2017 World Cup — Round 3 Complete

by admin on September 11, 2017

The 2017 World Cup fashion show… er, chess tournament… has now finished three rounds, and the upsets continue! Fabiano Caruana went down to defeat at the hands of Evgeny Najer, and both Levon Aronian and Anish Giri lost games to their lower-rated opponents before coming back to defeat them.

The final 16 are now set, and they are a group that nobody would have predicted before the tournament. Only seven of the remaining 16 are actually rated in the top 16 in the world:

Maxime Vachier-Lagrave (#2 in the world), Levon Aronian (#4), Wesley So (#8), Alexander Grischuk (#9), Anish Giri (#12), Ding Liren (#13), and Peter Svidler (#14).

They are joined by nine dark horses: Vladimir Fedoseev (#28), Vassily Ivanchuk (#32), Bu Xiangzhi (#35), Baadur Jobava (#40), Wang Hao (#44), Maxim Rodshtein (#50), Evgeny Najer (#52), Richard Rapport (#68) and the longest of the longshots, Daniil Dubov (#81).

Let’s look at the matchups for Round 4. There are few enough now that we can actually cover them one by one!

Vachier-Lagrave vs. Grischuk: This is the marquee matchup of the round, the only one that features two “favorites” playing each other. Vachier-Lagrave has had a sensational year to rise to #2 in the world rankings, but Grischuk is really tough in knockout events. He will draw the first two games and play to win in the rapid and blitz tiebreaks — and he’s very good at that. I give the edge to Grischuk.

Ivanchuk vs. Giri: Ivanchuk’s win against Kramnik was impressive. Giri was shaky against Sethuraman. So you might expect me to pick Ivanchuk — but I have a hunch here that Giri is ready to live up to his higher rating. Tossups always go to the younger generation! (OK, maybe not.)

Aronian vs. Dubov: On paper this is a huge mismatch, #4 versus #81. But really it’s a referendum on whether the world rankings mean what we think they do. This whole tournament suggests that the answer is no. Realistically I have to pick Aronian, but I will not be at all surprised if Dubov pulls off the upset.

So vs. Jobava: The match I most want to see. I’ll be rooting for Wesley So because he is the only remaining American. But I’ll also be rooting for Baadur Jobava because he is the most original, creative player in the field. Jobava also has a home-country advantage, because he is from Georgia. I’m picking So, but Jobava is the scariest opponent he has faced so far.

Najer vs. Rapport: I’ve been super impressed with Rapport, who has knocked out two tough Chinese in the last two rounds. Now he gets a “Cinderella versus Cinderella” matchup, and I think that he is the better of the two Cinderellas.

Bu vs. Svidler: Bu Xiangzhi is now known around the world as the man who knocked off Magnus Carlsen. But now he’s facing Svidler, who has played in many World Cups and almost always does well. Carlsen lost in part because of overconfidence, and I don’t think Svidler will fall into that trap.

Fedoseev vs. Rodshtein: I think that chess karma is coming for Maxim Rodshtein, who won his last match against Anton Kovalyov by forfeit. It’s always weird playing the next round after a forfeit win; you lose your rhythm, and you feel as if everybody else has been playing in the tournament while you have been in some strange alternate dimension. Not that Fedoseev needs karma to win this match. For what it’s worth, the computer says that he was the most “precise” player in round three.

Ding vs. Wang: Two of the remaining three Chinese face each other, and sparks will fly. These guys don’t play a lot of draws: in 11 classical games, Ding leads +5 -2 = 4. Michael Aigner told me not to overlook Wang because he hasn’t played as much in recent years. So I won’t ignore him. I’ll pick him to beat Ding in a mild upset.

What Does “Precision” Mean?

Usually I check the results at chess24.com or at the event’s website. Today, because both of them seemed to be temporarily down, I went to chessbase.com. I noticed that ChessBase puts up computer analysis of each game, and at the end of each one they had a “Precision Rating” for each player. For example, in the Carlsen-Bu game, Carlsen had a “precision” of 33 percent, while Bu’s was 64 percent.

Does anyone know what this is supposed to mean? A naive interpretation, which may be right, is that it’s simply how often the player played the computer’s top choice. If that’s what it means, it is a singularly useless statistic, because there are many positions where both the #1 and #2 moves are completely playable. Also, I think that human play and computer play diverge a great deal in lost positions. In some cases the computer’s idea of the “best move” may in fact be an obvious loss, while some other move poses some significant problems for the opponent to solve. The human master will always choose the latter.

However, the “precision” scores did correlate pretty well to winning and losing. In round three there were 22 decisive games. In every one of them, the winner was much more precise than the loser; the closest margin was Dubov-Artemiev, where Dubov had a precision of 52 percent and Artemiev 41 percent. (In that game Dubov played very speculatively and probably unsoundly, but pulled out a victory.) Only two other decisive games had margins under 20 percent: Lenderman 51 — Vachier-Lagrave 64, and Jobava 45 — Nepomniachtchi 30. Vice versa, in almost all of the drawn games, the difference between the two players was less than 20 percent.

Other interesting stats: The “most precise” game was played by Wesley So against Francisco Vallejo Pons; So won the precision battle by a margin of 98-37! The next two most precise games were played by Vladimir Fedoseev against Hikaru Nakamura. Fedoseev out-precisioned Nakamura by 95-76 in their first game, but only drew. Not satisfied that, he came back in the second game and out-precisioned Naka by 96-38, a game that of course he won. So this, I guess, makes Fedoseev the most precise player of the round! (Whatever that means.)

Print Friendly, PDF & Email

{ 7 comments… read them below or add one }

brabo September 11, 2017 at 8:36 pm


In my last tournament I had one game with a precision of 22% (loss) and one game with a precision of 95% (win) see https://live.chessbase.com/pgn/40th-eastman-open-ghent/0


admin September 12, 2017 at 10:24 am

Thanks! From the first article, it is clear that this is quite a new feature in ChessBase. I am still not sold on how meaningful it is. I would still like to know whether a winning player can have a lower precision score. So far, I haven’t seen one yet.

A very interesting test case would be my game against the world champion chess computer, Belle, in 1983. (Belle was the first computer with a master rating.) I’ve checked the game against modern engines (Fritz and Rybka) and they agree with a very high percentage of Belle’s moves, and a considerably lower percentage of mine. Nevertheless, Belle lost. It would be great if there were some way of submitting this game to ChessBase and seeing the precision scores.

What I think “precision” really means is “plays like a computer.” Which is really only relevant, for human competitions, if you want to detect cheating.


Joji September 12, 2017 at 2:27 pm

“Carlsen lost in part because of overconfidence”. It’s an easy enough excuse when a highly regarded team or individual loses to a much lower rank opponent. If that were the case, there were so many players playing over-confidently. It could very well be true but unless the players tell us ourselves, how can we really know? (rhetorical). 😉


admin September 13, 2017 at 10:03 am

I think there’s no question in this case. Just look at the game. Carlsen could easily have played Nf1 earlier; the reason he played the line he did is that he was goading Bu into playing the sacrifice on h3. To me, that looks like overconfidence.


Larry Smith September 12, 2017 at 5:08 pm

Also, assuming that your notion of “precision” is correct, I think that human and computer ideas of the “best move” diverge greatly in won positions, as well as lost (as you pointed out). If I can trade off pieces to get into a simple K + P win, I will gladly forego the more precise but more complex line. My precision rating will suffer, but my digestion will be much improved!


Todd Bryant September 13, 2017 at 8:18 am

Hey Dana, I just realized a clear example of the losing side having greater precision than the other side. This would happen when one side was slowly crushing for a long time, then blundered at the last minute. And that’s just what happened in Eljanov-Lenderman from round 1:


Precision: White=62% Black=39%


admin September 13, 2017 at 10:04 am



Leave a Comment

Previous post:

Next post: