Please put links to the official website and any other relevant links at the top of an event's forum page.

So for example, for the World Team Championship (2010), the links

Official website: http://wtcc2009.tsf.org.tr/,

Results: http://wtcc2009.tsf.org.tr/componen...,

Live: http://wtcc2009.tsf.org.tr/content/...

would be displayed.

It would be very nice to have all of this information displayed in one place. I always have to go over to http://www.chess.co.uk/twic to find these links; it would be nice if they were easily available here.

http://www.coruschess.com/index.php

A few people have already commented on the [C05]/[C06] situation in the Tarrasch French. C06 seems to have become redundant over the past couple of years, with all relevant games since 2007 going into the more general category C05.

A similar situation exists with [C17], another line in the French. Officially it covers deviations in the Advance line of the Winawer, ie 1.e4 e6 2.d4 d5 3.Nc3 Bb4 4.e5 c5 [5.a3], and then anything other than the mainline 5...Bxc3+. There are a number of [C17] lines, but the most important is 5.a3 Ba5 (variously called the Retreat Variation, the Swiss Var., the Armenian, and the Swiss-Armenian or Swarm).

Recently, games in this line have been classified with the mainline Winawer as C18. As far as I know, certain engine/databases (such as Fritz) make the same mistake.

In both of these cases (and there are others apart from the French) the ECO system has one code for the 'general' variation, and another code for a specific longer line -- but all games are going into the 'general' box.

A related problem is in openings with many transpositions. For instance, games starting 1.d4 e6 may be called the Horwitz (or is it Harrwitz?!) Defense, ie just each player's first move. Yet the vast majority of such games go on to become a French (2.e4 d5), a Dutch (2.Nf3 f5), a Nimzo-Indian (2.c4 Nf6 3.Nc3 Bb4), a QGD (2.c4 d5), a Bogo-Indian, a Slav, an Owen's or English Defence (2...b6), a Kangaroo (2.c4 Bb4+) and so on.

Similarly, the 'Zukertort Opening' name is now used for many games starting 1.Nf3. These were previously given the routine name 'Reti Opening' - even when neither is strictly correct. One thing affected by these classifications is the 'most frequently used openings' table on the CG Statistics page, where the King's Indian Attack is placed much higher than it should be.

My suggestion is that, as far as practicable, the opening classification should be based on the position reached after several (5? 8? 10?) moves. This would mean that games starting 1.c4 or 1.Nf3, with early d4 by white, would be assigned to whatever Indian or Queen's Pawn game they transpose into. But it would not have to deal with those positions where, say, 13...Nc6 is a Tarrasch, 13...Nbd7 is a Catalan, 13...dxc4 is a QGA, etc etc. Such positions do exist, but precise classification is a matter for experts.

As for 1.Nf3, I find the following scheme useful: if white plays for c4, a Reti or English; if e4, a King's Indian Attack (or transposition to a 1.e4 game like the Sicilian); if white plays an early d4, it either transposes to a 1.d4 opening ... or remains as a Zukertort. Plus, of course, 2.b3 is a Nimzo-Larsen and 2.g3 can be a Benko/Hungarian, or a Neo-Gruenfeld, or a lot of other things not worth going into here.

One of the two methods was designed by Eric Schiller, called the Caxton notation. Unlike ECO it has never been adopted by any large database, but after studying it we think it holds great promise. We humbly question some of Dr. Schiller's naming conventions, but for the most part it is exactly the required technology. The Caxton database needs some corrections to a few of the positions but the general idea is correct.

There is no magic number of moves (5, 8, 10) which define an opening. A certain variation may only be definite after 14 moves (some lines in the Ruy Lopez or Sicilian fall into that category) -- while others may be defined by the first two moves (1.e4 a6). The trick is to start at the END of the game and work your way towards the beginning. If you come across a position in the database, that's the one you want, regardless of other "hits" that occur previously in the game.

I'd read that Eric Schiller was using some ideas from linguistics/grammar to create a new classification system - then called CaXML, I think? Is Caxton the same, and is there anywhere I can read more about it?

Working backwards sounds good -- until you remember certain middlegame positions that can be reached in completely different openings (IQP positions in Panov-Botvinnik Caro-Kann or QGD Tarrasch ... it's even possible to reach identical positions in the Petrov and the French Exchange). I guess in such cases you either give primacy to one system, or go on using both with some link between them to indicate a transposition ...

And I presume the system ignores identical endgame positions. Unless the endgame is reached directly from the opening in <20 moves?

One approach is to pay homage to whatever sequence of moves led up to the position, so that if it began 1.d4 d5 2.c4 we call it "Queen's Gambit" event though it achieved a well known Nimzo Indian position. There is something satisfying about this approach, but it may naively assuming that the data are so well organized that pigeon-holing the games into their openings like that is even possible. If there are only two of three ways of reaching a position it might work--but what if there are thousands of ways?

Another way is to take statistics: ask the database, which is the most common way of achieving the given position? (Yet another use for the Opening Explorer data!) With that method, we'll have games that start 1.d4 d5 2.c4 and are labeled "Nimzo Indian", and ones that start 1.d4 Nf6 and are called Queen's Gambit, and we'll also have a large body of data behind us to support that the designations are correct. What's more, this means that from time to time the openings will require computation for a second time, just to make sure that a change in an opening's popularity doesn't change how games are designated.

There are lots of games in the closed Spanish that include the moves ..h6, ..Re8, ..Bf8, etc. It becomes impossible to keep track of which variation is which.

I must say I don't like to classify games in openings different from the ones played.