ForumHelpSearchMy ProfileSite InfoGuests InfoRepertoireLinks
I am pretty familiar with the theory of the French Defence, so I'm continually amazed by the volume of fresh ideas being generated in this opening, and perhaps even more so by the fact that most of its variations never seem to be truly resolved. That is, players are discovering new resources in even relatively ancient lines, and when something is apparently 'solved', that status sooner or later changes with a new idea. So take a look at the French if you're feeling limited by too many forcing lines in your repertoire.

I noticed a lively discussion in the Forum about whether the French is 'correct' at higher levels of play, which included the more specific question of whether it was an inferior choice for players above 2600 or even above 2500 (the thread is entitled ' French BAD for + 2600 and less bad for -2500'). There are many good contributions, but 'Dean' in particular laid out some statistics which showed a better-than-normal success rate for White in the 2700+ and 2600+ brackets, with a better-than-normal success rate for Black in the categories below that. Performance ratings were not included. I decided to collect my own stats, using Megabase (results from TWIC were extremely similar).

The first statistic that popped up is bad news for the French: With both opponents above 2700, the French scored 61%-39% for White (with a 44% drawing ratio, 11% less than the figure for all games), whereas 2700+ games in all openings had a 56%-44% ratio. Although this basically means that in a 20 games tournament Black would be only a point behind using the French instead of a generic opening, that is still quite significant, and the performance rating advantage for White is greater than White's normal advantage in the 2700+ database by 60 points, which is certainly nontrivial. The sample space was rather small, however, and this led me to wonder about the breakdown of these games. I have noticed for years that the elite players, who are frequently involved in closed events, have had a tendency to use the 3...dxe4 versions of the French to try to draw as Black (and taking short draws). In fact, no less than 30% of these 2700+ games are in those variations, as opposed to merely 8% of 3...dxe4 Frenches in Megabase as a whole. And the results for those variations were poor, as with other passive lines; whereas a more dynamic line, the Winawer Variation, had a 54%-46% ratio and 42-point rating spread, both of which are actually better for Black than for all the games in this 2700+ database. I've noticed before that when these top players use an opening to draw as Black (like the 3...dxe4 systems), they indeed get more draws, but have a mediocre or worse performance. The Petroff Defence, for example, draws at a spectacular rate, but performs a bit worse for Black than the average defense at 2700+ levels (and worse than that at lower levels). You'd think a high percentage of draws would help Black's performance rating; but I think that 2700+ players using the Petroff tend to accept the draw when offered, even if they have a small advantage, whereas with White they are more inclined to play for a win. Similarly, players at 2700+ may need to employ the French with more ambition in order to improve its performance - in my opinion, it isn't a very good drawing weapon.

The news gets much better for Black in the 2600-2699 region, with of course a much larger sample space. Now using the French results in a 57%-43% spread, one point worse than in the 2600 database as a whole (equivalent to 1 more loss in 100 games), and because of the rating differential the performance rating spread is only 9 points lower than the average of other openings (having the Sicilian Defence in the pool accounts for more than this gap.) In fact, this performance is as good as or better than other 1 e4 defences. Curiously, using the example above, the Petroff does poorly at this level with 59%-41% and a 41 point performance rating deficit below normal, even with a very large sample of games. I suspect that the same people questioning the 'correctness' of the French wouldn't assess the Petroff in the same way.

Finally, for all games in Megabase, the French produces nearly identical results as the database as a whole, and does better than most individual 1 e4 openings (because the Sicilian at 51%-49%, and taking up over 20% of all chess games played, makes the others look bad!). That includes, for example, the traditional Ruy Lopez after 3...a6 4 Ba4, considered about as sound as it gets. I should also say that, as was pointed out in the Forum, the French does as well or better than major 1 d4 defences (usually about the same).

My conclusions: The French has indeed performed worse than the average opening at the 2700+ level; I think that's at least partially explicable by a skewed choice of variations (within a very small sample space), and not too worrisome. On the positive side, it does well in all other ranges (up to 2700). At any rate, I think it's fair to call the French 'correct'; if it isn't, then arguably nothing besides the Sicilian Defence is! Still, I haven't put a great deal of time into this investigation, so you may want to generate your own figures, use other criteria, or challenge the assumptions. Forum material, perhaps?

Download PGN of February '10 French games

Advance Variation

Adams took a break from his recent adventures in the Tarrasch defence to try out the Advance Variation. Black played an early ...Rc8, to which White responded with a3 (unnecessarily, I think) and the game entered normal channels.

Black has the usual light square pressure in return for space, and is even poised to play ...g5-g4. In Adams - Polzin, Bundesliga 2009/10, he probably could have achieved some advantage, but made a crucial inaccuracy and was ground down in a cramped position.


Against the 3...c5 Tarrasch, White often employs 4 Ngf3 to avoid the 4 exd5 Qxd5 lines. This can make it difficult for Black to generate active play, but White's punishment is that his play is limited as well:

In McShane - Ni Hua, London Chess Classic 2009, I've shown some standard Black solutions to 4 Ngf3 using a series of recent games. These lines can be rather dull, which should please neither side. That's the way chess sometimes goes, and at least the positions remain imbalanced.

Winawer Variation

It seems that every line of the Poisoned Pawn Winawer could use its own column (and in some cases its own pamphlet!). White's attempt to expand with 14 h3 (and earlier h3s, transposing or not) constitutes a potentially serious danger to the variation. It plans g4, which expands on the kingside while both protecting g2 and denying f5 to Black's knight. Reader Franz Steenbekkers specifically asked about this line before last month's update, and we have an example from this month through which to address it:

In Khachiyan - Sevillano, Los Angeles 2010, I've included an array of possible continuations with dense analysis, largely original. The game itself is a real slugfest, and Black's options after 14 h3 are equally fascinating. One of the most important options has barely been seen or analysed; and the other most promising move, to my knowledge, isn't mentioned anywhere!

The Winawer 'Armenian Variation', is thought to be in theoretical trouble, but there are still some new ideas in the crazy main line:

In Amonatov - Gorovykh, Moscow 2010, Black holds his own and should have won easily versus a much higher-rated opponent. In general you feel that White should survive these speculative attacks and come out ahead, but that's not completely clear when you look at the specifics, and at any rate it's a difficult task for both sides over the board. This variation is well represented in the Archives, by the way.

Classical/Steinitz Variation

The Classical with f4/Nf3/ Be3 has arguably been the most topical line in the 3 Nc3 French Defence, and the theory expands with each passing month. Because the two sides' forces don't always come into direct conflict, I expect that trend to continue and new positional ideas to arise.

In Karjakin - Carlsen, Wijk aan Zee 2010 (apparently Carlsen's first French in a regular time-control game), White played 11 Kh1 here. Most of my presentation deals with 11 a3 (including potentially important new games) in a lengthy note. The game itself is positionally instructive, in that neither side tries the standard manoeuvres.

Another Classical Variation arose in Quesada Perez- Nogueiras Santiago, Ciego de Avila (Cuban Ch) 2010.

White went queenside and after Black inaccuracies achieved a central advantage; but castling kingside is probably the best theoretical continuation.


These days the most dynamic opportunity for Black in the ...dxe4 lines is the Burn Variation with ...gxf6, normally using Morozevich's favourite ...a6.

In Naiditsch - von Herman, Bundesliga 2009-10, Black went into the variation with 7 Nf3 f5 8 Nc3 a6, called the 'neo-Morozevich ', as opposed to the original 7...a6 (which I discuss in the notes). In New in Chess Yearbook #90, Larry Kaufman has an article about the variation with 7...f5, and I've referred to it as well as earlier theory. My conclusion is that White retains the advantage after 7...f5, precisely in the line Naiditsch employs. The game itself is a slaughter.

Exchange Variation

Finally, the game Roitzsch - Graf, Finsterbergen 2010, shows one way in which a strong grandmaster handles the Exchange Variation. I've looked very briefly at a couple of variations with c4, which occurs in this game, and included some analysis by a Forum contributor.

This is a standard position which I feel is equal. Black outplays his opponent in instructive fashion. In a future column I hope to get to the broader subject of c4 in the Exchange Variation, since it's recently been addressed in a few articles.

Till next month, John

Please post you queries on the French Forum, or subscribers can write to me at if you have any questions or queries.