Candidates Tournament 2018 : Analyses

Prior to the tournament, I gathered a database of games from the recent super-tournaments, from Tata Steel 2016 to Tata Steel 2018. This sample of 1407 games at classical time control will help me generate a few statistics about openings at the highest level, in order to appreciate the opening choices during the Candidates Tournament with some perspective.

The statistics will follow the following format :
opening, number of games, score for White, percentage of draws, average Elo (color), success index

The success index is the difference between performance and average Elo. The color for which average Elo is computed is the one making the last move in the definition of the opening. Transpositions are taken into account (e.g. 1.c4 e6 2.Nf3 d5 3.d4 Nf6 4.Nc3 c6 is classified as a Semi-Slav, not as an English opening).

Reference points :
all, 1407, 55%, 66% draw, 2750 W, +36
1.e4, 566, 56%, 65% draw, 2753 W, +34
1.d4, 582, 55%, 69% draw, 2747 W, +38
1.c4, 165, 55%, 66% draw, 2751 W, +37
1.Nf3, 84, 57% 59% draw, 2754 W, +24

************************************************
* The first game analyzed is Kramnik-Grischuk from round 1 *
************************************************

It is a very instructive game about the Reti opening, a model to be compared with the historical examples by Reti himself. It is also representative of the contemporary trend in chess, where an opening is not necessarily expected to give any advantage but is rather aimed at creating opportunities (and chances for the opponent to go wrong) during the middlegame. The analysis is found in the following study :

*************************
* lichess.org/study/qJOLmzLp *
*************************

Although it actually starts with 1.d4, it transposes into a flank opening. I will therefore give the statistics for flank openings today (others will follow with other games).

1.Nf3/g3, 48, 55%, 70% draw, 2764 W, +0
1.c4 c5 2.Nf3, 53, 50%, 74% draw, 2743 W, +27
1.c4 e5, 58, 58%, 65% draw, 2751 B, -53
1.c4 e5 2.Nc3 Nf6, 33, 55%, 61% draw, 2765 B, -37

There are two variations that deserve a special mention, one because it scores unusually badly for white and the other because it's the most frequent variation in all flank openings during the recent super-tournaments. The pure symmetrical English 1.c4 c5 2.Nc3 Nc6 3.g3 g6 4.Bg2 Bg7 5.Nf3 doesn't work at all for White, as it scores 2,5/9, and Black's performance reaches 2902 there ! The Rubinstein variation (1.c4 c5 2.Nf3 Nf6 3.Nc3 d5 4.cxd5 Nxd5 or 3.g3 Nc6 4.Bg2 d5 5.cxd5 Nxd5 6.Nc3), championed by Anand, Harikrishna, Svidler and Vachier-Lagrave with Black, scores +4=16-2 for White (Aronian and Svidler being the expert on this side), with normal success (+45) but a high frequence of draws (16/22).

More games and related statistics will follow soon !

interesting stuff!

Today I could analyze a second game :

*******************************
* Karjakin-Mamedyarov from round 1. *
*******************************


Please consider clicking on the heart below the game score in the study :) .

It is a Spanish Opening. The players know each other vey well, Mamedyarov was part of Karjakin's team for the World Championship 2016. 3...g6 was not a real surprise, but 5.Fxc6 has not been seen in a super-tournament. The opening is very interesting, I think that we will see 10...Ne7 and 11...Bxd4 again, but I'm curious to watch White's preparation after 12.Rd1. If 3...g6 is sufficient for equality after 12 moves against a 2770-player, why should we bother to play anything else against the Spanish ?
Minor pieces are exchanged quickly and the game turns into a lesson in queen endgames. It is extremly instructive because this topic is one of the most obscure in chess and the events in the game shed some significant light on winning procedures and drawing chances.


As an introduction, I will post the statistics for Open Games in recent super-tournaments.

The reference is 1.e4, 566, 56%, 65% draws, 2753 W, +34.
e5, 353, 56%, 69% draws, 2761 B, -51
Spanish, 172, 58%, 73% draws, 2764 W, +56
Italian, 108, 52%, 63% draws, 2762 W, +15
Petroff, 45, 60%, 76% draws, 2748 B, -65

Time for a third game :

*******************************
* Mamedyarov-Aronian from round 2. *
*******************************

link to the study : lichess.org/study/qJOLmzLp

In this game, the opening stage is very instructive. Both players toy with the opponent's preparations. The influence of Mamedyarov's second, Alexey Dreev, is felt, not for the last time in this tournament.

This is a Nimzo-Indian, so as an introduction, I will post the statistics for Indian games in recent super-tournaments.
reference: 1.d4, 582, 55%, 69% draws, 2747 W, +38.

Grunfeld, 59, 47%, 75% draws, 2758 B, -7
Nimzo, 49, 60%, 68% draws, 2742 B, -82
QID, 42, 58%, 46% draws, 2737 B, -55

The Grunfeld scores surprisingly well for Black, even if it is very drawish. Only the exchange variation with 7.Nf3 score 4.5/7 for White.
The Nimzo-Indian performs poorly nowadays for Black, due to the success of 4.Nf3 for White (9.5/13), especially a score of 6/8 against 4...b6 (the "Hybrid Indian", half-Nimzo half-QID, often played by Adams).
The QID itself is significantly less drawish than average but performs normally. The Petrosian variation (4.a3) is rare but scores better than the others (5.5/8).

The next game that is analyzed is :

*******************************
* Caruana-Mamedyarov from round 3 *
*******************************

link to the study : lichess.org/study/qJOLmzLp
Please consider clicking on the heart below the game score in the study :) .

Despite the frequency of 1.e4 e5 among top players, the Najdorf Sicilian is still a trendy variation. Here Caruana knowingly enters a variation in which he has an awful score (0/4 in recent years against top players), probably because Mamedyarov's choice in this line used to be predictible (6.f3 Qb6), but as in the previous rounds, Mamedyarov shows that he prepared fresh ideas for this Candidate tournament (the main line 6.f3 e5). Both players seemed to assume during the press conference that Black is better, but that is open to debate.

As an introduction to the game, I will show the statistics for Semi-Open games in recent super-tournaments.
reference : 1.e4, 566 games, 56%, 65% draws, 2753 W, +34

c5, 144, 56%, 57% draws, 2747 B, -43
c6, 37, 50%, 58% draws, 2709 B, +34
e6, 25, 50% 60% draws, 2700 B, +26
Caro-Kann advance, 18, 47%, 51% draws, 2744 W, -63
Sicilian Najdorf, 68, 57%, 53% draws, 2756 B, -64
Sicilian Bb5(+), 21, 55%, 63% draws, 2746 W, +28

It is quite surprising that, in super-tournaments, the Caro-Kann and the French perform so well for Black (a positive success index for Black is quite remarkable on move 1). The Advance variation of the Caro-Kann has the worst success index of the whole database : with it, White doesn't even score 50% against slightly lower-rated opponents. The Najdorf is struggling, on the contrary, with fewer draws than usual and simply more losses for Black.

As today is a rest day, I had enough time to analyze one more game :

***************************
* Karjakin-Aronian from round 4. *
***************************

link to the study : lichess.org/study/qJOLmzLp
Please consider clicking on the heart below the game score in the study :) .

The Vienna variation of the QGD is one of Aronian's specialty. Karjakin enters it willingly and traps himself with an original move order. Aronian admits that he was not very precise in converting his extra pawn, but Karjakin is clearly demoralized and doesn't take two real fighting chances : 19.Kg1 (not mentioned by the players) and 36.Rd2 (mentionned by Aronian, but Karjakin still believes that it is loosing). The rook endgame with three extra pawns on the same side of the board is instructive even for beginners.

As an introduction to the game, here are the statistics for Closed Games in recent super-tournaments.
reference : 1.d4, 582 games, 55%, 69% draws, 2747 W, +38

QGD Nf3, 122, 56%, 75% draws, 2764 B, -49
QGD (no Nf3), 20, 68%, 55% draws, 2711 B, -103
Slav & Semi-Slav, 73, 56%, 68% draws, 2751 B, -59
Catalan, 69, 54%, 65% draws, 2741 W, +31
without c4, 44, 60%, 71% draws, 2778 W, +29

I don't understand why top players nowadays still allow the QGD without Nf3 when they play with Black. Hou Yifan's specialty 4.cxd5 Nxd5 scores 6/9 for White (not counting Kramnik's defeat with it in the present tournament) and 4.cxd5 exd5 is even a bit worse than that for Black. The whole line has the worst success index for Black in the whole database and it can easily be avoided (3...Be7 instead of 3...Nf6). It's less drawish simply because Black looses more often.
The QGD with Nf3 would be even more drawish without the semi-Tarrasch where White scores 8/13 (61% is no big deal though). The Bf4, Ragozin and deferred exchange variations of the QGD happen to be very drawish in super-tournaments.
In the Catalan, the line 6...dxc4 7.Qc2 a6 8.a4 scores a slightly unusual 10/16 for White. Everything else looks like the reference.

All the statistics are now covered (Open, Semi-open, Closed, Semi-Closed and Flank openings).

I've added two more games to the study :

****************************
* Caruana-Karjakin from round 5.. *
* Caruana-Grischuk from round 6. *
****************************

link to the study : lichess.org/study/qJOLmzLp

There was no point for me to analyze Aronian-Kramnik, Kramnik-Caruana and So-Aronian that have already been analyzed by Peter Svidler. What can I add to the comments of a player who is (1) as strong as the Candidates themselves, (2) almost impartial (not a player's coach or second) and (3) willing to share his comments freely ?

Caruana - Karjakin is a small wonder of opening preparation. Caruana thinks that he's playing a surprising move (Qb3), Karjakin follows a line that he hasn't played before (on move6) and uncorks a novelty that wraps up the game into a single strategic concept, from move 10 to move 28. Reaching a dead draw as Black against Caruana without the slightest effort is a very nice achievement.

Caruana-Grischuk is also an interesting opening battle, in a variation in which Grischuk is an expert... with White ! The evaluation of the position is very complicated all along, until the final repetition. The tactical shots are not very hard to calculate, but assessing the various lines and Black's compensation after a pawn sacrifice is a topic for a big debate. This debate actually happened during the press conference and both players were forthcoming and friendly to each other, often making suggestions stressing the resources of the opponent's side. We will see more about this (very) unclear line in the future, for sure.

I don't have any new statistics left to introduce the games, but I can draw a few lessons about the first half of the tournament from my analyses.
(1) Three players haven't lost a game yet : Caruana, Mamedyarov and Ding Liren. Only one player hasn't won a game yet : Ding Liren. Everyone else has won and lost during the first half of the tournament. The elimination process, usual in old-fashioned Candidates Tournaments, has already been taking place.
(2) Caruana systematically downplays his chances during the press conferences, insisting on the fact that he was "lucky" in several games ; his post-game analysis does not show any ambition or self-confidence. It sounds as if it were important for him to be underestimated by his opponent. Now, his clear first position after 7 rounds will have the opposite effect. Handling the pressure later in the tournament seems to be an issue he was aware of from the start.
(3) Mamedyarov appears to be very well-prepared and in very good shape (i.e. very lucid and not too tired after the games). He catches his opponents in lines that are partially new to him as well, he seems to anticipate his opponent's preparation and to deviate accordingly. Even half a point behind Caruana after round 7, his chances appear extremely good.
(4) Ding Liren accumulates experience and doesn't try to turn his outsider position into a kingmaker role. The second half of the tournament can go either way for him.
(5) 1.d4 is played more often than 1.e4, and 1...e5 was played in all the 1.e4 games but one. In top-level practice, 1.e4 has become the quieter first move, and everything else is "wilder". The frequency of flank openings is above average, but it has probably more to do with the identity of the participants than with an objective trend.
(6) We have seen quite a large number of games where opening preparation turned out to be determinant. The idea that openings are less crucial nowadays than during the Kasparov era does not square with the facts during this Candidates Tournament.

This is an amazingly interesting tournament, indeed. The players are fighting toe to toe, and not holding back their punches. After round five, I thought that Kramnik would be a great title contender but now, he is lagging a bit behind. Caruana would certainly pose a challenge to Carlsen, if he wins.

Thanks for your Analysis A-Cielbleu

Two more games were added :

*****************************
* Aronian-Ding Liren from round 1. *
* Caruana-Ding Liren from round 9 *
*****************************

link to the study : lichess.org/study/qJOLmzLp

I had planned to analyze Karjakin-Kramnik from round 9, but Peter Svidler decided that the game was worth a video of its own.

The Aronian-Ding Liren is a flash back justified by three facts : (1) the Flohr-Mikenas attack is very trendy among top players ; (2) I had yet to analyze a game of Ding Liren's ; (3) the material imbalance makes it very instructive.

The Catalan is the most popular opening during this Candidate Tournament. I decided to analyze Caruana-Ding Liren because the opening was closely related to Caruana-So from round 1. If Caruana doesn't win the tournament, his missed opportunities against Ding Liren will be pointed out afterwards as the main cause of his failure.

With the new tournament situation, this game becomes very relevant.

******************************
* Ding Liren-Grischuk from round 11. *
******************************

It has been analyzed as the 10th chapter of the study : lichess.org/study/qJOLmzLp

Openings may seem dull and repetitive in this tournament, but it is an optical illusion. Actually a lot is happening in the opening choices and you can see it as soon as you dig beneath the surface. This game is no exception, as Ding Liren doesn't force a Catalan and drives the game to a Semi-Tarrasch, then a century-old variation of QGA. Grischuk knows this line, as he played it in the early 2000's but that was a long time ago...
Without any experience, Grischuk would presumably not have gone for Na5, a move that works well only for specific reasons. Grischuk didn't remember the follow up and ended up in a loosing position. Ding Liren missed five wins in ten moves.
But that's not all : we also get a free course in typical endgames with opposite color bishops and other pieces, and also an illustration of the drawing method with a knight against an advanced g-pawn.
It's probably the richest game of the tournament in terms of instructional value. But there are more games to come !