Candidate Moves

Candidate moves are the shortlist of moves worth calculating in a position — the two, three, or four options you identify before analyzing any single line deeply. The concept was popularized by Alexander Kotov in his classic Think Like a Grandmaster (1971), and it remains the single most useful piece of thinking technique ever put into words. In my coaching, the difference between a 1400 and a 1900 is rarely raw calculating power — it is that the stronger player considers the right moves in the first place.

What Are Candidate Moves?

In any middlegame position there are typically thirty to forty legal moves. Nobody — not a grandmaster, not a world champion — calculates all of them. Instead, a trained player glances at the position and their intuition surfaces a handful of moves that deserve attention: a capture that changes the structure, a check that forces matters, a maneuver that improves a bad piece. Those are the candidate moves, and everything that follows — calculation, evaluation, decision — happens only within that shortlist.

Kotov's central insight was that this selection step should be conscious and complete: before you calculate anything, you fix the full list of candidates in your mind, and only then analyze them one by one. The quality of your final move can never exceed the quality of your candidate list — if the best move never made the shortlist, no amount of brilliant calculation will find it. This is why I tell students that candidate generation is not a preliminary to thinking; it is thinking — it anchors the six-step method in my guide on how to find the best move. The lines you crunch afterwards are just verification.

Why Candidates First, Calculation Second

The most common thinking error I see at every level below master is the plunge: a move catches your eye, you dive five moves deep into it, discover a problem, surface, spot another move, dive again. Twenty minutes later you have analyzed two moves to exhaustion, your clock is bleeding, and the third option — often the best one — was never examined at all.

Breadth before depth fixes this. Spend the first minute of a critical think doing nothing but listing candidates — no calculating, no evaluating, just collecting. It feels unnatural, because the mind wants to chase the first shiny variation it sees, but the discipline pays for itself three ways:

You stop missing better options. The winning move is frequently the second or third candidate, not the first. If you plunge into candidate number one, candidates two and three may never exist for you.
You manage the clock. A complete list lets you budget: "three candidates, ten minutes, roughly three minutes each, then decide." The plunge has no budget — it ends when you are exhausted, usually deep in time trouble.
You calculate less, not more. Comparing candidates at shallow depth often eliminates two of them in seconds — one drops a pawn, one allows an obvious fork — and only the survivor needs real depth.

The practical takeaway: when the position is critical, force yourself to say "my candidates are…" and name them all before touching a single variation in your head.

How to Generate Candidate Moves

A candidate list is only as good as the generator behind it, and the generator has a strict priority order.

Checks, captures, and threats — always, every position, no exceptions. These are the forcing moves, and they come first for a simple reason: they constrain the opponent's replies, which makes them both the most likely sources of immediate tactics and the cheapest to calculate. Even in a quiet position, spend ten seconds confirming what every check, every capture, and every direct threat does — for both sides. Most tactical oversights are not calculation failures; they are forcing moves that never made the list. A capture that "obviously loses material" sometimes turns out to be the sacrifice that wins the game.

Then positional candidates. Once the forcing moves are catalogued, ask the strategic questions:

Which of my pieces is doing the least? The move that improves your worst piece is a standing candidate in every quiet position — reroute the passive knight, activate the rook, trade the bad bishop.
What are my pawn levers? A pawn break changes the structure and opens lines, and in closed positions it is usually the only path to progress. I organized much of Chess Structures around this question, because in most structures the correct breaks are known in advance — know your structure, and half your candidate list is pre-written.
What does my opponent want to play? The move that stops their idea — prophylaxis — belongs on the list even when it looks passive. Some of the strongest moves in chess do nothing except delete the opponent's only plan.

The opponent's intentions are themselves a generator: asking "what would they play if it were their move?" surfaces candidates you will never find by staring at your own pieces.

How Many Candidates — and Kotov's Tree Versus Reality

For most positions, two to four candidates is the honest number. Fewer than two means you probably stopped looking too early; more than five means you are not pruning and will drown in analysis. The count flexes with the position: in sharply forcing positions — a piece hanging, a mate threat on the board — there may be exactly one candidate, and you should not manufacture alternatives to feel thorough. Rich, tense middlegames are the opposite case: several breaks, regroupings, and forcing tries genuinely compete, and the player who saw three ideas beats the player who saw one. That is where tunnel vision gets punished.

Kotov's famous prescription went further than the shortlist: he described an analysis tree in which you examine each candidate exactly once, calmly and completely, never returning to a line already analyzed. It is a beautiful ideal, and generations of players have felt guilty for failing to live up to it — because nobody actually thinks that way, grandmasters included.

Here is how I actually think during a serious game. I list the candidates. I do a quick pruning pass — shallow looks that kill the obviously deficient options. Then I identify the critical candidate, the one the position hinges on, and I dive into it deeply — and yes, I jump back out when something feels wrong, check a detail in another line, and return. The jumping around that Kotov forbade is not indiscipline; it is how comparison works, because the evaluation of line A often depends on what you just learned in line B. Finally, once I have chosen, I run a blunder-check. The structure — candidates, prune, deep-dive, verify — is Kotov's real legacy. The rigid one-pass tree is not.

The Blunder-Check: The Last Candidate Filter

After you have chosen your move and before your hand touches the piece, run one final scan: what does this move allow? Not "is my idea good" — you have already decided that — but "which of my opponent's checks, captures, and threats does this move newly enable?" Does it undefend something? Does it walk into a knight fork or leave the back rank loose? Does it release a pin that was holding their position together?

This takes ten to fifteen seconds, and it is the cheapest rating improvement in chess. A huge share of decisive mistakes at club level are not deep miscalculations but one-move oversights — a blunder committed in a position the player understood perfectly, because the final safety scan never happened. The blunder-check is the last gate your candidate must pass: it earned its place on the list, survived calculation, won the comparison — now it must also survive the question every strong player asks by pure reflex. When I train students, making this scan automatic is worth more than any opening file I could give them.

Candidate Moves in Defense

Defensive candidates are the most-missed candidates in chess, for psychological reasons. Under attack, players collapse into one narrow question — "how do I stop the threat?" — and only passive moves make the list. But the defensive shortlist must include the aggressive options too:

Counterattack. Sometimes the best answer to a threat is a bigger threat. Generating counterplay instead of parrying passively is often the only defense that holds, because a defender with threats forces the attacker to spend moves defending too.
The zwischenzug. Before automatically recapturing or retreating, check the in-between forcing move. The zwischenzug — an interposed check or threat played before the "obligatory" reply — refutes an enormous number of plausible-looking combinations, and it is invisible to anyone who only lists compliant moves.
Returning material. Giving back a pawn or the exchange to break the attack is a legitimate candidate that amateurs almost never write down, because it feels like admitting failure. It is frequently the move an engine finds instantly.

When defending, deliberately widen the list: one passive defense, one counterattacking try, one forcing resource. If only the passive one exists, at least you now know the position's true character.

Where Amateur Candidate Generation Fails

Four failure patterns account for most of the bad shortlists I see in students' games:

Only "my plan" moves make the list. The player decided three moves ago to attack the kingside, so every candidate serves that plan — even after the position changed and the plan died. The list must be regenerated from the current position, not inherited from an old intention.
The opponent's replies are never used as generators. Strong players extract candidates from the opponent's ideas: their threat suggests a prophylactic candidate, their loose piece a tactical one. If you never ask what they want, you are generating from half the board.
"Ugly" moves are never considered. Moves that violate general principles — an exchange sacrifice, a retreat in mid-attack, a pawn move in front of your own king — get filtered out before conscious thought. These are exactly the moves that decide games at higher levels, because your opponent filtered them out too.
Only forcing moves are considered. The mirror image: in a winning attack, players list only checks and captures, when the strongest continuation is often a quiet move that brings one more piece to the party or cuts off an escape square.

The common thread is a candidate list built from bias instead of from the board. The fix is the same discipline every time: forcing moves first, then piece improvement, breaks, and prophylaxis — and no move rejected before it has been consciously seen.

How to Train Candidate Moves at ChessMind AI

Candidate generation is trainable, but only if you make the selection step explicit. Change how you solve: with every position in our puzzle collection, list your candidates — all of them — before calculating a single line. Solving this way is slower and dramatically more valuable, because it trains the exact skill you use at the board. Our positional puzzles are built around multiple-choice positional decisions — several reasonable candidates, one best — which is candidate comparison in its purest form, with no tactic to bail you out. The evaluation trainer sharpens the judgment that prunes your list: knowing quickly who stands better tells you whether to hunt for forcing candidates or quiet ones. And to learn whether candidate selection is actually your bottleneck — as opposed to calculation, openings, or endgames — take the GM diagnostic; it pinpoints where your thinking breaks down so you train the real weakness.

Frequently Asked Questions

What are candidate moves in chess?

Candidate moves are the small set of moves — usually two to four — that a player judges worth calculating in a given position. Instead of analyzing all thirty-plus legal moves, you first select the promising ones, then calculate only those. The term was popularized by Alexander Kotov in Think Like a Grandmaster.

How many candidate moves should I look for?

Two to four in a typical position. Highly forcing positions may have only one genuine candidate — do not invent alternatives just to feel thorough. Rich, complex middlegames may justify a fifth, and they are exactly where considering only one move gets punished.

What order should I check candidate moves in?

Generate forcing moves first — every check, capture, and threat, for both sides — because they constrain replies and hide most tactics. Then add positional candidates: improving your worst-placed piece, pawn breaks, and prophylactic moves that stop the opponent's plan. Calculate the most critical or most forcing candidate first.

Is Kotov's analysis tree method realistic?

As a discipline, partly; as a literal description, no. Kotov taught that each candidate should be analyzed exactly once, without jumping between lines. In reality even grandmasters shuttle between variations, because comparing lines is how evaluation works. Keep his real lessons — complete candidate lists and breadth before depth — and drop the guilt about revisiting lines.

How do I stop missing my opponent's moves?

Treat your opponent's checks, captures, and threats as part of your own candidate scan, and finish every decision with a blunder-check: "what does my chosen move allow?" That ten-second question, asked before every move without exception, removes more one-move oversights than any amount of extra calculation depth.

Do candidate moves matter in blitz?

Yes — you just compress the process. In fast chess the candidate scan (forcing moves plus one positional idea) often is the whole thought, and the blunder-check becomes even more valuable because there is no time to recover from an oversight.

Calculation — what happens after the shortlist is set.
Quiet Move — the candidate every routine scan misses.
Zwischenzug — the in-between resource that rewrites your tree.
Blunder — what the final candidate filter exists to prevent.
Combination — where disciplined candidate search pays off.
Prophylaxis — the opponent's-ideas scan that feeds your list.