Do Chess Puzzles Actually Help You Improve?

Tactical puzzles are the most popular training tool in chess, and also the most badly used. Most players do hundreds a week and see no effect on their games. The problem is not the tool: it is how it is used.

What a puzzle actually trains

A puzzle has a property a game does not: it tells you something is there. You know in advance that a winning move exists, and often even how winning it is.

That makes it excellent for one thing and limited for another:

• Excellent for building your pattern catalogue. Pin, double attack, X-ray, deflection, decoy, back-rank mate. Seeing them a thousand times makes them appear on their own.
• Limited for the skill that decides games, which is noticing that something is there when nobody warned you.

Hence the most common complaint: "I solve 2000-rated puzzles and hang my queen in a 1400 game." That is not a contradiction. They are two skills and you were only training one.

How to do them badly

This is the default pattern and almost all of us start here:

1. You look at the position for three seconds.
2. You play the move that looks right.
3. If it works, next. If it does not, "ah, of course", next.

This trains recognition and untrains calculation, because being wrong costs nothing. In a game, playing what looks right costs the game.

How to do them properly

The version that works is slower, more uncomfortable and considerably less fun:

One. Do not move until you see the end. Calculate the whole line, with your opponent's best reply at every move, until you reach a position where you can say "here I am winning and it is obvious why". If you have not reached that position, you have not finished calculating.

Two. Say the line out loud, or write it down. It feels ridiculous and it is the step that changes the most. It forces the calculation to be explicit instead of a cloud of intentions.

Three. When you fail, stay there. The instinct is to move on. The value is precisely there: which of your opponent's moves did you not see? Did you dismiss it or never look at it? The answer to that question is your blind spot, and it shows up in your games too.

Four. Redo the ones you failed, a week later. A pattern you got wrong and never saw again is a pattern you will get wrong again.

Ten puzzles done this way take twenty minutes and are worth more than a hundred in the same time.

The missing link: the check

Puzzles do not close the gap with real games on their own. A routine does, and it is deliberately boring. Before every move:

• What checks does my opponent have?
• What captures?
• What does their last move threaten?

Three questions, five seconds. It is the only thing that turns "I can recognise a pin" into "I do not get pinned".

Most pieces hung below 1600 are not lost through failing to see the pattern: they are lost through not looking.

Which puzzles to pick

At your level, or barely above. A puzzle you cannot solve in five minutes teaches you nothing you can use tomorrow. The useful zone is the one that is hard and that you get.

From real games. Composed problems are beautiful and strange; positions from real games have the shape of what is going to happen to you.

And occasionally, with no guaranteed solution. There are training sets where the right answer is "there is no tactic, play positionally". They are uncomfortable and they are the ones that most resemble playing.

The seven motifs that cover almost everything

Tactics look infinite and they are not. Almost everything that comes up in a game below 2000 grows out of seven motifs, and knowing them by name changes how you look at a position: you stop hunting for "something" and start hunting for specific things.

The pin. A piece cannot move because something more valuable sits behind it. It is the most frequent motif of all and the one that decides the most games, because a pinned piece stops defending whatever it was defending.

The skewer. The pin in reverse: the valuable piece is in front and moving it drops what was behind.

The double attack. One piece attacks two things at once. The knight does it better than anyone — the famous fork — but the queen, the pawn and even the king do it too.

The discovered attack. One piece moves and uncovers another's attack. It is the hardest motif to see because the threat is not made by the piece that moved.

Overloading. A piece is defending two things and cannot manage both. You force it to choose.

Deflection. You drag the defending piece out of the way, usually by giving something up.

The in-between check. Before recapturing, a check gets inserted that changes the whole calculation. It is the one that ruins the most "won" games, because your opponent has it and you did not see it.

Once you can name the motif of every puzzle you solve, practice changes in nature: you stop accumulating solutions and start accumulating patterns, which is the only thing that transfers to a position you have never seen.

How many a day, and for how long

The short answer is fifteen or twenty minutes a day, every day. Not an hour on Sundays: tactics are pattern memory, and pattern memory is built by spaced repetition, not by concentrated volume.

Fifteen minutes fits between five and ten puzzles done properly, which is far better than thirty done badly. And "properly" has a concrete definition: decided before moving. If you move to see what happens, you are not training calculation, you are training yourself to move fast.

For how long? Tactics are the first thing that pays and the first thing that plateaus. The first three months the effect is enormous — that is the phase where you stop losing pieces. After a year of daily practice the marginal return drops considerably and it is worth moving part of that time to analysing games and to endgames. Tactics are never abandoned, but they stop being the main course.

Puzzles and calculation are not the same thing

This is the distinction most people skip, and it explains why somebody with three thousand puzzles solved still calculates badly in a game.

A puzzle tells you there is something there. You know, before looking, that a winning move exists in that position. That is exactly the information you do not have in a game, and it is half the work.

A puzzle has no cost of error. You fail, you see the solution, you move on. In a game a botched calculation costs you the result, and that pressure changes how you think.

A puzzle is clean. There is no deciding whether it is worth looking: somebody already decided that it is.

So it is worth adding a second exercise, done on the same kind of positions, that trains the other half: take any position from a game — not knowing whether anything is there — give yourself three minutes to decide on a move, write it down, and only then look. That one does resemble playing.

How to know whether they are working

Not by the number of puzzles, and not by the app's rating, which rises with speed as much as with accuracy. Three signs that do mean something:

You lose fewer pieces to oversights. It is the first thing to move and it is easy to measure: count, over ten games, how often you lost material to a single move.

You see your opponent's tactics, not only your own. It is the most important sign and the slowest to arrive. Puzzles train you to attack; half the real benefit lies in noticing when the one about to win something is the other player.

You get them during the game, not afterwards. If, while analysing, you say "oh, there was a fork there", the pattern is learned but not automatic. When the pattern jumps out at you during the game, it worked.

If after three months of daily practice none of the three has moved, you are almost certainly doing puzzles at guessing speed. Halve the pace.

In short

Puzzles work, and work well, on three conditions: calculate to the end before moving, stay on the one you failed, and pair them with the checking routine that carries them into games.

Without that, they are an entertaining way of looking at positions. With it, they are the tool that most quickly reduces the number of games lost to a single move — which, below 1800, is almost all of them.

Related questions

How many puzzles does it take to see a change? It is not a question of total quantity but of consecutive days. At fifteen minutes daily, the change shows between week four and week six, and the first thing to change is not that you win more but that you drop fewer pieces.

Is solving puzzles against a clock useful? For training recognition speed, yes, and it is useful once the patterns are in place. Before that it does harm: it trains guessing. The rule is simple: if you still fail more than one in five, take the clock away.

What about mate-in-two puzzles? They are the best to start with and the least like a real game. They are excellent for learning to see every move in a position, which is the underlying skill; afterwards it is worth moving to material-winning puzzles, which is what actually comes up.

Can I replace puzzles with fast games? No. A fast game presents very few tactical positions per hour and presents them without warning, so as pattern training it is far less dense. They are different things and both are needed.