The Winning Academy 19: Typical Mistakes when Analysing with an Engine

by Jan Markos
7/3/2023 – They are so much better than us. They solve the most complicated positions in seconds.And they are so available. Anyone can buy them and use them at home. And frankly, we all use them. Chess engines. These artificial beasts are very useful when used wisely. But frankly, most club players tend to rely on them too much and use them mindlessly. In the following article I would like to show you three typical mistakes you should avoid when using a chess engine. And to show that even the best sometimes fail to tame the engines, I have chosen the examples from games played by Magnus Carlsen and his super-GM opponents. | Graphic: Europe Echecs

Let us learn together how to find the best spot for the queen in the early middlegame, how to navigate this piece around the board, how to time the queen attack, how to decide whether to exchange it or not, and much more!

When preparing an opening novelty with the help of an engine, you should always remember that your unprepared opponent will analyse his position with his own brain and will therefore prefer solutions that are natural for a human being, not for a silicon monster.

Therefore, always check the ideas that are natural to a human player, even if they are not among the first lines of your engine. If you do not, you will be caught off guard and surprised by your own novelty.

A typical example:

Carlsen-Giri, Carlsen Invitational 2020

White to move

For this rapid game, the World Champion had prepared a smart and complex novelty 13.Rg1!?, offering a piece sacrifice. Of course, Carlsen anticipated and analysed the direct 13…f4, and assumed that after 14.gxf4 gxf4 15.Bxf4 Bxf4 16.e3 Qe7 17.Qh5+ Kd8 18.Qxd5+ Bd6 19.c4 White has got good chances.

However, Giri was too smart to fall directly into a forcing line that his opponent checked at home with his engines. He played the normal 13…0-0. And he was surprised that Carlsen did not respond at once.

In his notes, Giri comments: “After this extremely natural move Magnus sank into thought. It is an unavoidable consequence of preparing with the computer that sometimes the most natural moves, if not entering the engine tab, are dismissed.”

And I think he is right. This is exactly what happened to Carlsen. After all, my engine does not list the castling among top 10 moves either, preferring various strange looking alternatives. However, the differences in evaluation are rather small, 13…0-0 is definitely not a bad move.

Although the World Champion reacted correctly with 14.h4, his psychological advantage was gone. Black won later after an interesting fight.

Here's the complete game:

Carlsen,M28630–1Giri,A2764
1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.Nc3 Bb4 5.cxd5 exd5 6.Bf4 Ne4 7.Rc1 Nc6 8.Nd2 g5 9.Be3 Nxc3 10.bxc3 Ba3 11.Rb1 f5 12.g3 Bd6 13.Rg1 0-0 14.h4 f4 15.gxf4 g4 16.Rb5 Ne7 17.f5 h5 18.Bg5 c6 19.Rb2 Rxf5 20.e4 Rxg5 21.hxg5 Ng6 22.e5 Bf8 23.Bd3 Nf4 24.Nf1 Qxg5 25.f3 Nxd3+ 26.Qxd3 Qc1+ 27.Kf2 Qxb2+ 28.Nd2 Bf5 29.Qxf5 Qxd2+ 30.Kg3 Qxc3 31.Kh4 Qxd4 32.Rg3 Bg7 33.f4 Rf8 0–1

A second typical mistake in opening preparation is to choose a line solely on the basis of the computer's evaluation. However, YOU will be playing the line at the board, not the computer. YOU will be the one making difficult decisions under pressure. Therefore, the question of whether YOU like the position and can handle it is just as important as the objective and abstract computer evaluation.

In the 2018 World Championship match, Caruana repeatedly tried to get an advantage against Carlsen in the 7.Nd5 line of the Sveshnikov Sicilian. The following pawn structure arose in several games of the match:

Caruana-Carlsen, World Championship match 2018, game 10, White to move:

Carlsen´s second Peter Heine Nielsen commented on this line: “While the engines indeed seem to give a preference for White, people tend to underestimate the human factor and that Black's position is easier to play, as he generally tries to aim for the opponent's king, while White has to balance positional gains on the queenside with also caring about the security of his own king.”

To put it shortly: for computers, the 7.Nd5 line is attractive and promising. For us mortals, it is extremely difficult to handle. Both was confirmed later in the game. After Black’s 23rd move, the following position arose:

Objectively, Caruana is winning. Any engine would take 24.Bxb5!, then find a couple of precise defensive moves and finally push the a-pawn, thus gaining a decisive material advantage.

For a mere mortal, however, it is extremely difficult to choose a move that deprives your king of an important defender. And in a World Championship match it is even more difficult. The danger of being mated in a few moves (and losing a game with the white pieces) is all too real.

Therefore, Caruana decide to pick a relatively safe option and played 24.g3. The game ended later in a draw.

Here's the complete game:

Caruana, Fabiano2832½–½Carlsen, Magnus2835
World-ch Carlsen-Caruana
London22.11.2018[Shankland,S]
The 10th game of the World Championship match was the most back and forth affair of any game thus far. There were plenty of games where one side did not make the most of their chances, but this has been the only one where I thought both players at some point had very real winning chances. 1.e4 c5 2.Nf3 Nc6 Magnus stays true to his Rossolimo/Sveshnikov repertoire for the 5th time in a row. Starting from game 5, I have expected him to pivot every single game, and been wrong every single time. Perhaps he will just play this the whole way through? 3.d4 Caruana eschews Bb5, which he chose in his first 3 White games. cxd4 4.Nxd4 Nf6 5.Nc3 e5 6.Ndb5 d6 7.Nd5 I was a bit surprised to see Caruana enter the same line as game 8. Even though he got an excellent position in that game, he had to figure Magnus would have come with improvements, and the line is supposed to be okay for Black. 7.Bg5 Leads to the mainline Sveshnikov. 7...Nxd5 8.exd5 Nb8 9.a4 Be7 10.Be2 0-0 11.0-0 Nd7 12.b4!? The first major deviation. Caruana chose Bd2 in the previous game. Still, the move does not look too impressive to me. 12.Bd2 Carlsen misplayed the position and quickly ended up worse. 12...a6 13.Na3 a5!? A very commital decision, but not a bad one. Magnus forces the queenside open right away, a fearless decision against a prepared opponent, but he had other options. 13...b6 Appeals the most to me. Black tries to keep the queenside closed and will look for his own counterplay with f7-f5. In fact, I actually prefer Black in a practical game here. 14.bxa5 Rxa5 15.Nc4 Ra8 16.Be3 Caruana was still playing very fast. He probably was in his preparation. f5! Black absolutely needs counterplay, and f5-f4 is coming. However, unlike game 8, here White cannot comfortabll play f2-f4 himself to keep his kingside secure. 17.a5 17.f4 Black is absolutely fine after exf4 18.Bxf4 Nb6 18...g5!? If Black is in a feisty mood 17...f4! 18.Bb6 Qe8 The battle lines are drawn. White clearly has an edge on the queenside, Black's pieces (particularly the c8-bishop and d7-knight) look a little clumsy, and the b6 square could prove to be a problem. But Black's kingside pawn mass promises him decent counterplay. 19.Ra3 A not so mysterious rook move. It looks pointless now but Black's plan is clearly to launch a kingside attack, and the rook will be a good defender laterally along the third rank. 19.Re1 This is the computer's recommendation by a wide margin, but I don't really understand the move? I guess it just serves as some vague prophylaxis against e5-e4. It claims white is better but I am unconvinced because after Rf6! Black is ready to challenge the b6-bishop via Be7-d8. 19...Qg6 20.Bc7 e4! Carlsen is in burn the bridges mode. His pawns are vulnerable and could quickly fall if the center were to open at an inopportune moment, but he is looking for direct counterplay against White's king, truly in the style of the Sveshnikov. 20...Ra6 A quiet defensive move like this one was surely fine, but I think the World Champion was right to go for it. 20...b5!? It was also possible to play in the same manner as the game, only without committing to e5-e4. This could make some sense as the central pawns proved weak. 21.Nb6 21.axb6? It's possible this move annoyed Magnus since e5-e4 no longer works, but it is also insufficient. Rxa3 22.Nxa3 The problem is here Black needs a new attacking plan since e5-e4 fails to bring its desired returns. Nc5! The b-pawn is well under control, and most importantly, White's pieces are totally unprepared to defend his king. How will the c7-bishop ever contribute to a proper defense? Rf6-h6 is coming and Black should win. 22...e4 23.Bh5! Qg5 24.Qg4! White is doing well. Getting his queen to g4 will really lesse any attacking chances Black can hope for 21...Nxb6 22.Bxb6 This leads to a similar position to the game, but Black does not need to play e5-e4 if he so desires. For instance, he seems to have a good position after b4 23.Rb3 Bf5! When it is clear there are reasons the pawn is good on e5 as well. 21.Kh1 b5!? A very commital move, but not a bad one. 21...Qh6 The machines prefer a slower buildup with something like this. Black certainly will have attacking chances by lifting the f8-rook, but I don't mind Magnus's decision either. 22.Nb6! 22.axb6? This would be asking for trouble. I find it unlikely White will survive after the energetic Rxa3 23.Nxa3 f3! 24.gxf3 Ne5! Black's attack looks devastating. 22...Nxb6 23.Bxb6 Qg5?! Technically, this move loses the game against best play, but it comes with a very nasty idea of playing Rf6-h6 and delivering mate on the h-file. A machine with its nerves of steel would have no trouble grabbing h5, but for a human, it looks absurdly dangerous. 24.g3 Caruana's move makes a lot of sense. Taking on f4 and bringing the rook to g3 should dispel any mating dreams. 24.Bxb5! White could have gotten away with this. Rf6 Extreme precision is needed to beat back the attack, but it is possible. 25.Re1! An important move. White hits the e4-pawn while simultaneously clearing f1 for the bishop. Bf5 25...Rh6 26.Rxe4 25...Rg6 26.Bf1 Bf5 26...Ba6 27.Rxe4!+- 27.a6 Rh6 28.a7 Qh4 29.h3 f3 This loks absolutely terrifying for a human, but apparently White wins with another only move: 30.Qd4!+- The point is to defend against the threat of Bxh3 followed by Qg4. Bxh3 31.gxh3 Qg4 32.Rxf3! No more exf3 thanks to the pin exerted by the queen being on d4. After Qxf3+ 33.Bg2 The a-pawn will carry the day. One certainly cannot fault Caruana for not seeing all of this. 26.f3! e3 27.a6 Rh6 28.Re2‼ An incredibly difficult only move that would have to have been prepared in advance. White is ready to play Qg1 and hold the position together. Even then, the variation continues. Qh5 29.Qg1 Bh4 30.g4! fxg3 31.Raxe3! White wins. But this is really just a machine line, and outside of human capability. 24...b4 25.Rb3 25.gxf4 I would have been tempted to clear the 3rd rank for defensive purposes, and this probably is an easier route to equality. Still, while Caruana's choice gives him some only moves to find, he did find the moves and he objectively was not worse. 25...Bh3 26.Rg1! The rook is needed for defensive purposes. 26.Re1 Bf6! Moves like Bc3 or Be5 could come next. White is in trouble. 26...f3 White is under a lot of pressure. Around here I thought Caruana might crack and give Magnus the first decisive result, but he defended very well. 27.Bf1! This is a tough move. I would have been tempted to just let the bishop stay on h3, since White can always sacrifice an exchange on g2 very happily and otherwise it's not clear what active role the bishop plays. But it was absolutely critical to play Bf1. 27.Bb5? A move like this looks sensible... Rf6! Until you realize White will simply be mated on the h-file! Rh6 and Qh5 is coming, and there is nothing to be done. 27...Bxf1! 27...Qh5 Caveman play does not work here since White is in better shape to keep the h-file under control. The queen coming to f1 is very important, for instance after 28.Rxb4 Rf6 29.Bxh3 Qxh3 30.Be3! Black will not get Rh6 very easily, and e4 is hanging to boot. White should win. 28.Qxf1! Another important decision. White cannot allow a disaster on the h-file. His position looks awful with the queen and the rook so passively placed and the d5-pawn falling, but he remains solid and can rely on the a5-pawn for counterplay. 28.Rxf1? Keeping active pieces comes at a heavy price. After Qg4! 29.Rxb4 Rf5! White will promptly be mated by Rh5 and Qh3. 28...Qxd5 29.Rxb4 Qe6 30.Rb5= The position is dynamically balanced, but White has to be more careful than Black. He managed without a ton of trouble. Bd8 31.Qe1! The queen returns to life. Bxb6 32.axb6 Rab8 33.Qe3 Qc4 34.Rb2 Rb7 35.Rd1 Qe2! Setting a devilish trap, though it is not enough to claim an advantage. 36.Re1! Far from the only holding move, but I am awarding an exclamation point for not falling for Black's idea. 36.Qb3+? Kh8 37.c4 Looks like it traps Black's queen. But the nasty response Rxb6!-+ Would promptly turn the tables! 36...Qxe3 37.Rxe3 d5 The position is balanced as neither side can easily mobilize their pawn majority. Black's center looks impressive, but he can't do anything with it so long as he is left with a passive rook on b7. 38.h4 Rc8 39.Ra3 Kf7 40.Kh2 Ke6 41.g4 Rc6 42.Ra6 Ke5 43.Kg3 h6 44.h5 The position is some kind of mutual stalemate where neither side can do anything. Magnus tried to make something happen with Kd4? But only made trouble for himself after the strong response 45.Rb5! When Black has to be somewhat accurate to not get into trouble. Still, he has more than one route to a draw, and Magnus found a very easy one. Rd6! Overprotecting the d5-pawn White was planning to harass with Raa5. 46.Ra4+ Ke5 47.Rab4 Ke6 Nice and easy. White has no good way to challenge Black's central pawn mass without allowing liquidation. 48.c4 dxc4 49.Rxc4 Rdxb6 50.Rxe4+ Kf7 51.Rf5+ Rf6 Black loses a pawn but simplifies into an easily drawn 3 vs 2 on the kingside. Caruana did not even bother trying. His winning chances are approximately zero. 52.Rxf6+ Kxf6 53.Kxf3 Kf7 54.Kg3 With only 2 games to go, each player has only one game to try with White. Carlsen gets to go first, and I'll be interested to see what first move he chooses, as well as if we will see a 6th 2.Nf3 Nc6 Sicilian in the final game. ½–½

Thirdly, I would like to draw your attention to a typical mistake you can make when analysing your own games with an engine. Do not blindly follow the advice of your engine when it comes to identifying the critical moment or the crucial mistake! Why not? Let me explain with a simple example.

In the World Blitz Championship 2022, Carlsen desperately needed to win against a much lower-rated opponent, Madaminov. However, he failed to gain any advantage after the middlegame and was forced into a relatively easy and completely equal endgame with only 10 seconds on the clock. (Madaminov had over a minute left. After each move both players received a 2-second increment).

Carlsen-Madaminov, World Blitz 2022, White to move:

Still, the World Champion won in his trademark, seemingly effortless manner. Where did Madaminov actually go wrong? Please, have a look at the live video of the endgame here before continuing reading.

According to my engine, the critical position arose after Carlsen's 52.Kd3:

Madaminov played 52…Bc4+? and was lost after 53.Ke4 Kd6 54.Kd4. However, the ChessBase Let´s Check function claims that the diagrammed position is still drawn after 52…Bd5! 53.f4 Bc4+ 54.Ke4 Kd6 55.Kd4 Ke6!, attacking the f4-pawn. (The best engines in the world need depth over 40 ply to be sure.)

So should we assume that Mamadinov's decisive mistake was his 52nd move? Well, in an abstract, scientific sense yes.

However, from a practical point of view it makes little sense to assume that 52...Bc4+ was a crucial mistake. In a blitz game it is virtually impossible to see all the subtleties of the position and make all the precise, machine-like moves.

In fact, Madaminov's mistake was to get into such a dangerous position from a relatively safe endgame. So where did he go wrong? Well, as any good coach would easily see, he made a very dubious decision after Carlsen's 46.Kc3:

Black can play almost any move with his bishop. However, Madaminov played 46…a4?, putting another pawn on a light square, making it more vulnerable and creating a hole on b4.

The machine assumes that the position after 46…a4 is still equal, and therefore this move is as good as any bishop move. But it does not tell you that the path to equality is now much narrower than it would be with the pawn on a5.

Therefore, 46…a4 is a mistake despite the engine disagrees.

Similarly, an experienced GM might argue that Madaminov committed a mistake even earlier. Let’s have a look at the position after Carlsen’s 41.hxg5:

Madaminov played 41…Kf8?!. However, 41…f6! leads to a very simple draw, as after 42.f4 fxg5 43.fxg5 Bf5 White has a weakness on g5 and this fact severely limits the activity on his king. In the game, Carlsen transferred his king all the way to c3. That would never be possible with such a weak pawn on g5. In addition, after 41…f6! a pair of pawns gets off the board, which brings Black closer to trivially drawn endgames.

Again, any engine would claim that 41…Kf8?! and 41…f6! are of equal value. However, they are all wrong. The position after 41…f6 is much easier to manage in a blitz game.

Here's the complete game:

Carlsen,M28591–0Madaminov,M2438
World Blitz 20222022
1.d4 Nf6 2.c4 g6 3.g3 Bg7 4.Bg2 d5 5.cxd5 Nxd5 6.Nf3 0-0 7.0-0 Nb6 8.Nc3 Nc6 9.Bf4 Nxd4 10.Nxd4 e5 11.Bxe5 Bxe5 12.e3 c6 13.Qc2 Qe7 14.Rad1 Bg4 15.Rd3 Rad8 16.h3 Bxd4 17.Rxd4 Rxd4 18.exd4 Bf5 19.Qc1 Rd8 20.Rd1 Qb4 21.a3 Qb3 22.g4 Bc8 23.Ne4 Nd5 24.Re1 Qb6 25.Nc5 Qc7 26.Qg5 Kg7 27.Nd3 Be6 28.Nc5 Bc8 29.Nd3 Be6 30.Qd2 b6 31.Rc1 Qd6 32.Ne5 Ne7 33.Nxc6 Nxc6 34.Rxc6 Qxd4 35.Qxd4+ Rxd4 36.Rc7 Rd1+ 37.Kh2 Rd7 38.Rxd7 Bxd7 39.g5 h6 40.h4 hxg5 41.hxg5 Kf8 42.Kg3 Ke7 43.Kf4 Kd6 44.Ke3 Kc5 45.Kd3 a5 46.Kc3 a4 47.Bf1 Be6 48.Be2 Bb3 49.Bg4 Bd5 50.Bd7 b5 51.Be8 Bb3 52.Kd3 Bc4+ 53.Ke4 Kd6 54.Kd4 Ke7 55.Bc6 Kd6 56.Be4 Ke6 57.Kc5 Ke5 58.Bc6 Kf4 59.Bxb5 Bb3 60.Kb4 Kxg5 61.Bxa4 Bd5 62.Bb3 Bxb3 63.Kxb3 Kf4 64.a4 Ke5 65.a5 Kd6 66.Kc4 f5 67.f4 1–0

----

Fire is a good servant but a bad master, it is said. Engines are the same. If you follow their advice blindly, they may become a hindrance to your chess growth. But if you use them wisely enough, they will be of great help to you.

The role of a good coach or a more experienced chess buddy therefore remains vital. Not because people play better chess. We simply understand better what it actually means to be human.

Let us learn together how to find the best spot for the queen in the early middlegame, how to navigate this piece around the board, how to time the queen attack, how to decide whether to exchange it or not, and much more!

Links


Jan Markos is a Slovakian chess author, trainer, and grandmaster. His book Under the Surface was the English Chess Federation´s 2018 Book of the Year. His last book, The Secret Ingredient, co-authored with David Navara, focuses on the practical aspects of play, e.g. time-management over the board, how to prepare against a specific opponent, or how to use chess engines during the training process. Markos was the U16 European Champion twenty years ago. At present he helps his pupils from several countries to achieve similar successes. Apart from focusing on the royal game, he is also the author of several non-chess books, focused on critical thinking, moral dilemmas, and phenomenology.

We use cookies and comparable technologies to provide certain functions, to improve the user experience and to offer interest-oriented content. Depending on their intended use, analysis cookies and marketing cookies may be used in addition to technically required cookies. Here you can make detailed settings or revoke your consent (if necessary partially) with effect for the future. Further information can be found in our data protection declaration.