How does one determine if a book move is objectively better than another?

Question

For example, I have heard the following statements:

1. After 1. e4 e5 2. Nf3 Nc6, the move 3. Bb5 is marginally better than the move 3. Bc4.
2. Openings moves such as 1. e4, 1. d4, 1. Nf3, and 1. c4 are equally good and none is marginally better than the others.

Are these two statements accurate? If so, is there any way to "prove" them?

I am not sure about the following argument:

Since after 1. e4 e5 2. Nf3 Nc6, 3. Bb5 is played more often by grandmasters than 3. Bc4, 3. Bb5 must be the better move.

As by the same logic, we can say 1. e4 is better than the other three popular first moves.

Answer 1

1. Here is one approach to prove if one opening book move is better than the other.

Algorithm

1. Create positions arising from the opening that you want to study.
2. Get stockfish and play a match on those positions.
3. Keep a record of win/loss/draw and score percentage for each opening.
4. Compare the score percentage among openings.

With that you will be able to determine if one opening is better than the other.

Opening

1. start: e2 e5 Nf3 Nc6 Bc4
   e4 e5 Nf3 Nc6 Bc4 Nf6
   e4 e5 Nf3 Nc6 Bc4 Bc5
   ... add more, more is better to reduce uncertainty

2. start: e4 e5 Nf3 Nc6 Bb5
   e4 e5 Nf3 Nc6 Bb5 Nf6
   e4 e5 Nf3 Nc6 Bb5 a6 Ba4
   ... add more

Engine
stockfish 14.1-1 vs stockfish 14.1-2, name is different but they are the same stockfish 14.1 engine

Evaluate stats
Get the score percentage of white in opening 1 and 2. If white score percentage in opening 1 is more than the white score percentage in opening 2 then opening 1 (e2 e5 Nf3 Nc6 Bc4) is better.

Tools
You may use ordo or bayeselo to generate stats from the output pgn file. Or load the pgn on softwares that support opening stats views like scid and others.

2. Another approach is by viewing the score percentage of the move from the given player. Example get the games of Magnus load it in scid, chessbase, or hiarcs explorer or any other game viewer software that is capable of showing opening stats. Below I use hiarcs explorer.

According to Magnus

Bb5 is 72% for white and Bc4 is 69% for white. Games are from 2004 (the year he got the GM title) to present. All games are longer than rapid time control.

Answer 2

This question hinges strongly on what you mean by 'objectively better'.

If by 'objectively better' you mean you are more likely to reach a position where you can pressure the opponent, then the answer is yes, and the Ruy Lopez is "objectively better" than the Italian.

On the other hand if by 'objectively better' you mean distinguishing between a win and a draw, then the answer is no, since chess is a draw anyway (and almost surely 1. e4 is not "objectively better" than 1. c3 or 1. b4).

Answer 3

If there were an "objective" (by which you probably mean "algorithmic") method to determine the merits of chess moves then chess would lose much of its appeal. As you point out, perfect play leads almost surely to a draw, so many moves must be objectively equivalent, and any preference for one move over another must be based on an assumption of imperfect play. When you speak of "pressuring" the opponent, you are thinking of increasing the possibility of a mistake. The statistics of past games might be considered an "objective" measure of that, but I always feel that resorting to statistics is an admission of ignorance.

Answer 4

Begin by defining your criteria for objectivity. Do they apply to everyone, or are you just asking 'why is a book move better - for a current-day GM -- than another?'

I see chess as a research-problem, and so would add the factor 'how well has this book-move been researched in the past and with what results?'

Compare the question of ulcers. 20 years ago, stomach ulcers were frequently diagnosed as stress-related, and treated accordingly. Then someone researches the question and discovers that they are (to a great extent) caused by bacteria, and can be treated by antibiotics, a treatment that would have been 'objectively worse' five years earlier. (Your have similar situations with Semmelweiss in medicine, or with Knorosov in Mayan language, or with Ventris, who once said that whatever language Linear-B was used for, it wasn't Greek, and then tries it, and finds that he was wrong ...)

What research has been done, and how good is it?

Which probably means that 'one' (unless 'one' is very good at judging that question for chess) won't be able to decide.