When you like "In Flanders Fields", you may very well like "The Green Fields of France" (by the Fureys).

<Miss Scarlett>

On update, the <Olga will not display odds game properly> bug is fixed now.

So we should have no further shenanigans on uploading new odds games.

<Susan Freeman> Thank you, but that's all down to <Steve> who knows how the machines work eh?

I can function only as his loyal "bird dog."

====

Also User: plang submitted this interesting game with a note saying that it would make a "good game of the day."

P Tregubov vs B Lalic, 2000

I second the request by "Z" <realdumptrump>.

<Z realdumptrump> There might be millions of games here with illegal PGN.

I could copy/paste them all, auto correct them with my chess engine and re-install a legal pgn.

But it might take me 300 years.

Is it possible to write code that would clean them all automatically?

<Susan Freeman, Admins>

Please carefully consider <realdumptrump's> analysis of the perennial problem caused by <illegal pgns> in our database and the inability of <Olga> to play them.

###################

I said this to <realdumptrump>:

Bernhard Stark vs O Sarapu, 1987 (kibitz #6)

<<Z> Thank you for information.

What do you think the best solution is at the end of the day?

It seems to me that the <Upload Utility> and target pgn player <Olga> need to be synched. Right now they are not working well together.

If <Olga> is going to be strict on playing pgns with errors, then the <Upload Utility> needs to be given a strict gate keeping capability to keep out pgns with mistakes on them.

It's a disaster to have a strict <Olga> married to an <Upload Utility> with weak gate keeping capablity.

This is how we get pgns published that cannot be played by <Olga>.

It is terrible to have to manually repair pgns that are already published. I did hundreds of those repairs in the past, and it required tons of extra work for me and tons of extra work for <AnnieK> at the admin end.

In your idea of a loose setting, <Olga> would have the capability to autofix pgn errors that the <Upload Utility> let past the gate, right?

I like that idea. My chess engine autofixes pgn errors for example.

So when I submit a pgn, I always pass it through my chess engine first (paste in, then copy paste back out). <Olga> has had no trouble playing any of my pgns that have first been cleaned (auto corrected) by my chess engine. >

#######################

Then <realdumptrump> replied here, and *this* is the most important information in my opinion:

Bernhard Stark vs O Sarapu, 1987 (kibitz #7)

<Right, I logged back on for a quick reply... This can get in the weeds fast, but let's take a common concrete error - using 0-0 (number) vs O-O (letter) fot castling.

The CG DB should only have the correct version. The uloader can be on the lookout for this confusion, and if need be fix it (even without complaint - maybe).

If the DB were perfect Olga could be strict. But the DB isn't - so Olga should be loose, by default.

Probably the Olga programmer(s) weren't aware of the common PGN error, or that the DB had such PGN.>

###################

Please, please listen to <realdumptrump>.

Working with <Annie K> and <Chessdreamer> in the past, we had to "manually repair" hundreds upon hundreds of published illegal pgns at cg.com that <Olga> would not play. To be honest this was unpleasant, back breaking work, and there should have been no reason for it to have to be done in the first place. But it did have to be done and the problem has not gone away. It needs to be addressed at source level.

<CIO>'s game Olga-ends at 5.Nf3 because 5...de4 isn't parsed as a pawn capture (officially, it should be 5...dxe4).

Simple stuff, but what a pain-in-the-neck to patch one-game-at-a-time.

* * * * *

pgn4web is "loose", or permissive, about notation - a deliberate design choice. In fact, one might even consider it a bit too "liberal":

(From its website)

<<about ambiguous moves>

Ambiguous moves are errors in the PGN data where not enough information is provided to determine the intended unique move. For example in the following sequence the move 3. Nd2 is ambiguous because both the knights from b1 and f3 could move to d2:

1. d4 d5 2. Nf3 Nf6 3. Nd2

The ambiguous move error 3. Nd2 should be replaced by either 3. Nbd2 or 3. Nfd2.

When encountering an ambiguous move pgn4web first flags the error with an alert message; rather than stopping the game replay after the error, pgn4web then guesses which of the possible moves was intended and continues the replay; if the guess was correct, the game will replay in full; if the guess was not correct, it's possible that an invalid move alert will follow later and stop the game replay; it should be noted that the game continuation after an ambiguous move alert it's never fully reliable. >

https://sourceforge.net/p/pgn4web/w...

And we know how some on <CG> consider such "liberal" agendas.

What else?

If d8=N+ is in the scoresheet, then it should be (under)promoted to knight and a check. (Even possibly a double check... d8=N++)

click for larger view

Given how many hours <Chessdreamer> spent hunting and collecting inaccurate pgns that <Olga> could not play, and how many hours <Annie K> and I spent repairing and republishing those pgns, I would say that solving these problems is of the highest priority.

Should we just hope <Chessdreamer> does all of this exhaustive work again for free? He even went through each pgn and published the errors, work that made it possible to fix those faulty pgns in the first place.

Or should we solve the problem at source- make code so that the combination of the <PGN Upload Utility> and <Olga> can work together without producing hundreds- or is it thousands? of games that <Olga> chokes on?

The nightmare returning:

<Olga chokes on it>

L McLaren vs P W Stuart, 1987

-------------

<Olga chokes on it>

I forget off the top of my head what he called it - but Daniel had a pgnfix program which would periodically go through all the DB and fix up the PGN in some fashion.

<1.) The PGN parsing in OLGA 1 is handled by chess.js and OLGA 2 by a library called kokopu. >

<https://github.com/jhlywa/chess.js> <https://www.npmjs.com/package/kokopu>

Improvements were made to Olga 1.0 by upgrading to a newer version of Chess.js. Still possible to upgrade again in the future.

Kokopu seems pretty solid.

< 2.) In all cases, the PGN should be accurate. >

I suggest we build a PGN submission tool using OLGA 2. All PGN submitted through it would be valid for all viewers (assuming the viewer can handle the game, like Chess960 variants).

As for the old games with errors, we should just have a script run over them and identify them. Auto fixing, them is tough.

If the move is ambiguous like stated above, yes OLGA could technically deduce what the correct move is and then continue the game moving from that point. This is a ton of extra code, and overhead. If we could not ensure that the PGN was correct when OLGA gets it, yes I would have added support for errors. But OLGA was <designed> to be clean and small. It reduces the size of OLGA if it can just assume the PGN is "good". When I first built OLGA, we found a lot of games that had bad PGN because chess.js did not handle ambiguous moves.

So I believe the logic that is involved with serving OLGA the games needs to be improved. And FOR SURE CG should not allow any games to be submitted with errors.

<3. (Extra change) I have a good strategy on how to move forward but it's a lot of work. >

Just so everyone knows - My idea for fixing the games that are broke is to make an area where anyone from CG.com could go to - "Broken Games Zone". There you could see the giant list of games needing work, you choose a game, you fix it (with the help of OLGA 2 submission tool), and then you "Submit the fix". OLGA would have already verified that it is at least a valid PGN.

Then the "Fix Submission" would go out somewhere public (front page of something) and users could "Vote" on fixes. After a fix gets an "Validated By CG User" vote, it's sent to the final gate, a simple "Submit, or Comment" that will allow the users with elevated privileges to add the game into the database (or comment what about the perceived issue).

I have a ton of ideas about how to improve kibitz (heart and soul of CG) as well. :)

< CG is in a bit of an awkward phase at the moment. OLGA 2 alone cannot fix the issues. If you care about CG and want to support its continued existence, Please contact Susan about your interest to help in any way you can! >

There is a starting FEN which with Steve's improvements to Olga 1, should now work. And it has an ambiguous move (maybe multiple). I do not believe Olga 1 (chess.js) will ever support.

Go to The Bomb to watch some top ranked speed chess.