How to Improve at Chess — The Plan That Actually Works

There is one question that comes before all the others, and it almost always arrives after a bad run: how do you actually get better at chess? Not "which opening should I play" or "which book should I read", but what to do, in what order, so that in six months the board looks different.

The short answer is that chess improves through diagnosis, not accumulation. The long answer is this article.

Why playing more is not enough

The most common way to plateau is also the most reasonable-looking one: play a lot. Five hundred blitz games a month, two hundred and fifty losses, and by the end of the month the rating is exactly where it started.

The reason is that an unanalysed game teaches you nothing. At most it teaches you that you lost. It does not tell you where it went wrong, or why you picked the plan you picked, or whether the mistake was calculation, evaluation or the clock. Without that information, the next game starts equally blind.

What does teach you is the analysed game. And it does not take many: one well-analysed long game is worth more than fifty blitz games played back to back.

The method, in four steps

1. Find out what is actually costing you

Before studying anything, take your last twenty serious games and sort every loss into one of four buckets:

• I hung something. A piece, a pawn, a mate in two. Calculation or attention.
• I came out of the opening badly. You reached move fifteen undeveloped, planless, or with a broken structure.
• I did not know what to do. The position was balanced, there was no tactic, and you played a move for the sake of moving.
• I lost an endgame that was drawn, or winning.

Count how many fall into each. The biggest bucket is your study programme for the next three months. Not the one that looks most interesting: the biggest one.

This sounds obvious and almost nobody does it. Most players study openings because they are the most entertaining and the easiest thing to find on video, while losing eighty per cent of their games to hanging pieces in the middlegame.

2. Train what is missing, not what you enjoy

With the diagnosis done, the work becomes concrete:

If you hang pieces, your problem is calculation and checking. The fix is not prettier mate-in-three puzzles but pure calculation exercises, plus a routine: before every move, look at every check, every capture and every threat your opponent has. It sounds mechanical because it is, and it works because it is.

If you come out of the opening badly, you do not need more variations: you need to understand the three or four ideas behind the one you already play. What structure it aims for, which pieces want to go where, which break is the key one. A course that explains an opening through its plans is worth a thousand times more than a database of twenty thousand games.

If you do not know what to do in quiet positions, what is missing is strategy: evaluating a position, weak pawns, strong squares, your opponent's bad piece. It is the hardest part to learn alone and the one that changes fastest with good material.

If you lose endgames, start with the three that come up constantly: king and pawn versus king, rook and pawn versus rook, and minor-piece endings with pawns on one wing. That is enough to stop giving away half points.

3. Put the time where it pays

A realistic training week for someone with a job or studies:

• Two hours of tactics and calculation, split into short twenty-minute sessions.
• One hour on whatever came out of the diagnosis.
• One long game — thirty minutes or more — plus its analysis, which takes almost as long as the game.

That is four or five hours. You do not need more, and in fact more hours badly distributed pay less: half an hour of tactics a day beats three hours in a row on Sunday, because patterns stick through spaced repetition, not through accumulation.

4. Measure in a way that does not lie

Rating is noisy: it moves twenty points for reasons that have nothing to do with your level. As a short-term measure of progress, it is poor.

These work better:

• How many times you hung material in your last ten games, compared with the ten before.
• How many games you reached move twenty with a playable position.
• How many endgames you finished with the result the position deserved.

These are numbers that move before the rating does, and that move because of what you did.

The three mistakes that cost the most time

Studying fashionable openings. An elite player's repertoire is built against an opponent who prepares with an engine. Yours has to be built to reach positions you know how to play. One solid, understood opening is worth more than three memorised variations.

Outsourcing analysis to the engine. The engine tells you which move was better; it does not tell you why you chose the other one. Analyse alone first — write down what you thought, what you rejected and why — and only then turn it on. The difference between the two versions is exactly your study material.

Changing plan every three weeks. A mediocre method sustained for six months pays more than a perfect method abandoned three times.

When it is worth having someone look

You can improve alone, and a lot. But there is a point where self-diagnosis stops being enough: you cannot see the mistake you do not know is a mistake. That is where a lesson with someone who looks at your games saves months, because it changes the order of the work — and the order is half the result.

It also works the other way round: to be told that what you are doing is right and that you should keep at it, which is the hardest thing to believe when the rating is not moving.

How to divide a week

The method above has to fit into real life, and in real life almost nobody has more than five or six hours a week. Here is how they divide:

Two and a half hours: long games and their analysis. Two games of thirty minutes or more, with twenty minutes of analysis each. It is half your time and it is the block that produces the improvement; everything else supports it.

An hour and a half: tactics. Fifteen or twenty minutes, six days. Not one long session: pattern memory is built by spaced repetition.

One hour: the subject of the quarter. One only, sustained for three months. Endgames, or one structure, or the middlegame. Changing it every three weeks is the commonest way to study a lot and accumulate nothing.

Half an hour: repertoire. Read your own file and add whatever came up this week. Not studying new theory: maintaining the file.

What is left out of that week says something too: there is no hour assigned to watching content. Videos and annotated games are excellent and they are what you do with spare time, not with the method's time.

The four habits worth more than any method

Underneath all the plans there are four habits, and in practice they explain more of the difference between two players than any choice of material:

Look before you move, always. Your opponent's checks, captures and threats, in that order, before every move. It is boring, it is mechanical, and it is the habit that wins the most games between 800 and 1600.

Finish what you start. A half-watched course, a subject abandoned in week three, an opening changed before it was understood. Chess punishes scattering more than it punishes lack of talent.

Analyse your own games. It has been said three times in this article and it will be repeated whenever necessary, because it is the only activity that works on your errors rather than on humanity's.

And play even when you do not feel like studying. Consistency beats intensity by a distance, and a weak week with two games played is worth more than a perfect week that never happened.

What progress looks like, so you do not quit too early

And a part almost no article mentions, which decides whether you continue: progress does not feel like progress while it is happening.

What it feels like is this: first, a few weeks where everything is harder because you are thinking about new things and that consumes attention. Then a plateau where the number does not move and it looks like nothing worked. And only then a jump that arrives all at once and that, in hindsight, looks as if it appeared on its own.

That order is the normal one, not the exception. The plateau is the phase where what you learned is becoming automatic, and it is exactly where most people conclude their method is no good and change it, a week or two before harvesting.

The way to survive it is not to measure with the rating. Measure with this, once a month:

• How many of your last twenty losses came from dropping material?
• How many won games did you lose?
• Can you explain in one sentence why you lost each of them?

Those three numbers move before the rating does, they move more cleanly, and they are the ones that actually say whether the work is paying.

In short

Improving at chess is not a mystery and it is not talent: it is a short cycle you repeat.

1. You play serious games.
2. You analyse them yourself first, engine second.
3. The analysis tells you what to study.
4. You study that, not something else.
5. You go back and play.

Whoever runs this cycle for six months improves. Whoever plays five hundred blitz games a month does not. And the difference between the two is not how much time they spend: it is where they put it.