary about: I don't really trust it on its own.)

Perhaps some way of 'signing off' a database entry digitally (and its relations to other entries) might be the best way to ensure and maintain some degree of quality.

And yet, ultimately, anything that can be misunderstood, will be. If there is no understanding of what a piece of information 'means', and within what limits it may be interpreted, even quality input may produce garbage output. If a problem is listed as by 'P. K. Traxler,' there will be someone assuming that 'P' represents a personal name.

And, less obviously, because some database might use a FEN-related position format, not everyone will realize that castling and en-passant information has been synthesized in some way, and in bad cases may lead to the position being interpreted quite differently that was intended.

I seem to have become long-winded -- I better stop now.
 
 
(Read Only)pid=2557
(13) Posted by Michael McDowell [Tuesday, Aug 5, 2008 17:13]

Conrad Bayer
Leipziger Illustrierte Zeitung 1855

(= 8+10 )

Mate in 9

1.Rb7 Qxb7
2.Bxg6+ Kxg6
3.Qg8+ Kxf5
4.Qg4+ Ke5
5.Qh5+ Rf5
6.f4+ Bxf4
7.Qxe2+ Bxe2
8.Re4+ dxe4
9.d4.

Bayer’s “Immortal Problem”. Brian Stephenson tells me the following:

“In WinChloe the source is given as '1st Prize, Era Ty., 1856'. It lists the following places where the problem has been seen.

237 p.62, Beispiele zur Ideengeschichte des Schachproblems, J. Breuer 1982
p.1392, Le Guide des Echecs 1993
4 p.11, Na Malenkom Shakhmatnom Pole, V. Archakov 1983
27 p.11, Shakhmatnaya Mozaika, V. Archakov 1984
F p.204, Die Schwalbe 121 (fév. 90)
I p.132, Die Schwalbe 225 (juin 07)”

All presumably giving the same source (certainly true of Breuer and Archakov), which is incorrect. I could also mention
96 p.114, The Chess Problem, H.G.M.Weenink 1926 (source given as Era 6th July 1856),
82 p.50, The Chess Problem in the 19th Century, E.I.Umnov 1960 (source given as Era 1856), and
788 p.275, The Golden Book of Chess Composition 1850-1913, Y.Vladimirov & A.Selivanov (source as in WinChloe).
Doubtless the list could be greatly lengthened.

I have seen the Era column for 6th July 1856 and yes the problem is there, but it is given with a note saying that it was originally published in a Leipzig column in 1855. David Hooper and Ken Whyld give the problem in the 2nd edition of “The Oxford Companion to Chess” (p.33) under the entry for Conrad Bayer, saying that it was “…first published anonymously in the Leipziger Illustrierte Zeitung, 1855”. It is a little embarrassing to have two (non-problemist) chess historians correcting a long-standing error involving a very famous problem!

Can WinChloe be amended to (a) give the correct source details and (b) record that some of the places mentioned give incorrect source details?
 
   
(Read Only)pid=2575
(14) Posted by Sarah Hornecker [Tuesday, Aug 5, 2008 17:44]

Illustrirte Zeitung, 16th August 1851 (issue 424, p.160) according to Wikipedia which was properly checked by the User Stefan64
http://de.wikipedia.org/wiki/Diskussion:Unsterbliches_Schachproblem
 
   
(Read Only)pid=2577
(15) Posted by Michael McDowell [Tuesday, Aug 5, 2008 20:46]

That is very interesting! Bayer contributed many originals to the Era, and would presumably have sent the #9 himself, therefore why would he get the date wrong? Also Hooper and Whyld were very thorough, and would go back to original sources rather than take things on trust.

This needs further investigation I think, either by rechecking the mentioned column, or if that is not possible, by checking a magazine from the period (Schachzeitung?) to see if the problem was quoted before 1855. Can anyone help?
 
   
(Read Only)pid=2578
(16) Posted by [Tuesday, Aug 5, 2008 21:47]; edited by [08-08-05]

Michael McDowell:

>This needs further investigation I think, either by rechecking the mentioned column,
>or if that is not possible, by checking a magazine from the period (Schachzeitung?)
>to see if the problem was quoted before 1855. Can anyone help?

It's not in any of the transcriptions from (Berliner) Schachzeitung I have (which is
almost everything up to 1870). That means it has not been published in that exact form,
or mirrored, rotated or translated in any form, among the main series of problems. (My
coverage of studies/endgames from this source is incomplete, though.) This is *very* odd
-- when was it called 'unsterblich' and by whom? (Not in Max Lange's Handbuch either...)

It is listed in Harold van der Heijden's HHDB III as a study, source: Illustrirte
Zeitung, 1851.

Breuer has it as winning '1. Sendungspreis, Era-Turnier, 1856'. Surely Sz would have
reprinted it if it really had?

A double-check in Lowenthal's Era-book does not locate it. Bayer did win the main prize
there, but his set contained six problems in the range #3-5.

If this problem was indeed reprinted in the Era, was there a statement that it was
part of the winning set?
 
   
(Read Only)pid=2579
(17) Posted by Michael McDowell [Tuesday, Aug 5, 2008 22:30]

It was definitely NOT an entry in the Era tourney.

Harold includes in his database any directmate which is also sound as a study.
It strikes me as a very dubious practice. Isn't this rewriting history? Shouldn't the composer decide on his stipulation? Is it not absurd to call the following a study?

P.F.Blake
2nd Prize, German Chess Association 9th Ty., 1910

(= 6+10 )

Mate in 2

1.Sa4 with 2 threats, 5 models and completely accurate play.

Anyway, that's a different topic!
 
   
(Read Only)pid=2580
(18) Posted by Sarah Hornecker [Wednesday, Aug 6, 2008 00:37]

I remember having read recently that the Era gave it as a reprint from something in 1855. Sadly I don't know where I read that.

Also, Wikipedia talk page gives Wiener Schachzeitung 1907, p.212 as a source for Illustrirte Zeitung 1851. However, the microfilm of Illustrirte Zeitung is available somewhere, it seems, so if anyone wants to investigate, that'd be great!

I think, it's perfectly acceptable to include problems that would be sound as studies (for easier searching for anticipations), but he should note that these were published as problems.

Best,
Siegfried
 
   
(Read Only)pid=2581
(19) Posted by Steven Dowd [Wednesday, Aug 6, 2008 15:00]

Plus I believe there was just a tourney that combined studies and problems, white to win, black to draw? Siegfried knows more, I am sure. Those would have to go in with multiple stipulations.

I do find the Bayer problem phenomenal and probably as "unsterblich" as the "Partie"....

It certainly reflects the era, no pun intended, of when it was composed.
 
   
(Read Only)pid=2583
(20) Posted by Hauke Reddmann [Wednesday, Aug 6, 2008 18:07]

Me2 :-)
That was in the SCHWALBE. (Duplum Tourney in honor of Josten's 70th birthday)

Hauke
 
 
(Read Only)pid=2584

No more posts


MatPlus.Net Forum Internet and Computing Databases and accuracy