# The value of opening lines

Do some opening lines favor white and others black? Or are they optimized to maintain a neutral position? To answer this question I selected 190 commonly played opening lines (out of thousands) from the "opening theory" book available online at wikibooks and played each one of them on engines in analysis mode and recorded the value of white's position at the end of each line. The data are shown in a pdf file on freepdfhosting.com that can be viewed or downloaded using the link shown below:

Opening data sorted by move sequence: http://freepdfhosting.com/634a84afbc.pdf

The positional values are computed on a scale in which the material value of a pawn is set to 100 points. At the start of the game, white has a positional advantage of 20 points by virtue of holding the first move. Since white's win percentage is higher than that of black's ceteris paribus, we can assume that a 20 point advantage is meaningful in the game. For the sake of the rest of the analysis we shall assume that an advantage of less than 20 points is not statistically important in terms of game outcome. In particular, we should also note that position value estimates have a variability that indicate a measurement error of plus or minus 5 points.

Accordingly, in the second data sheet I have color coded the data to separate the opening lines in my sample into three distinct groups.

Opening data sorted by position value: http://freepdfhosting.com/b1c3a63855.pdf

In the middle group, in black letters, we have about 130 lines representing 66% of the sample with white's position value in a range range from 0 to 40. Since white started with a positional advantage of 20 and ended with a value somewhere between 0 and 40, white's positional advantage did not change by more than 20 points by virtue of the opening moves.  We therefore classify these openings as neutral in the sense that the opening moves did not substantially affect positional advantage.

At the top of the list are about 30 opening lines color coded in red in which the opening moves caused white's positional advantage to decrease by more than 20 points. These lines favor black. If the line is a "defense" then we can consider the defense to be an effective one. If it is an "attack" or a "gambit" by white then we can consider the attack to be ineffective. In any event, black's response has been stronger than white's attack and white has completely lost it's initial first move advantage.

The bottom group, colored in green, contains about 30 opening lines which favor white. In these lines white's positional advantage has increased by more than 20 points by virtue of the opening moves. We can therefore consider white's "attacks" in these lines to be effecive or black's "defense" to be relatively ineffective.

In summary the answer to my question seems to be that mostly (in 2/3 of my sample) opening lines played faithfully do not favor white or black but about 1/3 of the lines favor one side or the other and these are symmetrically distributed. Half of them favor white and the other half favor black.

The first pdf file in which the data are arranged by move sequence also serves as a compact ready reference to 190 commonly played opening lines.

Cha-am Jamal, Thailand

• 7 months ago

• 2 years ago

• 2 years ago
[COMMENT DELETED]
• 2 years ago

Caveat emptor: chess engines are weakest at opening evaluations.

When you did this, did you turn the engines' opening books off, or leave them on? In either case, what effect do you think that had on the results?

It used to be the case that their weakness was endgames, but once engines became strong enough to increase search depth to beyond a dozen moves in a reasonable amount of time, this was no longer the case, and now with tablebases, most engines will outperform any human almost all of the time.

But until search depth can be around 30 moves deep in a reasonable time, engines will remain remarkably poor at evaluating opening positions for many reasons, including problems caused by search horizon, faulty quiescence optimization, aggressiveness of alpha-beta pruning, etc. As evidence, note that to have any chance against top players, the very best engines need very thoroughly developed, specialized openings books.

How in the world did the monster obtain a -5 score for 1.e4 e5 2.Nf3 Nc6 3.d4 exd4 4.Bc4 Nf6 5.e5 d5 6.Bb5 Ne4 7.Nxd4 Bd7, but obtained a 0 score for 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.d4 exd4 5.e5 d5 6.Bb5 Ne4 7.Nxd4 Bc5, when the best continuation after each of these lines reaches exactly the same position, and that position is widely regarded to be very comfortable for White, to the extent that Black is thought to have to fight hard for a draw? Does the computer diverge in each line? I find that unlikely; one line, yes, both, hard to believe. That could explain the different scores, but it doesn't explain why both are so low when most players would kill to achieve a position like White's in every game.

You are spot on in your comments regarding the 20 point lead, and in your treating +/- less than 20 as a more or less irrelevant difference, and this is indeed a good amount of work, but I wouldn't let an engine decide the value of any opening less than 12 to 15 moves in. Even then, I treat engine evaluation with suspicion, because I know a lot about how their evaluation functions work - or don't work, to be more precise.

I'd love to see your step by step methodology. Which engines, on what machines, with what specs, analysis mode set for how long, with what other programs running. You mentioned engines, plural. Did you run multiple engines on each line, or one engine per line, or was it the same engine for all lines, or some lines one and some another? Did you ever average results, and if so, how (mean, mode, median)? If you have that information, and enjoy writing such things up, I and I am sure others would be very interested in reading it. Others of us could duplicate your work under the same or similar conditions, or vastly different ones, but using the same methodology, or vice versa, and see what the results are.

Thanks for the thought-provoking contribution!

• 3 years ago
[COMMENT DELETED]
• 3 years ago

the list of openings that is sorted by move sequence is like a bird's eye view of the opening book with the added bonus of having white's positional value at the end of the line shown and also the eco code and opening name in case you want to do some research on other variations.

• 3 years ago

dr frank, in online games we are allowed to use opening books and databases during the game and so i think it would be helpful to have this list in front of you. in the file where the openings are sorted by move sequence you can follow the opening as it unfolds on this list to see what it is homing in on and to make sure that you don't cooperate in the construction of a move sequence that changes the value against you. for example if you are playing black and the ruy lopez is unfolding you may want to avoid c63 schliemann by not playing 3...f5 and instead playing something like a6, d6, or Bc5 or any of the other ruy lopez lines on the list that do not favor white in terms of white's positional value. on the other hand, if you are playing white and black does play 3...f5, you will know exactly how to proceed to take advantage of that move to increase your positional value. and so on.

• 3 years ago

This data sheet has a practical use, in that one can memorize the best lines for white and black and use those lines when playing chess. Please let me know if I am right or wrong!

• 3 years ago

yes, drfrank, the data show that the c57 two knights/classical variation line shown in the datasheet is the best line for white in this sample and that the b02 alekhine's defense line shown in the datasheet is the best line for black in this sample. please note that i am being careful to include the phrase "line shown in the datasheet" as there are other lines that might fit the eco code and description. however, that is not the point of this post. the point i was trying to make is that there are a significant number of opening lines  - among the most commonly played lines - that favor one side or the other.

• 3 years ago

I'm not sure I completely understand this. Forgive me if I'm not too bright but are you saying that those lines in green are best for white and that white's very best line is the Two Knights/Classical? Is this what the chart shows?

• 3 years ago

Hi Jamalov,

Great work! I would like to make these available for download on our website Chess Mastery with your permission.

Of course, you will be given full credit, a link to this article or any other blog or website you have.

Thanks again.

• 3 years ago

i keep a printout of the first pdf file (sorted by move sequence) next to the computer when playing online games and that makes it hard to surprise me with exotic openings and also easy for me to spot when my opponent has made an openng error. if you don't have enough of these lines memorized you might want to download that file and print it out.

• 3 years ago

josechu, i know some guys who have a lot of this stuff memorized but i don't know how they do it.

• 3 years ago

Wow, Jamal! There ought to be a Nobel prize for something like that. Imagine being a professional chess player and having to memorise all that stuff!

• 3 years ago

thank you dan, it was a labor of love but i always wanted the these openings displayed in a compact manner for quick reference that that is how this thing got started.

• 3 years ago

That was a very interesting analysis! It must have taken you quite a while.