Grading the Openings (Part Two)

by admin on March 6, 2015

First of all, let me announce that my last post, Grading the Openings (Part One), unexpectedly turned into the biggest hit I’ve ever had on this blog. The blog had 1136 visitors yesterday, which is three times more than I have ever had in a single day (except during the 2012 World Championship match, when I was translating Sergei Shipov’s commentaries). Thanks to all of you, especially to the many first-timers who found their way here because of a Reddit link. I hope you’ll enjoy today’s post as well.

Today is the post I really look forward to, where I am going to name names and point fingers at the good, the bad, and the ugly of chess openings. As I mentioned last time, all of these statistics are based on a collection of all games in ChessBase, played since 2000, where both players were rated 2200 or above. There are just short of a million games meeting that description. Also, a disclaimer: I absolutely do not recommend that you choose your openings based on these statistics. They are just for fun.

And now let’s have the award envelopes…

Most Popular ECO Code

popular

This category is won by the Alapin Variation of the Sicilian Defense (ECO code B22), 1. e4 c5 2. c3, in a landslide. Interestingly, the top five variations are all in ECO code B. In order, they are:

  1. B22 (Alapin Sicilian) — 17775 games
  2. B07 (Pirc, all variations except Classical and Austrian Attack) — 13644
  3. B33 (Sicilian, primarily Sveshnikov and Pelikan) — 12252
  4. B01 (Scandinavian) — 11306
  5. B06 (Robatsch, 1. e4 g6, all variations that don’t transpose into Pirc) — 11234

Interesting as the results are, to me they are a little bit skewed by the choices the ECO designers made. For example, the Scandinavian is not all that popular as an opening. But when you put all the Scandinavian Defense games into one ECO code, you’re essentially comparing it with variations of other openings, and then it looks tremendously popular. If I were designing the codes, I would at least split the Scandinavian up into two variations, 1. e4 d5 2. ed Qxd5 and 1. e4 d5 2. ed Nf6.

Because these are all openings in group B, I’d also like to list the most popular opening variations in the other categories.

  1. D02 (1. d4 d5 2. Nf3 various continuations — not 2. … Nf6) — 8527 games
  2. C41 (Philidor!) — 8395
  3. E11 (Bogo-Indian) — 8313
  4. A30 (Symmetric English — not 2. Nc3 Nc6 or 2. Nf3 Nf6 3. d4) — 8005 games

I was stunned to see Philidor’s Defense up there. I had no idea that it has become so popular among masters. Among amateurs, sure. Dang, maybe I’ll even start playing it.

Least Popular ECO Code

unpopular

I thought this category would be a lot of fun, but unfortunately I don’t quite trust the results. I think that there may have been problems with categorizing some of the more unusual variations. Nevertheless, we do appear to have a clear winner: A76, the Classical Benoni with 9. … Re8 (Czerniak Variation) where White doesn’t play 10. Nfd2. It has only been played 18 times in the database. I guess the variation is played out (especially for White; I wouldn’t want to play it). Also, 10. Nfd2 is the best option for White, so not too many masters play anything else. Here are the top five unpopular openings, according to the database:

  1. A76 (Benoni Czerniak without 10. Nfd2) — 18 games
  2. E78 (King’s Indian, Four Pawns Attack) — 27 games
  3. E79 (King’s Indian, Four Pawns Attack) — 27 games
  4. D69 (Queen’s Gambit Declined, Ultra-Classical Variation that goes 1. d4 d5 2. c4 e6 3. Nc3 Nf6 4. Bg5 Be7 5. e3 O-O 6. Nf3 Nbd7 7. Rc1 c6 8. Bd3 dc 9. Bxc4 Nd5 10. Bxe7 Qxe7 11. O-O Nxc3 12. Rxc3 e5 13. de. This is truly analyzing the opening to death!) — 36 games
  5. A74 (Classical Benoni with 9. … a6 instead of 9. … Re8 just before the illustrated position) — 53 games

I absolutely, flatly don’t believe #2 and #3. I think these positions are being played a lot but being classified as Benonis or even as weird Sicilians. Or else maybe Rob made some mistake in tabulating the data.

Best Opening Variation for White — Common Variations

goodforwhite

From here on, I am going to separate the ECO codes into “common” (played more than 1000 times) and “uncommon” (played less than 1000 times). The reason is that the less common variations tend to have more extreme statistics simply by virtue of less data; it doesn’t mean that they are actually stronger or weaker. Also, this was a fairly convenient division because roughly half the ECO codes are common (283) and roughly half are uncommon. And an opening that appeared less than 1000 times in a million games is showing up less than 1/1000 of the time, which is indeed pretty uncommon.

In the common category, the best opening variation for White was the A44, the Czech Benoni (1. d4 c5 2. d5 e5), shown above. White scores 63.9 percent here. It’s a little bit hard to say why it’s that good for White. Of course, Black has given White a space advantage and has not extracted any concessions in return, but still I wouldn’t have expected it to be the worst opening for Black of all 283 common ECO codes. It’s been a rough millennium for the Benoni!

Here are your top five opening variations for White (or worst for Black):

  1. A44 (Czech Benoni) — 63.9 percent
  2. D37 (Queen’s Gambit Declined with 4. Nf3 but not the Semi-Slav, Semi-Tarrasch, or Ragozin) — 63.0 percent
  3. A39 (English Symmetric, 1. c4 c5 2. Nc3 Nc6 3. g3 g6 4. Bg2 Bg7 5. Nf3 Nf6 6. O-O O-O 7. d4) — 62.6 percent
  4. A52 (Budapest Gambit with 3. … Ng4) — 62.2 percent
  5. E69 (King’s Indian, Fianchetto Variation, Main Line with 9. h3) — 62.2 percent

#2 doesn’t surprise me. This is an artificial category created by taking away Black’s best moves, a little bit like A76 discussed above, where we took away White’s best move. #3 does surprise me. Who would have thought that a symmetric position would lead to White’s third-highest win percentage? This truly shows that symmetric openings do not have to be drawish. #5 is a favorite of Botvinnik, who played it against both Tal and Bronstein in World Championship matches. We’re going to see his name again. Botvinnik had really good taste in openings.

Best Opening Variation for White — Uncommon Variations

goodforwhite 3

Botvinnik strikes again! The champion in this category is a variation that he cooked up for his 1951 World Championship match with David Bronstein. It’s code A94, the Dutch Stonewall Variation with 9. Ba3. Unlike the previous category, I can completely understand why this is a winner for White. Black’s position is riddled with weak dark squares, and White is about to trade off Black’s dark-squared bishop. How can this be okay for Black? It isn’t. This is the only ECO code where White scores greater than 70 percent.

Here are your top five:

  1. A94 (Dutch Stonewall with 9. Ba3) — 70.5 percent
  2. D99 (Gruenfeld Defense, Smyslov Variation) — 68.1 percent
  3. C81 (Open Ruy Lopez, Howell Variation 9. Qe2) — 66.0 percent
  4. B39 (Sicilian, Maroczy Bind with 7. … Ng4) — 65.9 percent
  5. D51 (Queen’s Gambit Declined, with 4. … Nbd7 but not Cambridge Springs) — 65.7 percent

What’s fascinating about this category is that even though these variations are “uncommon” today, they have rich histories behind them. All of them except #4 have appeared in World Championship play. #2 was played three times by Smyslov in the 1948 World Championship tournament. #3 was played four times in the 1948 World Championship tournament. #5 was played in the final game of the Capablanca-Alekhine match of 1927. Wow! I think they are less common now because tastes change, and because White won too often for Black players to continue playing them.

Okay, this entry is already getting pretty long, so I will stop here. Next time, in Part Three: the best opening variations for Black, the most frequent “drawing variations,” the variations with the least draws, and some final comments. Stay tuned!

Print Friendly

{ 6 comments… read them below or add one }

Franklin Chen March 6, 2015 at 7:32 am

I would like to use the Ba3 idea against the Stonewall as an example of why these opening statistics are flawed.

I used to play the Stonewall as Black. I don’t think much of the dark Bishop trade idea for White. It is safe for White, definitely, but in practice, the whole b3 idea for White is not so great, if Black plays in a non-passive way with plans such as a5, Ne4, and other ideas, making things look like a closed Catalan or Queen’s Indian with Queen side and center activity where it’s hard for White to make real progress. White can get a Knight to e5 but it will be traded, or White can maneuver the other Knight around, but then Black has time to do something in return and the e5 weakness is mainly optical unless White really can create a break with f3/e4.

I believe there is another reason that Ba3 scores so well for White. It is because strong players have been taught that it’s good, and therefore Black if following “theory” will reflexively prevent Ba3. So allowing Ba3 is a signal that Black does not know theory, and therefore that there may be other things Black does not know about playing the Stonewall properly also (e.g. maybe Black is tied to old Qe8/Qh5 ideas or passive do-nothing play).

Reply

admin March 6, 2015 at 10:39 am

This is a really great comment, and relates to something I’ve been thinking about a lot lately. We see from the data that 9. Ba3 is positively associated with Black losing the game. But is it the cause? Your answer is no: there is another factor, Black’s lack of preparedness, that is a common cause of both things. It causes him to allow White to play 9. Ba3 and it causes him to play poorly later on. Unless the data are analyzed in a much more sophisticated way than I have done, we cannot tell the two explanations (A causes B, or A and B are both caused by C) apart.

That’s why I said at the outset, “I do not recommend that you choose your openings based on these statistics.”

Reply

Michael Aigner March 6, 2015 at 1:03 pm

Let me continue the discussion on the Stonewall Dutch. First of all, the position shown is 8.Ba3. Both sides moved 4 pawns and castled. White moved 3 minors and black moved 2. QED

I completely agree that the position above is evidence that black doesn’t know his theory. Indeed, after 1.d4 f5 2.Nf3 Nf6 3.g3 e6 4.Bg2 c6 5.c4 d5 6.O-O, the move Bd6 has been played about 2200 times in Chessbase Live Book and Be7 less than 100 times. Note that various move orders are possible to arrive at this position. White frequently also plays Nc3 or Nbd2 in the first 6 moves, although both preclude the Ba3 idea we are discussing.

So is the entire Ba3 concept rubbish? No. Continuing the above variation, 6… Bd6 7.b3 Qe7 8.a4 a5 9.Ba3 is a topical position. Is this the 9.Ba3 that you meant? But this position is hardly as promising for white as you proclaimed. After 9… Bxa3 10.Nxa3 O-O, white scores a whopping 51% in 85 games. In fact, 8.Bb2 is preferred over 8.a4, although I see both often enough.

My next question is which variation corresponds to ECO code A94? The answer surprised me a bit. I see both 8.Ba3 and 9.Ba3, plus a few games that aren’t even Dutch until black plays 9… Ne4 and 10… f5 (Rausis-Cekro 2002). Bah humbug!

Reply

Michael Aigner March 6, 2015 at 4:21 pm

Please let me continue the train of thought in the final paragraph that I rushed earlier.

Clearly some of the ECO codes that were invented in the prehistoric era have fallen out of favor. Why are the variations leading to Ba3 on move 8 or 9 so special that they get an entire ECO code? Out of the 20 variations dedicated to the Dutch, a quarter have only 100 games included in Megabase, give or take (restricted to year 2000+ and rating 2200+). On the other hand, a half dozen codes have 2000 or more games (A80, A81, A84, A87, A88, A90).

If I was designed Dutch opening categories, I would feature five variations: Stonewall, Leningrad, Classical, anti-Dutch systems, and one for other moves. One could break these down a little more, but ten variations max (i.e. Dutch = A8 without A9).

By the way, my favorite variations of the Dutch are A80, A85 and A87, with the occasional A90.

Reply

admin March 6, 2015 at 5:14 pm

This whole exercise has made me appreciate both how flawed and how ambitious the ECO code project was. Clearly there are goofs. Benoni should have at most 10 codes instead of 25. Dutch should have at most 10 instead of 20. Scandinavian should have 2 instead of 1. On the other hand, if you made all of these changes, you would lose the balanced system where each letter gets 100 codes. You might end up with 63 A’s, 127 B’s, etc. (I’m just making these numbers up.) Is that a problem, or is that the way it should be? Obviously the ECO people liked the balance. If you want to maintain the balance, then it’s a really challenging math problem to fit the openings in any kind of sensible way, and I think they did an impressive job.

But there ought to be a way for the chess community to vote on changes, so that outmoded variations don’t stay around forever.

Reply

Franklin Chen April 22, 2015 at 6:21 am

I’ve been enjoying Carlsen’s success with the Stonewall as Black: http://dejanbojkov.blogspot.com/2015/04/the-clash-of-top-seed.html

Reply

Leave a Comment

Previous post:

Next post: