Sunday, August 19, 2007

How usefull is opening study?

Both Takchess and HisBestFriend raised a question about the usefullness of opening study.
I'm not a grandmaster, so my verdict has not the power of law. Yet if I look at the figures, I'm inclined to think are they insane? If I see grandmasters memorizing lines 20 moves deep or more, I get the feeling they are fooling themselves and each other. Isn't it just a matter of fashion? Let's try to use some reason.

Fischer suggested that openingstudy would kill chess in the end. In order to prevent that, and to become independant from the study of openings, he invented his shuffle-chess. Since Fischer chess has 960 different start positions, the amount of openingstheory becomes 960 x so vast. That raises the following question: can we create the same effect without altering the rules?
At every ply in the opening there are about 20 possible moves. If 5 of them are mainstream theory and 15 "unorthodox", you need only 3 plies where you make an unorthodox move. That gives you 15 x 15 x 15 = 3375 new possibilities. Are all these possibilities automatically bad since they are not part of the "theory"? That is hard to believe.

Have a look at the following table from my database with 650,000 mastergames.
It lists the moves that are played by white after 1.e4 c5 2.Nf3 d6

Move 3





1: d43460381.6%53.5%+0.19
2: Bb5+449010.5%50.7%+0.17
3: c315923.7%50.9%+0.03
4: Nc37921.8%52.9%+0.21
5: Bc45111.2%56.2%+0.23
6: g31700.4%52.0%+0.03
7: d31250.2%37.2%+0.06
8: b3370.0%41.8%+0.05
9: Be2 220.0%29.5%+0.09
10: b4210.0%38.0%-0.18
11: c4 170.0%38.2%+0.05

81.6% of the whiteplayers play 3.d4
Why on earth is that so? I have never heard a reasonable explanation why d4 should be the move. If you look at the figures, both in practice and according to Rybka 3.Bc4 should be the move. So why is everybody trading a fat centerpawn against the pathetic farmer at c5? Leaving black with a stronger center and an open line further away from the king? With a ready to exploit minority attack?

Is it just fashion? Do people fear to think for themseleves? Even masters and grandmasters? The figures suggest that even they behave like sheep.

While I got better at tactics, I felt more and more independant of the opening. Of course I don't know how that will be at grandmaster level, but I would test the hypothesis for sure if I was one of them! If it is possible to get rid of openings study by such simple means, you could use your time for different topics, more worth to study.

After 2 hours calculating and 19 ply deep Rybka comes up with 3.Nc3 as the first choice (+0.09) and 3.Bc4 as a second choice (+0.07). I suspect that it will come closer to zero if I let it run longer. To my knowledge these openings haven't even a distinctive name!

So if you know what you are doing and your opponent does not, openingstudy can give you an edge. But if you don't want to study you can force anybody in unknown territory in just a few moves. Without being worse than in the mainlines. Then you must both play chess as early as move 5.


  1. I am pleased to say I am not a sheep playing 4: Nc3 792 1.8% 52.9% +0.21 and enjoying it. One can get lost in the study of all openings and their variants 20 deep. This which would just be silly. My thinking is that through studying openings one studies postions and why one move is good and not the other. This should be an active learning process not a bowing to Fritz.

    I believes it is useful knowledge and not limited only to the openings. The knowledge that one should immediately challenge blacks forward pawn on e4 in the falkbeer transcends the falkbeer postion. Thinking about these sort of things helps in a philosophy of how I want to play a game.

    It may also help in understanding some tactics normally found in the framework of a game.

    It is about what is a good move and why. Not a blind thoughtless memorization of moves.

  2. very nice!

    tripple bravo!!!!!

    this table greatly reminds me of Muellers table in the back of his Fundamental Endgames, where he and Lambrech quantify the frequency of endgame types, in the big or mega database of millions of games, for purposes of their cannonical text.

    warmest, dk

    PS: i just met Blunderprone and his wife and daughter briefly, as they are out in Seattle to see their daughter who was at Microsoft as an intern, off from MIT for the summer, and fast though it was, was marvelous.

    i said, as i hugged them all goodbye, to him: "you are the first from here [blogger] that i have met, but wont be the only one" [thinking of you tempo, just fifty meters from another major port, off of the vast and at times stormy and mysterious and wild Pacific Ocean:

    if i travel far and long enough west, i know that i can reach the Netherlands from Seattle's Elliot Bay].

  3. Great post...It is quite liberating to be off of the opening study habit.

  4. Excellent article.

    I believe that it is of great practical value to have some idea of how you are going to start the game. The question is how much time and effort is opening study worth compared to other portions of the game. Having a simple opening system as White allows you to get to a playable middlegame that you are familiar with with a minimum of effort. Having a simple system as black does the same. And, as black, a little more preparation may be justified to keep from getting rolled right off the bat.

    Learning something like the KIA as white is a great basis for an opening system. You can play it against anything. Over time, if you like, you can add specific lines against various openings.

  5. Just some thoughts about 3.d4 (I prefer 2.d4). Of course you give away a center pawn but you get a semi-open file instead. In the Morra Gambit I use to play, this is a major strategic element and gives many opportunities for tactics. A similar idea may be behind 3.d4. If you leave that c5 pawn, the white position is somewhat under strangulation, so 3.d4 may be just a freeing motif.

    Christian aka Mousetrapper