How to calculate candidate moves in chess: checks, captures and threats before you commit

Short answer

Calculating in chess is two jobs that most club players collapse into one: building the list of moves worth calculating, and calculating each of them. This guide covers the scan that produces the list — checks, captures and threats, in that order, yours first and then your opponent's — three positions where it decides the game, and how to drill the routine until it runs on its own.

The essentials in a few lines

• Hanging a piece is rarely a calculation error. It is a list error: the refutation was on the board and was never considered.
• The order checks, captures, threats is not arbitrary — it runs from most forcing to least forcing, and the most forcing lines are the ones you can calculate to the end with the fewest branches.
• The half nobody does is the second one: running the same scan for your opponent, in the position that exists after the move you intend to play.
• Width before depth: two moves ahead with a complete list beats eight moves down a single line, which is the signature error of amateur calculation.
• The final check is not calculation. It is a five-second question before you touch the piece: after this move, what is left undefended?

The error is in the list, not in the calculation

Watch what happens when you review the game with an engine. Nearly every time, the reaction is not «I couldn't see it to the end» — it is «I never even looked at that move». Those are two different problems, and they take different training: deep calculation is trained with long exercises, candidate selection with a scanning routine, which is the subject of this guide.

The idea of candidate moves comes from Alexander Kotov, who described calculation as a tree walked exactly once, with no going back to branches already visited. The tree part has aged badly — strong players backtrack constantly, and saying so is uncontroversial today — but the part that matters here still stands: before calculating deeply, you have to decide what will be calculated, and that decision deserves a method of its own.

The method is a three-step scan, known by the acronym CCT: checks, captures, threats. It does not pick the move for you and it does not replace positional judgement. It does exactly one thing, and it is the thing that is missing: it guarantees that no forcing move in the position stays off the list, for either side.

Step 1: checks

Cheap does not mean good. Most checks in any position are bad moves, and checking for the sake of checking is a beginner's habit. But listing them costs seconds and answers the question that matters: is there a forcing sequence here at all? If there is, it almost always starts with a check or a capture.

The board below is the line known as the Fried Liver, and it shows precisely what the check buys. After 5...Nxd5?, the sacrifice 6.Nxf7 only works because of what follows: 7.Qf3+ is the move that drags the black king into the centre and keeps the initiative. Without that check, the sacrifice is just a piece down.

One honest note from the engine: after 8.Nc3 the evaluation settles around +0.7 — a clear advantage for White, a knight for two pawns and a black king stuck in the open, but not a forced win. And Black's real mistake was 5...Nxd5; the defence 5...Na5 keeps the position playable.

Step 2: captures

Legall's trap is the classic example and still turns up in club games two centuries later. With the black bishop pinning the knight on f3 against the queen, White plays 6.Nxe5, which any quick glance dismisses: the knight moves, the queen falls.

And it does fall. What the quick glance fails to do is keep counting: after 6...Bxd1 7.Bxf7+ Ke7 8.Nd5 is mate, and the queen Black won never gets to matter. The capture scan finds 6.Nxe5 precisely because it does not filter by first impression — it lists every capture and judges afterwards.

There is a detail most versions of this trap leave out, and it matters for anyone learning the method rather than the trick: even with the best defence, 6...Nxe5 7.Qxh5, White is a pawn up. Black's mistake was not walking into mate; it was 5...Bh5, and the engine already gives White almost two pawns there. The mate is a bonus, not the refutation.

Step 3: threats — and the scan from the other side

A threat is a move that wins nothing yet, but wins something next move if nobody does anything about it. It is more expensive to list than checks and captures, because it is not forcing — which is why it comes last. In exchange, it is where most losses come from once you have stopped hanging pieces outright.

The Noah's Ark trap in the Ruy Lopez is the portrait of this. White regains the pawn with 8.Qxd4, a natural move that centralises the queen and seems to settle the position. The engine disagrees emphatically: the evaluation jumps from level to nearly three pawns for Black, because 8...c5 and 9...c4 build a pawn box around the bishop on b3 and it has nowhere to go.

None of Black's moves is a check, and none is a capture. They are threats — and that is exactly why 8.Qxd4 gets played so often. A player who scans only the opponent's checks and captures finishes the count convinced everything is fine, two moves before losing a piece.

Width, depth and when to stop

Engines call this quiescence search, and the idea transfers well to human calculation: stopping in the middle of a capture sequence produces a meaningless evaluation, because the position being judged is not a position that will ever appear on the board. If your line ends with a piece under attack, you have not finished.

The symmetric error is calculating fifteen moves of one beautiful line and playing it without ever looking at the second option. Listing three candidates and calculating each until it goes quiet costs about the same time and gives a far better result. Depth without width is the signature of amateur calculation.

It is also worth separating calculation from evaluation. At the end of each line you have to say whether the resulting position is good, and that is not calculated: it is judged, from material, king safety and piece activity. Perfect calculation with a wrong evaluation at the end produces exactly the same loss as bad calculation.

• Stop when there is no check, capture or threat pending in the final position of the line.
• Before going deep on a line, make sure the list has at least two candidates.
• If a line runs past six moves, be suspicious: either the position is extremely forcing, or you are following a wish.
• Write the evaluation at the end of the line in words — «pawn up, endgame»; «level, exposed king» — not as a feeling.
• Only then compare the candidates. Comparing is the last step, not the first.

How to drill the scan in four weeks

In week one, use tactics you would solve anyway and force yourself to say every check and every capture in the position before answering. It feels awkward and it feels like a waste of time — which is the point: the routine only becomes automatic when it runs in positions where it is not needed.

In weeks two and three, move the routine into your own slow games, and run the opponent's scan after choosing your move but before playing it. Mark the games where you ran the routine on every critical move: that count is the number that matters, not the result.

In week four, measure. Review the month's games and count how many serious errors were list errors — opponent moves you had not considered — and how many were evaluation errors. The ratio tells you what next month's training is, and it differs from player to player.

Frequently asked questions

Openings mentioned in this guide

• Italian Game (Giuoco Piano) (C50)
• Ruy Lopez (Spanish Opening) (C60)
• Scotch Game (C45)