< Earlier Kibitzing · PAGE 428 OF 428 ·
|May-30-15|| ||zanzibar: After renormalizing the PGN, SCID's <Tournament Finder> reports the following changes:|
1956 .. 12 tournaments -> 11
1957 .. 34 tournaments -> 11
The PGN is now much more manageable, since it finally agrees with <CG>'s view of the games.
|May-30-15|| ||jessicafischerqueen: |
<Z> Where did this pgn come from?
<Re: R2> Pretty sure, and then some...
[Event "USSR Championship"]
[Site "Leningrad (RUS)"]
[White "Isaac Boleslavsky"]
[Black "Lev Polugaevsky"]
1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 d6 6.g3 g6 7.Nde2 Bg7 8. Bg2 O-O 9.O-O Bd7 10.h3 Rc8 11.Nd5 Ne5 12.Nd4 Nc6 13.Ne2 Ne5 14.Nef4 Nxd5 15.exd5 Qb6 16.c3 Rfe8 17.Re1 Qa6 18.a4 Qa5 19.Ne2 b5 20.b4 Qc7 21.a5 Bf5 22.Be3 Nd3 23.Rf1 Nb2 24.Qb3 Nc4 25.Bf4 Qd7 26.Kh2 Bd3 27. Rfe1 Ne5 28.Nd4 Bc4 29.Qd1 Nd3 30.Re3 Bxd4 31.cxd4 Nxb4 32.Qd2 Nxd5 33.Bxd5 Bxd5 34.Rae1 Rc7 35.g4 Qc8 36.Kg3 Rc2 37.Qb4 Qb7 38.Bxd6 e6 39.Bc5 Qc7+ 40.Bd6 Qc6 41.Be5 Rc4 42.Qd2 Rc2 43.Qb4 Qc4 44.Qd6 Qa2 45.Rf1 Qxa5 46.h4 Qd8 47.Ra1 Qxd6 48.Bxd6 Rd8 49.Bb4 Rc4 50.Bc5 b4 51.Bxa7 Ra8 52.Ra5 Rc6 53.Re1 b3 54.Ra3 b2 55.Rb1 Rc7 0-1 >
from the ZipFile snapshot of 2015-02-08.
I just downloaded all the games from this event from <Rusbase>. They list their source of the pgns as <Chessbase>, but none of the rounds are given in any of those pgns.
The "snapshot" is of our own cg.com pgns right? Which means that we must ask <Phony Benoni> where he got the round information?
Please correct me if I'm wrong on this.
On the main topic, and now that you have restated the main goal again, I have to agree:
I hope our webmaster will work with you to normalize the tournaments and to find that "bug" that apparently drops games right out of <TI Events>.
Perhaps it's time we stopped promoting new events for a while? At least until our webmaster reports that he has fixed that "bug"?
I don't see much point in promoting further events if the events already promoted are being ruined by the "bug."
What do you think?
|May-30-15|| ||zanzibar: <Jess> I'm actually working with several snapshots. |
That particular PGN comes from the ZipFile area <CG> provides to premium members. Specifically the chronological selection containing tournaments from 1956.
It's from Feb 2015 as I said.
* * * * *
Yes, the fact that <CG>'s games had the round numbers, but <RUSbase> didn't, was terribly impressive to me when I first learned it.
It's still impressive to me now, even though I'm a little more seasoned at this point.
I think <Phony>'s technique just needs to be better documented, and maybe discussed and debated again in the Bistro.
Certainly, tournaments with rounds derived from tournament books, etc. should be clearly distinguishable from tournament games where the round numbers were reconstructed.
<Phony> has explained his procedure in this post:
Biographer Bistro (kibitz #11004)
* * * * *
Well, we tried that last time, but I'll go along with my comrades again. I think we patched over the seriousness of the problem last time, if truth be told.
Download the games, and try to use <CG> games like <MillBase> games, and the problems are readily apparent.
But most <CG> users must tend to use the online version. Then the problems are better hidden, lurking in the background.
Corrections might make a link go stale in one of <Tab>'s tournaments. Or, a missing round might confuse a user (e.g. <Fusilli>) because the first three games of a player in a tournament weren't all wins, as reported in the intro.
But how often do people note round numbers on <CG>? The main view, pgn4web doesn't even display them.
And how often are people finding games via a tournament page vs looking up the pairing/year? We're well trained in the latter method, since we can't use Event/Site lookup. So most people use a tool they know works all the time.
OK, it's late, and I might be a bit fuzzy. But let me finish with these thoughts:
1) Another boycott would be political only.
2) There are so many "corrupted" tournaments, that any new ones introduced would amount to a negligible increase.
3) The main source of the "bugs" appears to be with the correction process (derespecting tournament integrity).
So, the most effective measure to stop tournament corruption would be to stop corrections.
And all the normalization stuff should be done at <CG>'s end, using well debugged programs.
|May-30-15|| ||Tabanus: Polugaevsky vs A Bannik, 1956 was played in round 2 according to Deplher Kranten (who may have lumped rds 1 & 2), but the pgn says rd. 1.|
Five other games from round 7 have the right round number.
|May-30-15|| ||Tabanus: <1) Another boycott would be political only.>|
Well, I think the bug should be fixed, else there is no point in continuing. And I don't like a program to make perfect tables by inventing wrong data.
|May-30-15|| ||jessicafischerqueen: |
<z>, <tab> Thank you both.
|May-30-15|| ||zanzibar: <Tab>, <Jess> if we biographers decide a boycott would be effective, then I'd support it. |
But I would like to see a list of when tournaments were created, so I could crosscheck. It would be revealing to see if there was any time structure to the corruption.
I understand both of <Tabs> points. Let me offer this two thoughts...
If we have bulk submission, then biographers could work on getting a collection together, and perfected, locally (ie at home/work on their own computer). They could continue working no mater what <CG> was doing. Moreover, they could always keep a snapshot copy of their assembled tournament - like keeping a collection copy, but now guaranteed non-corruptible (at least by <CG>).
As for "wrong data". This wasn't made by a program, per se, people can easily make these schedules up by hand. We'd have to ask <Phony> how he did.
But my point is, how do we know it's "wrong"?
Suppose, you had 5 years of an annual tournament, years 1+2 and 4+5 with tournament books. Suppose further, you find the schedule exactly matches yours.
Now, can you use your schedule for year 3?
What if, in addition, you find 5 (or 10) newspaper accounts all agreeing for year 3?
The argument could be made to adopt the schedule. It's wouldn't be as definitive as having the tournament book, and would need to be qualified.
Of course, the simplest approach would be to reject it. That argument remains as well.
But we have adopted it already, I would like to see the issue reopened and debated.
BTW- I assume you're aware that <CG> changes moves in games on what is perhaps even less certain conditions?
|May-30-15|| ||zanzibar: <RE: Tournament Restructuring>|
Do people agree with the three-tier structure?
* * * * *
Consider the following tournament:
EU-ch 4th (2003)
Should the hand-crafted tournament output of biographers go into the same hopper as this?
A tournament with no introduction,
with players listed by rating (w/o rating) and not alphabetically,
with all the games in one big list without round numbers or dates,
and which took place who knows where or when (other than year) from the tournament page?
|May-30-15|| ||Retireborn: <z> If by schedule you mean starting dates etc I would caution against making inferences. For example, I know Hastings traditionally starts on Dec 28th, but ssometimes it has been Dec 29th and in 1999 it didn't start until Jan 2000 for some reason.|
As for round numbers, Berger pairings are often used with round robins to determine who plays when with what colour (conveniently set out here:-)
So if you have a certain amount of information (usually gleaned from biographical books, magazines etc) you can sometimes deduce the Berger numbers for every player in the tournament and hence every round pairing.
I have done this myself in the past for some tournaments eg Moscow 1956 and I assume <Phony> uses a similar technique. More recently I am usually relying on CG for round information!
There are some tournaments (esp. Interzonals) where the Berger is modified or not used for political reasons, though (players from same country made to play each other early.)
|May-30-15|| ||zanzibar: Sorry <RetireBorn>, I was specifically referring to pairing schedules. |
We should do our best as biographers to strive to fill all such info from reliable sources.
But I'm open to debate on the interpolation issue involved here for round numbers.
As for dates, I've argued that they aren't so important, once the bracket is known. I like to have them, if possible.
But sometimes it seems too much work for the reward. And also, the Dutch papers I've come to rely so heavily on, seem to not be available for Sundays. So I'm often left with a Saturday/Sunday ambiguity in the end.
BTW - I'm such a nerd on this stuff that I'd find mention of modified Berger pairing (scheduling) in the intros an interesting aside.
|May-30-15|| ||Phony Benoni: <I think <Phony>'s technique just needs to be better documented, and maybe discussed and debated again in the Bistro.>|
Well, OK. Don't say I didn't warn you.
But let's acknowledge the elephant in the room right off. This method works <IF AND ONLY IF> the tournament used the same Berger Table you are using, and used it considtently. Now, these tables have been around for a good long time, and when I have been able to verify information I have found that the USSR Championships often used them. But even when the models predict every color accurately, and are consistent with any scraps of information that come up, still they possess only a high degree of probability, not a certainty.
My sin in the past has been not sufficiently recognizing and acknowledging this uncertainty, which is why I keep harping on it now.
That being said, here a a page of Berger Tables:
The mission is to find each player's pairing number which is made possible by certain recurring patterns.
In an 18-player tournament, players 1-9 get 9W/8B, players 10-18 get 8W/9B. Generally players alternate colors, but since there will inevitably be situations where two players are due for the same color most will jog at some point. Here is a list of the "jogs":
Rds. 1/2 #1=2W, #10=2B
Rds. 3/4.#2=2W, #11=2B
Rds. 5/6.#3=2W, #12=2B
Rds. 7/8.#4=2W, #13=2B
Rd. 9/10.#5=2W, #14=2B
Rds. 11/12.#6=2W, #15=2B
Rds. 13/14.#7=2W, #16=2B
Rds. 15/16.#8=2W, #17=2B
Players #9 and #18 alternate the entire way. I think the pattern is obvious, and it can be applied to any size tournament.
There are easy ways to do this, if you have the right information. For instance, if you have a complete record of one players pairings by round, you simply look at his color pattern to determine his pairing number, find the appropriate number from the Berger Table for each round, and assiggn the number to his opponent. The rest is mechanical.
Tal played in the 1956 USSR Championship. He is a very well-documented player, with several complete game collections. If you can find one with all his games in round number order, you can use this method to construct a model for the whole tournament.
Another method can be used if you know the full pairings for rounds 1 and 2 (or any set of consective odd/even numbered rounds). Examine the games to see who gets 2W/2B, assign them their proper pairing numbers, then assign pairing numbers to their opponents, and eventually you will be able to construct a complete chain. This is quite helpful when newspaper coverage is very complete in the first few rounds, but slacks off later.
This is all fine and dandy, but what about the situation where all you have is a list of games in order by White's name? There is a way, but I haven't used it for years and need to brush up. Let me grab some lunch first.
|May-30-15|| ||Tabanus: <I understand both of <Tabs> points.> And I'm starting to understand you more, or so I believe :)|
<Do people agree with the three-tier structure?> Seems like a good idea? 1) One for constantly added new (2015) tournaments, 2) one for voted in "historical tournaments" and 3) one for .. maybe a kind of pool of complete tornaments w/o table and intro that we could (make and) pick from for adding to 2).
|May-30-15|| ||Retireborn: <z> re: dates, I'm happy enough just to know what month a tournament started in, myself.|
Modified Berger for IZTs was designed to prevent numerous Soviets arranging useful results for each other later in the tourney (or, to be fair, to protect them from such accusations) and was applied to all nationalities. I recall an interview with Hort about Palma 1970, where he was unhappy because he had to play Filip in the first round, and all his subsequent opponents played Filip right before him, which (according to him) was like a free day for them.
<Phony> Very interested to learn more about your Berger-determining method.
|May-30-15|| ||chessgames.com: Is this all the same player?
|May-30-15|| ||sneaky pete: Yes, better known as
|May-30-15|| ||Tabanus: H Foerder is same as Yosef Porat (cf. Olimpbase).|
Heinz Foerder and Foerder are more doubtful, esp. the latter needs more proof or investigation.
|May-30-15|| ||Tabanus: If Heinz Foerder/Porat emigrated to Palestine in 1934, it's far from certain that Heinz Foerder vs Alekhine, 1935 is him.|
|May-30-15|| ||Phony Benoni: Again, I have done this exact procedure for years, and hope I'm remembering it correctly.|
<Determining pairings with only a list of games>
This will be for an 18-player tournament. Some adjustments have to be made in other sizes, but this will demonstrate the idea.
<STEP 1: List the players in two groups>
Group I: 9 Whites, 8 Blacks (players 1-9)
Group II: 8 Whites, 9 Blacks (players 10-18)
This is easy enough if your list is arranged by the name of the white player.
<STEP 2: Determine player #18>
In the Berger Table for 17/18 player tournaments::
You will see player #18's games listed in the first column. Note that he gets Black against each player in Group I, White against each player in Group II. Look at your list of Group II players, and determine which got White against all the other players in Group II.
<STEP 3: Divide the groups into four subgroups>
To do this, look at each player in Group I and determine which players in Group II he had White against. The pattern works out like this:
Subgroup I-Odd (Players 1, 3, 5, 7, 9) have White against Subgroup II-Even (players 10, 12, 14, 16)
Subgroup I-Even (players 2, 4, 6, 8) get White against Subgroup II-Odd (players 11, 13, 15, 17)
Even though you don't know the exact pairing number of each player yet, you can tell which subgroup is which because Subgroup I-Odd will have five member, Subgroup I-Even four members.
Examine the pairings between the members of each subgroup. There is a distinct pattern which allows you to assign the exact pairing number. Heres' Subgroup I-Odd:
1 3 5 7 9
1 X B B B B
3 W X B B B
5 W W X B B
7 W W W X B
9 W W W W X
So, if one players in Subgroup I-Odd gets 2 Whites and 2 Blacks against other players in the subgroup, his pairing number is "5",.
Again, I think the pattern is obvious, and it holds for all subgroups.
For tournaments of other sizes, some adjustments are necessary. For example, with a 16-player tournament you separate into two Groups of eight players each, then determine player #16 and put him aside for the moment. Then start the subgroup analysis with Group II. Since it now has an odd number of members (without player #16), the subgroups will be of different sizes and you will be able to tell which is Subgroup II-Odd (players 9, 11, 13, 15) and Subgroup II-Even (players 10, 124, 14).
This isn't hard (even working by hand) once you've mastered the idea, and no doubt an algorithm would not be difficult to devise. The question is whethr it's worth it.
|May-30-15|| ||zanzibar: Just in case anybody else might be working on it - I'm looking at <Lone Pine (1981)> at the moment. |
Looks fixable - and perhaps a good example to write up. But still a work in progress.
Here's what I have penciled in:
Player merge: <Brooks> -> <Michael A Brooks>
@p43194 -> @p16824
<Shirazi--Ivanovic> @g1122918 -> R7
<Shirazi--Alburt> @g1017247 -> R4
That leaves just one game with missing round:
Trouble is Thinnsen would like it to be R1, while Wells would like R4.
Ah, I see from <Phony>'s collection it was a makeup game, on April 3.
Now that I understand the concept of xround (cross-round) games, let's call this R5.5.
That fixes <Lone Pine (1981)>, with the exception of adding a footnote about the <Thinnsen--Wells> game.
|May-30-15|| ||Retireborn: <Phony> Many thanks for that, will give your method a try in the near future.|
<Tabanus> wiki page suggests Foerder/Porath did play in the 1935 olm.
|May-30-15|| ||thomastonk: <Foerder/Porat> In contemporaneous sources besides Porat his name is also written as Porath, and so some databases prefer this version, e.g. http://www.olimpbase.org/players/sw... .|
<tabanus> Heinz Foerder vs Alekhine, 1935 was played on 18. August 1935 in the third round according to "Tagesbote" of 28. August 1935. However, there is a little difference in the game score beginning at Black's 20th move. Foerder played for Palestine, of course.
|May-30-15|| ||Retireborn: <thomas> I suspect Porat/Porath is just a matter of which language you are transliterating the Hebrew name into.|
|May-30-15|| ||Tabanus: Ok. That leaves only Foerder to solve.|
|May-30-15|| ||MissScarlett: Compare and contrast
Heinz Foerder vs Alekhine, 1935
Y Porat vs Alekhine, 1935
|May-30-15|| ||Retireborn: Porath was playing in the zonal tournament Hilversum in 1947, but van Steenis wasn't, nor does the score match any other game I know of.|
Game is also here:-
The other "Foerder" game is here:-
No idea how reliable those sites are, but absent any other evidence I think I'd assume that "Foerder" = Porath.
< Earlier Kibitzing · PAGE 428 OF 428 ·
Advertise on Chessgames.com