A difficult endgame: Fabiano Caruana vs Maxime Vachier-Lagrave

by Karsten Müller
4/20/2021 – Fabiano Caruana had a lot of time to prepare for his game against Maxime Vachier-Lagrave in round 8 of the Candidates, and he used it to find a new and dangerous idea in the Poisoned Pawn variation of the Najdorf, a favourite of the French player. "MVL" survived the opening and escaped into an endgame that objectively was a draw but difficult to defend. Karsten Müller took a closer look.

ChessBase 16 - Mega package Edition 2021 ChessBase 16 - Mega package Edition 2021

Your key to fresh ideas, precise analyses and targeted training!
Everyone uses ChessBase, from the World Champion to the amateur next door. It is the program of choice for anyone who loves the game and wants to know more about it. Start your personal success story with ChessBase and enjoy the game even more.


Karsten Müller, born 1970, has a world-wide reputation as one of the greatest endgame experts. He has, together with Frank Lamprecht, written a book on the subject: “Fundamental Chess Endgames” in addition to other contributions such as his column on the website ChessCafe as well as in ChessBase Magazine. Müller's ChessBase-DVDs about endgames in Fritztrainer-Format are bestsellers. The PhD in mathematics lives in Hamburg, where he has also been hunting down points for the HSK in the Bundesliga for many years.
Discussion and Feedback Join the public discussion or submit your feedback to the editors


Rules for reader comments


Not registered yet? Register

malfa malfa 4/24/2021 12:50
@ Frits Fritschy

The ECE reports two positions from Arab codes, one dates from 1140, the other and more famous from 1257: the first postion is Kb5, Rh1 vs Ka7, Nb7; the second is Kc6, Rh8 vs Ka7, Nb7, which a very old book of mine attributes to Al-Adli, so I suppose you are referring to the latter. Again please note that in both these positions the king of the inferior side is on the rim, which as I wrote makes a huge difference from the situation when the king is on b6 or c7 as in the game: in the latter case it actively helps building the fortress.
Frits Fritschy Frits Fritschy 4/23/2021 11:22
Without pawns, the first endgame with the 'fianchettoed' knight I think of is the famous Al Adli position, named after the 9th century Arab player who rightly thought it was a win for the rook. With pawns on adjacent files, things don't change when you move the game position one or two files to the left. Move them three files to the left, then white wins. Move everything one file to the right and things get tricky. It's fun to find out what difference this makes, going through the tablebases.
malfa malfa 4/23/2021 08:52
@ fgkdjlkag

Throughout this discussion I have been referring to the simpler endgame of pure R vs N, without pawns.
Regarding what to look for, if you search this pattern in the Encyclopedia of Endgames you will find that there a lot of lost positions with the fianchettoed knight, but in all those cases the weaker player's king is confined on one side of the board. My memory was that there are also samples with a more active king, but it was a wrong one. I am sure I have learned this setup elsewhere ages ago, but I do not remember where else.

However it is sufficient to look on a big database all the R vs N positions with Ng7 and Kf7 or Kg6 (and of course their symmetrical ones) to find find plenty of them and almost invariably drawn, so I suppose that generally the players who were struggling to equalize more or less knew what they were doing, didn't they? And why? I think mainly because I am not the only one to know that in that kind of endgame it is a standard method of defence, though of course not unique: it is simple, hardly prone to errors, and effective.

When it comes to the endgame of Caruana-MVL, with one pawn for each color, anyone may develop his own idea of why Black was unable to make use of the above setup in a more complex situation: miscalculation? ignorance of basic patterns? Time pressure? Stress? Whatever it was, I notice that, based on what I read in this discussion, apparently there is no consensus regarding MVL's clock situation in the critical phase of the endgame.
fgkdjlkag fgkdjlkag 4/23/2021 06:44
@malfa, from the press conference it sounds like neither Caruana nor MVL were aware of the ideas that you describe. I looked at some positions with pawns on adjacent files with rook vs. knight, and I do not find any ideas in common from one position to the next. It looks like the right locations for the pieces have to be worked out depending on the specific position. I am not sure where you are getting the idea that there was some knowledge that would have made it obvious? Since we know now that the knight was supposed to be on g7 of course it is easy to see in retrospect, but not when in that position for the first time, and MVL was low on time.
malfa malfa 4/23/2021 01:37
@ Frits Fritschy,

LOL! Great movie, incidentally...
Of course all this stuff is pure speculation, yet the obscure mechanics of mistakes is one of the reasons why chess between humans, in all its imperfection, is such a terrible and fascinating game.
Frits Fritschy Frits Fritschy 4/23/2021 12:47
Your answer reminds me of the movie 'being John Malkovich'. We indeed should maybe not venture too far...
malfa malfa 4/23/2021 01:46
@ Frits Fritschy,

welcome to the club of those who would desperately want to know what is in MVL's head ;-)
Or maybe not, as this would imply a direct contact with "the French school of suffering"...
For what is worth, given also the safe clock situation, the picture I have figured out to myself portraits a typical case when too much knowledge becomes an obstacle to see the obvious :-)
Frits Fritschy Frits Fritschy 4/22/2021 11:32
To me, winning the won positions in this endgame seems much more difficult than drawing the drawn positions. Caruana had to find a string of subtle moves and that he faltered just once is quite amazing. There can be many reasons Vachier-Lagrave couldn't find the draw: nerves, tiredness, tunnel vision, loss of concentration because he knew it was a theoretical draw. Not time trouble by the way; on move 55 he had 35 minutes left. But in endgames like this it's natural to look for fortresses. (See for instance the famous endgame Emanuel vs Edward Lasker, New York 1924) It is not difficult to foresee that with the knight on g7, the fortress can only be broken by the white king going to f8. That's a long way from the pawn on g2. For a top GM (not for me, of course) it should be peanuts to find that you even can sac the knight then. An average IM could of course lose this as well, but drawing it along these lines should be possible at that level too. Even without a thorough knowledge of the handbooks.
malfa malfa 4/22/2021 05:58
Anyway I have checked the ECE, and I found that, when invoking it as a reference, I was wrong: in the only examples where a fianchettoed knight appears, the inferior player always loses, because in those positions his king is invariably confined to one side of the board, therefore it is of no help with building the fortress!

But then this might give a clue to explaining MVL's misfortune: not that he was not familiar with the basics of the R vs N endgame, it was just the opposite (no wonder, I should add)! During the game he might have recalled positions like the ones above as patterns of what was not to be done, and somehow sticked to them no matter that his king was in fact more actively placed.
malfa malfa 4/22/2021 04:02

definitely I didn't want to criticise so much a stronger player than me, rather to express my amazement that he did not automatically put the knight in the right place. I would say that stress and high stakes into play took their toll, of course, were it not for the circumstance that if there is a thing I know about that kind of endgame is the practical guideline I mentioned, whereas conversely even during the press conference neither Fabi nor MVL ever gave the impression of knowing it. No personal analytical skills at stake, just a little fragment of classical knowledge, though one which I am pretty sure is well represented also in the (OK, not so) modern Encyclopedia of Chess Endgames. Surely I am just an old fox that for once knew one bit more of two calculating machines. They are perhaps too young to have learnt from vintage manuals, however that this happened precisely to MVL, and not another chess star, doubles my surprise, because as far as I know he is regarded as a renowned endgame specialist.
sshivaji sshivaji 4/21/2021 07:06
@malfa, You are a well-known endgame analyst who has analyzed and submitted corrections to many endgames. Well done! This sort of thing is obvious with your analytical work, but less obvious for players who are preparing many lines usually not related to the endgame. I would not be too hard on MVL, for an FM such as myself who does not extensively analyze endgames, the fianchetto of the knight is not something I recall in this endgame.
psamant psamant 4/21/2021 06:51
Man... this was some game! A real tough one for black to hold on, even inn the middle game phase. And the endgame was just too complex, with so many only moves to be found!
malfa malfa 4/20/2021 06:40
What puzzles me the most in this endgame is how a SuperGM of MVL's caliber did not recall one of the basic concepts in the pure R vs N endgame, i.e. the recommendation to "fianchetto" the knight as an optimal defensive setup. This holds all the more with one pawn each like here, since the f6 pawn is an enforcement of the fortress while White's 'g' pawn is pretty unusable as a battering-ram.