Analyse a chess position in a fixed order, before you calculate anything: count material and name the imbalance, check both kings, read the pawn structure and weak squares, rate piece activity and find your worst piece, list every check, capture and threat for both sides, then pick two or three candidate moves and calculate only those. Evaluation tells you what to look for. Calculation confirms it. Skipping the opponent's half of the list is what costs club players most of their games.
Interactive lesson — The overloaded defender: one piece, two jobs: Play it on the board below, and make the key move yourself. Black’s rook on e8 is doing two jobs at once. It guards the back rank, where the king has no air, and it is the only piece defending the knight on e5. Rxe5. You take the piece the rook is defending, and now the rook has to pick which job it keeps. Black saves the knight, and the back rank is left with nobody watching it. Rd8, check. The king is walled in by its own pawns, so the rook has to come racing back. The only legal move on the board, and it costs Black the rook. Rxe8 is mate. The rook could not be in two places at once, and that is the whole idea.
Most players think analysing a position means calculating. It does not. Calculation is the last of six steps, and the five before it are what stop you from calculating the wrong moves. The evidence for that ordering is visible in Chess DNA's published dataset of 6,288 real amateur games, every move re-checked by Stockfish 17 at depth 18: a move that costs about 10% of your win probability gets labelled an inaccuracy and one that costs 20% a mistake, and in the amateur games we see, those drops overwhelmingly follow moves that were calculated carefully but chosen from a wrong assessment of the position. This guide gives you the checklist, a worked example, and an honest account of where engine numbers help and where they mislead.
There is a reason strong players stare at a new position before their eyes start moving pieces around. Calculation is expensive: it eats clock, it is error-prone, and it only tells you where one specific line leads. Evaluation is cheap, and it answers the more useful question, which is what kind of move does this position want. Once you know the position wants a slow queenside build-up, you stop calculating kingside sacrifices.
The habit that separates a 1200 from a 1700 is not calculating deeper. It is asking two questions before every quiet move: what are the permanent features here, and what does my opponent want to do next? The second is the one almost everyone skips, because looking at your own ideas feels productive and looking at your opponent's feels like pessimism.
One caveat. Forcing positions override everything. If there is a check, a capture of a defended piece, or a mate threat on the board, concrete calculation comes first, because tactics beat assessments. The checklist is for the quiet moves, where nothing is forced and drift is the real enemy.
Run these in order. The whole pass takes under a minute once it is a habit, and about three minutes the first few times you try it.
Notice that five of the six steps require no calculation at all. That is the point: you are building an accurate picture of the terrain, so the calculation you do afterwards is aimed.
Counting material is the easy half. Naming the imbalance is the useful half. "Equal material" hides three very different positions: rook and pawn against two knights, bishop pair against bishop and knight, and queen against three minor pieces are all roughly level on paper and completely different in practice. So say it in words: "material is level, but I have the bishop pair and he has an extra pawn on the queenside".
Two practical rules. First, minor-piece quality depends on the structure, so an extra pawn in a blocked position is worth much less than the same pawn in an open one. Second, when you are up material, your evaluation should push you toward simplification, and when you are down, toward complication. Getting the material verdict right therefore sets the whole direction of your play, not just the score.
King safety outranks almost every other positional factor, because it is the only one that can turn into a loss in three moves. For each king, check four things: is it castled, is the pawn cover intact, are there open files or long diagonals pointing at it, and how many of its own pieces are within one move of defending it. A king with three pawns in front of it and a knight on f8 is safe even in a sharp position. A king with an advanced h-pawn and no minor pieces nearby is a target even if the position looks quiet.
The comparison matters more than the absolute. You are not asking "is my king safe?" but "whose king is safer, and by how much?" That single comparison decides whether you should be opening the position or keeping it closed, and it is the reason so many attacks fail: the attacker opened lines while his own king was the more exposed of the two.
Pawns are the most permanent feature on the board, since they cannot move backwards. That permanence is what makes them the right foundation for a plan. Scan for isolated pawns, doubled pawns, backward pawns that cannot advance and cannot be defended by a pawn, and passed pawns. Then scan for holes: squares in each camp that no pawn can ever guard again. Those holes are where your knights belong.
The structure also tells you where to play. A pawn majority on one wing, or more space on one side, means that is your side of the board, and the pawn break that opens lines there is usually your plan. If you want the full treatment of turning structure into a plan, chess middlegame strategy walks through the pawn breaks that each common structure hands you.
Count good squares, not pieces. A rook on an open file, a bishop on an unobstructed diagonal and a knight on a protected outpost are worth far more than the same three pieces shuffling behind their own pawns. Check whether rooks are on open or half-open files, whether either side has a bishop hemmed in by its own pawn chain, and whether any piece is out of play entirely.
Then name your worst-placed piece. This is the single most productive question in quiet positions, because the answer is usually your move. It is also a safety net when you cannot find a plan: improving the worst piece never backfires the way a premature attack can. Ask the same question about your opponent, too, since it tells you which trades help you. Trading your bad bishop for his good knight is a positional gain even when the engine's number barely moves.
Now the step that most club players skip. List every check, every capture and every threat available to your opponent, then the same for yourself. The framing that makes this stick is a question: if I passed and did nothing at all, what would my opponent do? Whatever that move is, it is the threat you have to answer or accept knowingly.
This is where most rating points are lost and won. Deep strategic misjudgements are rare below 1800. One-move oversights are constant, and they are all catchable by a five-second forcing-moves scan. If you want the drill that trains the scan into a reflex, chess calculation training covers the candidate-move and forcing-move routine in detail, and chess visualization training covers seeing the resulting positions clearly enough to judge them.
Only now do you calculate. Pick two or three candidate moves that fit the assessment you just built, and calculate each one until the position goes quiet, not until you run out of ideas. At the end of each line, stop and run a short version of the same checklist on the resulting position, because a variation is only as good as the evaluation at its end.
If none of your candidates improves your assessment, that is information: look for a fourth candidate instead of talking yourself into the least bad of three. And keep the calculation proportional. In a quiet position, three moves deep with an accurate final assessment beats seven moves deep with a hand-wave at the end.
Take a position from a common Italian Game line: 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.c3 Nf6 5.d4 exd4 6.cxd4 Bb4+ 7.Bd2 Bxd2+ 8.Nbxd2 d5 9.exd5 Nxd5 10.0-0 0-0. Black to think. Here is the checklist, run out loud.
Material: dead level, and the imbalance is small: the dark-squared bishops have been traded, so each side has one bishop and two knights. Nothing here pushes the position in either direction.
Kings: both castled short, both pawn shields untouched, no open file pointing at either king. Verdict: equal, and neither side has a reason to rush.
Pawn structure: this is the real story. White has pawns on a2, b2, d4, f2, g2, h2, which means the d4 pawn is isolated: no c-pawn and no e-pawn can ever defend it. Black has no d-pawn or e-pawn, so the c- and e-files are half-open for both sides. The isolated d-pawn cuts both ways. It gives White extra central space and control of the c5 and e5 squares, and it gives Black a permanent target and a beautiful blockading square on d5.
Piece activity: White is slightly more active, with a bishop on c4 aiming at f7 and the d5 knight, and a knight on d2 that wants b3 or e4. Black's knight on d5 is excellently placed, on the blockade square in front of the isolated pawn. Black's worst piece is obvious: the bishop on c8, still at home with the rook on a8 doing nothing behind it.
Threats: nothing forcing, but White's Bxd5 is available and White's ideas of Nb3, Re1 and a queen to b3 or d3 all point at f7 and d5. If Black passed, White would improve with Nb3 and Re1 and start pressing.
Candidates: the assessment says solve the worst piece while keeping the blockade, so the natural candidates are 10...Be6, 10...Nb6 and 10...Nce7. Now a short calculation. 10...Be6 develops the bad bishop and adds a second defender to d5. If White answers 11.Bxd5 Bxd5, Black's bishop takes over the blockade square and the isolated d4 pawn becomes a long-term liability, which favours Black. So White will keep pieces on with something like 11.Ne4 or 11.Qb3, and Black continues untangling. 10...Nb6 gains a tempo on the bishop but abandons the blockade square, which contradicts the structural verdict from step three. That contradiction is exactly the kind of thing the checklist is for: without it, the tempo-gaining move looks obviously good.
Notice what happened. Five assessments took maybe forty seconds, produced three candidates, and rejected one of them on structural grounds before any calculation. That is the workflow.
An engine's evaluation is a single number in pawn units: +0.5 means White is about half a pawn better, -1.5 means Black is clearly better, and a mate score replaces the number with a move count. Modern engines derive that number from a neural network rather than a hand-written list of features, so it is not a literal sum of "material plus king safety plus structure". But it is still an answer to the same question your checklist asks, and the two map onto each other closely enough to be useful. Wikipedia's overview of evaluation functions is a good primer on the machinery.
Where engines are unarguably better: anything concrete. In a position with tactics, the engine's number is the product of a forcing line it has verified, and if it disagrees with you, you are wrong. That is why an engine pass over your games is the fastest way to find real errors, and it is what tools like the Stockfish-backed review in chess engine analysis are for.
Where human judgement still earns its keep: quiet positions. An engine will sometimes hand you a +0.3 based on a fleeting activity that no human can hold, or shrug at a structural weakness that will decide the game in thirty moves because it can hold the position with a defensive sequence you would never find. The practical rule is to read the number as a headline and the main line as the story. If you cannot follow the engine's line, its number has not taught you anything yet. And treat small differences as noise: below roughly a third of a pawn, at any depth, the engine is telling you the position is balanced, not that one move is better than another.
The checklist step you are getting wrong is fairly predictable from your rating.
Step five, almost exclusively. Material is counted, kings are ignored, and the opponent's threats are not looked at. The fix is mechanical, not conceptual: before every move, name the opponent's most dangerous reply out loud.
Over-weighting material and under-weighting king safety. This is the band that grabs a pawn on the edge while a file opens toward its own king. It is also where "activity" gets confused with "moving my best piece again". Ask which of your pieces is doing the least work.
Structure. Players in this band can calculate reasonably and rarely hang pieces, but they trade into structurally bad endgames, keep the wrong minor piece, and open lines on the wrong wing. The tell is a game where nothing dramatic happened and you were simply worse by move 30.
Candidate selection. The assessment is usually right and the calculation is usually sound, but the third candidate never got looked at. Widening the candidate list costs clock and gains accuracy, which is a trade-off worth making in critical positions only.
The general way to catch your own band-specific error is to check your evaluation against reality before checking it against an engine. Write your assessment down, play the move, and see whether the game went the way your assessment predicted. Recurring mismatches are your weakness, and they show up as a shape across many games rather than a single blunder, which is the argument for reviewing games in batches instead of one at a time. How to recognize chess patterns covers the recognition side of that loop.
Two honest recommendations, for two different jobs.
If you want to poke at a single arbitrary position, set it up on the free Lichess analysis board. You can drag pieces onto an empty board or paste a FEN, run an engine in the browser, and try variations, all without an account. Chess DNA deliberately does not have an arbitrary-position or FEN setup board, so for that specific job the Lichess board is the better tool and we would rather say so.
If you want to know which kinds of positions you personally misjudge, that is a different question, and it needs your own games rather than a made-up position. Chess DNA imports your games from Chess.com or Lichess, runs Stockfish 17 at depth 18 over every move, and then groups the errors into recurring patterns across eight skill dimensions, so instead of "you dropped 1.4 pawns on move 22" you get "you consistently misjudge king safety when the position opens". That is the diagnosis the checklist above cannot give you about yourself, because your blind spots are invisible from the inside.
Analyse your own games free with Chess DNA →
For the manual review process that pairs with it, see how to analyze your chess games, and for a comparison of the board tools available for stepping through a position, chess analysis board covers the options.
Work through a fixed checklist in the same order every time. Count the material and name the imbalance, check king safety for both sides, read the pawn structure and the squares no pawn can defend again, rate piece activity and find the worst-placed piece on each side, list every check, capture and threat for both players, and only then pick two or three candidate moves to calculate. The order matters: the first five steps tell you what kind of move the position is asking for, so the calculation in step six has something to aim at.
Chess evaluation is a judgement of who stands better and by how much. Engines express it as a number in pawns, so +0.5 means White is about half a pawn better and -2.0 means Black is winning, while a mate score is written as a move count instead. Human evaluation is the same judgement expressed in factors: material, king safety, pawn structure, piece activity and immediate threats. The number is a summary of those factors, not a replacement for them, which is why strong players still say the position in words before they trust the digits.
Evaluate first, then calculate. Evaluation tells you which candidate moves are worth calculating at all, so calculating first means burning clock on random variations. The one exception is forcing positions: if there is a check, a capture or a mate threat on the board, look at those concrete moves immediately, because tactics override any positional assessment. A useful habit is a ten second scan for forcing moves, then a thirty second evaluation, then calculation of two or three candidates.
Yes. To poke at one arbitrary position, the free Lichess analysis board lets you set up any position or paste a FEN and run a browser engine on it, with no account needed. Chess DNA works differently and does not have a free-position setup board: it imports the games you actually played on Chess.com or Lichess, analyses every move with Stockfish 17 at depth 18, and reports the recurring patterns in how you misjudge positions. Use the analysis board for a single position, and game-history analysis to find which kinds of positions you get wrong repeatedly.
Play through the game yourself first and mark the moves where you felt uncertain, then turn the engine on and compare. Look only at the two or three biggest evaluation drops rather than every small inaccuracy, and for each one write the human reason the move failed: a missed threat, a wrong pawn break, an unsafe king. A drop you cannot explain in words has not been learned yet. Reviewing three moves properly beats skimming forty engine suggestions you will forget by the next game.
Usually because it sees a concrete line you did not. Engines search millions of positions, so in any position with tactics their number reflects a forcing sequence rather than a structural judgement, and there the engine is simply right. In quiet positions the disagreement is more interesting: the engine may value a temporary activity no human can maintain, while your structural verdict about a weak pawn or a bad bishop is a better guide to the next twenty moves. Check the engine's main line before you overwrite your own judgement.
