Friday, August 21, 2009

Not d4 time

I had six Blacks at Benasque: all six were opened with 1.e4. A small sample, of course, far too small to be of any statistical significance, but nevertheless the absence of alternatives raised my eyebrows just sufficiently far to induce me to ask I asked myself whether it represented any sort of pattern. I started looking through my scorebook, beginning with the most recent games and working backwards, looking for games that my opponents opened 1.d4.

I found examples soon enough, but oddly - or, at the time, I found it odd - none of them involved players rated higher than myself. The six black games at Benasque, by contrast, had all involved players whose FIDE ratings were lower than mine. So I looked further - and though, eventually, I did come across games with lower-rated opponents playing 1.d4, another pattern emerged. While higher-rated opponents tended to accompany the move with 2.c4, the lower-rated players did not. None of them did.

There were Colles and Torres. There were games where c4 was played rather late on. But there were no Black games with a d4/c4 set-up as a first choice: not until I got as far back as February 2006, in which month I played two of them. A Slav and a Semi-Slav, both games against players graded not far below me (in the 160s ECF). February 2006. That's nearly three and a half years ago.

Is that extraordinary? It seems extraordinary to me. We all know that 1.e4 is a rather more popular choice, at club level, than the d-pawn. But where the d-pawn is selected - have the Queen's Gambit and Indian set-ups really become so very unpopular? At grandmaster level they took over from the alternative ways of handling 1.d4 eighty years ago and more - has this process, among club players, been reversed? (Or, perhaps, did it never really happen?)

I investigated my last 100 competitive OTB games with the Black pieces (which involved going back to 25 February 2005, picking up a couple more d4/c4 games from lower-graded opponents as I did so) and tabulated them by opening choice and strength of opponent. In the latter category I divided opponents up into three categories:
  • those substantially stronger than myself (which I categorised as above 2200)
  • those substantially weaker (below 2100)
  • those of relatively similar strength (between those two figures).
It's not quite so simple as that, since many players I meet have at least two Elo-based grades, Spanish and FIDE (I too have both) but where there has been a conflict as to which category the player should go into, I chose - wisely or otherwise - to privilege FIDE, which in each instance has put the player in a higher category. (None of these players were lower-graded players playing 1.d4.) One player, from France, who beat me in a Ruy López, had no grade I could locate and so I used my judgement, putting him in the middle category. Where I had only ECF grades for my opponents and some kind of conversion was required, I used my judgement and decided that ECF 168 was just below 2100, while 188 was just above 2200. (One player, as it happens, was put in two different categories for the two different games we played, due to his grade changing between them.)

The openings were also divided three ways:
  • 1.e4
  • d4/c4 set-ups
  • others
This involved making a couple of executive decisions regarding move-orders. 1.d4 2.Nf3 and 3.c4 - a sequence I employ myself - was categorised as falling under 1.d4 and 2.c4, given that it's intended to reach the same variations and positions. Similarly, 1.c4 and 2.d4. However transpositions from the Colle to an e3 Slav - there were a couple of these - were not, as I considered that the intention had been to play a Colle.

The table I thus compiled is below.


e4

d4+c4

others

above 2200

17

11

2

2100 - 2200

17

6

3

below 2100

28

4

12


What, if anything, does it tell us? On one perfectly reasonable approach, not very much, since 100 games is still a very small sample. (I'd be interested in the opinion of anybody with qualifications in statistics, but I think I'd want perhaps 10,000 before coming to any serious conclusions.) Moreover the fact that all were played against the same opponent might in itself cause the statistics to be skewed, given the possibility of preparation, of opening choices being influenced by the identity (and for that matter the playing strength) of the particular opponent.

Still, they are the statistics I have, so they're the ones I'm working with. Note that because they are based on exactly 100 games, it's simple to add together figures to get overall percentages, so for instance we can see that 1.e4 was selected in 62% of games, d4/c4 in 21% and "others" in 17%.

Nothing utterly startling there, though the figures for "others" would, I think, be rather lower at professional level. But what does stand out, for me, is that for players rated below 2100, d4/c4 systems are outweighed by "others" by a ratio of 3:1. There's an obvious contrast with the higher-rated categories: in the middle-strength category the ratio is 2:1 the other way and in the highest, the imbalance against "others" is more than 5:1.

There's a slight falling-off in the use of 1.e4, the higher we go: it was employed in 57% (17/30) of all games in the highest category, but 65% (17/26) in the middle and 64% (28/44) in the lowest. Whether or not that would be reproduced with a bigger and better sample, I'd be interested to know: I was a little surprised to see players below 2100 apparently no more fond of 1.e4 than players between 2100 and 2200. A better sample would of course show whether that was actually so.

But it's the differential between d4/c4 and "others", among the lower-graded players, that really commands the attention. Partly because of the way in which it reverses the situation obtaining in higher-strength categories - partly in and of itself.

That 1:3 ratio, if it's an accurate reflection of the general reality, would tell us a lot about how lower-graded players approach their openings - and it might also tell us a lot about how they should approach their openings. It's been written* that at club level, 2.c3 in the Sicilian is the main line. If that's true, it may be similarly true that at club level, where 1.e4 is not selected it's the alternatives to d4/c4 that are the main line: not the traditional d4/c4 set-up at all. To a surprisingly large degree, but a large one nonetheless.

How many club players, studying their repertoire as Black, spend three times as much time studying those alternatives - Reti/English, Colle, Torre, Trompovsky and so on - than the Queen's Gambit and Indian lines which have been dominant for several generations at a higher level? My guess would be along the lines of "none, or close to none". But it might be wiser if they did.

Of course, if everybody did, then people would start playing the "main" lines after all, but as it stands, if your opponents are overwhelmingly below 2100 in strength - and this is presumably the case for most club players - then it's quite likely, at the moment, that when you're looking at opening theory that will be of actual use to you, you're looking in very much the wrong place.

It's a lesson it's taken me thirty years to learn, that you basically want to prepare for opponents you're actually likely to meet. Of course it's true at the same time that the higher-graded your opponent, the more they're likely to know and understand, and so it can be helpful to spend more time thinking about what to do against them than you is proportional to the regularity with which you meet them - especially if, like the the present writer, your opening choices vary to some extent, contingent on the strength of the opposition. But still, if you are thinking practically then it's a point to bear in mind, and perhaps at the forefront of your mind.

Are you really going to study contemporary theory in the Anti-Moscow Gambit when the only people who play it are people who you meet maybe once every two seasons and who are graded wildly higher than you anyway? Or might you be better off thinking about the Colle System that you come up against several times a season - and never really know what you should do against?

[* that is, written to me by Tom in an email a couple of years ago. Tom draws it to my attention that in the same email, he also mentioned the lack of mainline d4/c4 openings at club level. I should have read more carefully, or remembered more fully!]

7 comments:

Jonathan B said...

"It's been written* that at club level, 2.c3 in the Sicilian is the main line. If that's true, it may be similarly true that at club level, where 1.e4 is not selected it's the alternatives to d4/c4 that are the main line: not the traditional d4/c4 set-up at all

... if your opponents are overwhelmingly below 2100 in strength - and this is presumably the case for most club players - then it's quite likely, at the moment, that when you're looking at opening theory that will be of actual use to you, you're looking in very much the wrong place."


Your experience re: how club players tend to play 1. d4 2. others and not 2. c4 pretty much matches my own.

Indeed, it's partly that observation that prompted me to take up the Dutch. Something I intend to write about at some point.

Interesting that we should have come to these similar conclusions independently.


btw:
I'm not qualified in statistics but I did take a stats course at the LSE last year. I recall that, depending on the conditions, statistical relevance can be achieved with surprisingly small samples. Naturally, I can't actually recall what these conditions are.

Tom Chivers said...

On a comparable note, in Ruston (2092) - Rogers (2337) black annotated his 1... c5 in reply to 1.e4 as follows: "For the first time in 22 years and the second time in my life. I was inspired by the assurance that these days, no one under 2500 dares to play 2 Nf3 and 3 d4."

Here for the game.

dfan said...

This 1800 (US) player just switched from d4/c4 openings to e4, mostly because I felt it would improve my game to finally try e4. I feel a little bad about following the herd. At least I'm not playing c3 or closed Sicilians...

Cox's Dealing with d4 Deviations was a godsend to my repertoire for exactly the reasons you mention.

Anonymous said...

It worked for Mark Hebden and Julian Hodgson =).

I noticed a similar trend when playing black as 1. e4 e5. How higher-rated the opponent how more likely they played the Ruy Lopez.Lower rated player tend to mostly play the Italian opening.

In a way I think both your observation and mine aren't unlogical.

In comparison d4 + c4 does require more understanding and flexibility then "D-pawn specials".Higherrated players will more able to cope with these demands.

Take for instance 1.d4 Nf6 2. c4 white has to prepare for the various indians while the QG still can be reached. These lead to wildly different positions with different plans.

After 1. d4 Nf6 2. Bg5 black options are more limited and the positions will be less diverse.

Anonymous said...

Guilty.
I'm roughly 2050, and gave up c4/d4 because I was fed up with dubious black defences that always beat me, like the Budapest. I also had inspiration from a very good book by Summerscale called "A killer chess opening repertoire". My results improved a lot since then, as the lines avoiding c4 are generally easier to learn and more likely to surprise your opponent - who with black actually like facing the Tromp or Colle?

I picked up the Alekhine years ago and I remember spending ages learning the "main line" - 1.e4 Nf6 2.e5 Nd5 3.d4 d6 4.Nf3. This is the only line that GMs play, yet I have encountered it maybe 5 times in 20 years of OTB chess, club players preferring to exchange on d6, avoid thrusting the e pawn or go all out for the four pawn attack.

I've found internet chess to be a good tool for finding out what people really play as opposed to what the books (written by GMs) say people will play.

Adam B.

Lauri said...

Internet chess is weird pool of ideas as there might be anything. It is also interesting to dip into this pool as some ideas seem to catch on. But yes: you get idea on what is topical: in scotcht game rarely people go for Nf6 lines and go typically for Bc5 lines etc..

ejh said...

That's pretty topical in GM chess as well, though.

It's also the line I play, though that may not be consequent.


I'm roughly 2050, and gave up c4/d4 because I was fed up with dubious black defences that always beat me, like the Budapest.

1.d4 2.Nf3 3.c4 should solve that problem.