<FIDE Women's Grand Prix
Sharjad, UAE>

Biographer Bistro

And problems like this will occur:

Ilijas Subasic is a fringe case; we don't have a lot of players with an unknown year of death. Those few situations could be solved by adding the year of death and noting in the biography that it's approximate.

If we were really ambitious we could have a "circa" checkbox next to dates, and then it could say he lived to be ~58 years old.

But since I'm in the neighborhood I was wondering about something...

<CG> has a couple of duplicated games for <Tromso Women's>, e.g.

< 2014.08.02 (R1.22) 0-1 Chelsie Monica Sihite -- Darlin Rocio Villar Acevedo (B18) 27

2014.08.02 (R1.22) 1-0 Chelsie Monica Sihite -- Darlin Rocio Villar Acevedo (B18) 27 >

C M Sihite vs Darlin Rocio Villar Acevedo, 2014 (1-0)

http://www.chessgames.com/perl/ches... (0-1)

Almost certainly the first, White win, result is correct. But FIDE has the second result.

Since both <CG> games, and the <FIDE> game, give the game as 27 moves, it appears that <CG> does some kind of <post-production> to fix the result.

Now, maybe the <post-production> is done elsewhere, like at <TWIC>.

1) Can somebody explain exactly what <post-production> is being done?

2) Why isn't the stale game deleted afterwards?

<virginmind: <cg.com> Alina Motoc and Alina L'Ami is one and the same person. Her name was Motoc until she married Erwin L'Ami a few years ago and so became Alina L'Ami. I've noticed you're keeping separate player pages for <Motoc> and <L'Ami> - maybe you should do something about it.>

In the leaderboard is:

Alina L'Ami 0/2 (+0 -2 =0)

1) Can somebody explain exactly what <post-production> is being done?>

First, I don't understand the significance of 27 moves. Yes, the game was 27 moves long (both versions).

Anyhow, "post production" does take place--sometimes TWIC fixes up games and even helpfully adds tags. Sometimes we fix things up. But most likely in this case, FIDE first sent out one version and later sent out the other.

The logic of the game loader works like this: if the moves are different, or the result is different, it's regarded as a unique game and added to the database. Even if the players are the same, the round is the same, etc., we make a new game record.

<2) Why isn't the stale game deleted afterwards?> Perhaps you're thinking that the old version should be automatically removed, but that's a very dangerous idea. We can't assume that whatever copy of a game we're receiving today should replace the one in the database; it's entirely possible that the new version is the bogus one. You'd think that the second version is always a correction of the first, but it's not unheard of to see a good game suddenly become mangled. If I understand your narrative, that's what happened in this case, no?

You commented:
<First, I don't understand the significance of 27 moves. Yes, the game was 27 moves long (both versions).

Anyhow, "post production" does take place--sometimes TWIC fixes up games and even helpfully adds tags. Sometimes we fix things up. <But most likely in this case, FIDE first sent out one version and later sent out the other.>>

As for the significance of the 27 moves, that's to contrast with a case like this:

<2014.08.02 (R1.67) 1-0 Hager Omer -- Jane Garwell (D38) 19

2014.08.02 (R1.67) 0-1 Hager Omer -- Jane Garwell (C45) 49>

<CG> has two versions again, but the different move numbers suggests something besides the result are different. Originally I thought that <CG> had downloaded the game before the round was over, so the 19-game was still unfinished.

But looking at the game in SCID I could search on the position, and found the 19-move <CG> actually matches this game:

<2014.08.02 (R1.13) 1-0 Alyssa Wang -- Mustafa Mansour Zienab (D38) 19>

So a mismatch in the move number seems to suggest a cross-up with another game.

(Looking at the <CG> gid number, the 19-move move game was loaded before the correct 49-move game)

* * * * *

Now, I've yet to see a <FIDE> result from this tournament change. Of course, I never downloaded a version until after the round was completed. But I got a couple of rounds "hot off the press", and the PGN was unchanged thereafter. Which means the errors were even frozen in place (e.g. the FIDE names during the first couple of rounds in the Open Section).

Not to be too contrary, but I remain unconvinced about the last statement in the quote - about the likely cause of the change being <FIDE>.

If that were the case <FIDE> would have to change a correct result into an incorrect result in the original <Chelsie Monica Sihite -- Darlin Rocio Villar Acevedo> game. That seems unlikely.

It's hard for me to know which of the two duplicates was uploaded by <CG> first without <CG>'s gid. I have to visit <CG>, searching one game at a time by hand, if I want to know for sure what the gid is.

Can I assume that when I do a PGN download the games loaded sequentially, by gid, for a tournament?

Here is another example:

< 2014.08.02 (R1.14) = Jovana Vojinovic -- Mathilde Chung (D00) 31

2014.08.02 (R1.14) 1-0 Jovana Vojinovic -- Mathilde Chung (D00) 31>

Here the draw seems wrong, and it has the lower gid.

* * * * *

Actually, I have about a dozen other mismatched <FIDE> vs <CG> game results - where for most of them <CG> has the correct result.

That leaves me wondering how <CG> results can to be different than <FIDE>'s? In these dozen games there aren't any earlier <CG> duplicates to match the different <FIDE> result.

I can only assume you switched to download the <TWIC> version of the games, or that you ran an engine to tweak the final results for those games where the final position is out of whack with the final result.

(But isn't that a dangerous practice? Maybe not to 1st order in this tournament - since FIDE provides the timing info in the PGN, and so time forfeits can be eliminated. That would only leave resignations in a winning position, a 2nd order effect)

I haven't checked <CG> vs <TWIC> or <FIDE> vs <TWIC> for the Women's but I will.

Some questions to finish (sorry for the length, but this stuff is rather involved):

Can I always use the gid to see which game was downloaded first?

Does <CG> do a "post-production" with an engine to flag games with an final position eval not matching the final result?

Also - does <CG> automatically flag duplicated games like the two set of pairs above for manual inspection?

* * * * *

Of course I ask all these question in order to learn more about the process. It does seem to me that the tournament should be cleaned up while it's still fresh. The sooner the better because most people who are interested will be downloading it the first moment its available. With some diligence, and the proper toolset, I think most of these inconsistencies could be caught before the end of the tournament.