How Deep Should You Study a Chess Opening?

The question everyone asks wrong

“How deep should I know my openings?” is usually answered with a number. Ten moves. Fifteen at a push. Twenty if you play the Najdorf.

A number is the wrong shape of answer, because depth is not the thing that costs you. What costs you is the number of positions, and depth and positions are not the same thing at all. One extra move of depth in a line with four plausible opponent replies is four extra positions. The same move in a forced sequence is one.

The useful question is not how deep. It is how likely.

Your moves are free, theirs are not

Here is the piece of arithmetic that settles it.

When it is your move, you play the move you prepared. The chance of reaching the next position is unchanged, because you decide.

When it is your opponent’s move, the chance divides. If they have a reply that occurs in 45 per cent of games, then after that move you are in a position you reach in 45 per cent of the games that got that far. Every one of their choices does this again.

Four opponent choices at plausible rates and you are looking at a position you reach in roughly one game in eleven. Not one in three, which is how it feels while you are adding the moves.

What that means for move fifteen

Depth is not expensive because the moves are hard. It is expensive because it is compound.

Prepare a line to move eight and you are covering positions you meet regularly. Extend the same line to move fifteen and most of what you added sits behind three or four opponent decisions, each of which had to go your way. You are spending real review time, every week, on positions that turn up a few times a year.

There is a second cost that is easier to miss. Those deep positions are also the ones you forget fastest, because you reach them least often in your own training. So the deepest part of a repertoire is simultaneously the least likely to be needed and the most likely to have decayed by the time it is.

The exception, and it is a real one

Forcing lines break the arithmetic, in your favour.

If your opponent has only one move that does not lose, then the probability does not divide. It stays where it was. A sharp variation where each side has a single move keeping the balance can be twenty moves deep and still be worth every one of them, because reaching move twenty is as likely as reaching move ten.

This is why the same advice does not fit every opening. In a Najdorf mainline, depth is cheap and necessary. In a London System, where the opponent has five reasonable setups at almost every point, the same depth costs five times as much and buys far less.

So the honest rule is not a number of moves. It is: prepare deep where the opponent has no choice, and stop early where they have many.

A rule you can actually apply

At the board you will not be multiplying percentages. Use this instead.

Stop adding to a line when you reach a position where you could explain the plan without help. Which pawn breaks matter, where the pieces belong, what the other side wants. If you can say that, the next few moves will follow from understanding and do not need to be memorised.

If you cannot say it, the answer is not more moves. It is time with the position, an engine and a database, until the ideas make sense. Adding another five moves to a line you do not understand just moves the point of failure later.

Where depth still pays

None of this is an argument for shallow preparation. It is an argument for putting depth where it earns its keep.

The lines worth going deep in are the ones you meet most, which are usually the ones nearest the root. Most players have this backwards. They know their pet variation to move eighteen and have never decided what to do against the second most common reply on move three.

If you want to find those gaps rather than guess at them, the method is in how to build a chess opening repertoire. And once you know how deep to go, the next constraint is how much you can hold at all, which we work through in how big a repertoire should actually be.