How to Analyze Your Chess Games: A Step-by-Step Guide

To analyze a chess game, replay it without an engine first and mark the two or three moments where you were unsure or your position turned. At each one, write down your opponent's threat and your candidate moves. Then let Stockfish check only those moments, name the cause of each mistake in plain words, and save one position to practise. That takes about fifteen minutes.

This guide is written for players who know the rules and play online, roughly 600 to 1600 in Rapid. You need no opening theory and no paid account. Every board below is checked: each position is legal, every line in it was played out move by move with a chess library, and the engine and endgame verdicts were checked with Stockfish 17.1 and an endgame tablebase.

Why analyzing your own games works

A puzzle tells you that something is there to find. A real game does not. Many games at this level turn on a decision you made without noticing it was a decision: a developing move played while a threat was on the board, a "free" pawn taken without checking the reply, a trade that looked equal.

Reviewing your own games is how you find out which of those decisions keeps going wrong for you. Your games are the one source of positions that already fits your openings, your time habits and your blind spots. The goal of a review is not to grade the game. It is to turn "I lost again" into one sentence you can act on in the next game.

The 15-minute game review at a glance

Minutes	Step	You finish with
0–4	Replay the game with the engine off	2–3 marked critical moments
4–8	At each moment: threats, then candidate moves	What you saw, and what you missed
8–12	Engine check of the marked moments only	The better move and the reason it is better
12–14	Classify each mistake	"Missed threat", "bad trade", ...
14–15	Journal entry and one saved position	Tomorrow's practice

Fifteen minutes is a habit-sized target, not a limit. A long, complicated game can take longer; a quick loss to a one-move blunder can take five.

Step 1: Replay the game before you turn on the engine

Open the finished game, then hide the evaluation bar and the move classifications. On Chess.com and Lichess both can be switched off in the analysis settings; if you cannot find the switch, cover that part of the screen. If you see the engine's verdict first, you stop remembering what you were thinking and start agreeing with the arrow.

Play through the moves at a normal pace and stop wherever one of these is true:

• you spent much longer than usual on a move;
• you were choosing between two moves and were not sure;
• your position suddenly felt uncomfortable, or material changed hands;
• you decided to trade pieces, or went into an ending;
• the clock forced you to move without thinking.

These stops are your critical moments: the positions where the game was really decided or could have been. Usually there are two or three. Mark them and move on; you will come back to each one.

Step 2: At each critical moment, check your opponent's threat first

Before looking for your own best move, answer one question: what does my opponent's last move threaten? Many games at this level are lost by a move that was fine in itself but ignored the threat on the board.

Black to move

87654321

♜︎♝︎♛︎♚︎♝︎♞︎♜︎♟︎♟︎♟︎♟︎♟︎♟︎♟︎♞︎♟︎♕︎♗︎♙︎♙︎♙︎♙︎♙︎♙︎♙︎♙︎♖︎♘︎♗︎♔︎♘︎♖︎

abcdefgh

In this position White's queen on h5 and bishop on c4 both aim at f7, a square only Black's king defends. White is threatening 4.Qxf7 checkmate. The natural developing move 3...Nf6?? walks straight into it: the knight attacks the queen, but 4.Qxf7 is mate. In fact 20 of Black's 28 legal moves allow 4.Qxf7#.

A player who reviews this game and writes "lost to an early queen attack" has learned little. The useful note is: "I developed without asking what the queen and bishop were aiming at." That habit costs games long after the opening, too.

Step 3: List your candidate moves

Candidate moves are the two to four moves worth comparing in a position. Listing them stops you from playing the first move you see. In the diagram above, Black has four sensible candidates, and each meets the threat in a different way:

• 3...g6 blocks the queen's line to f7 and attacks the queen.
• 3...Qe7 or 3...Qf6 adds a second defender of f7.
• 3...Nh6 defends f7 with the knight.

Stockfish rates all four as roughly level. The engine's favourite is less important than the method: threat first, then candidates, then a quick check of your opponent's best reply to each.

When you search for candidates, use the checks, captures and threats scan, for both sides:

1. Checks: which checks are available, to either king?
2. Captures: which pieces can be taken, and which are defended only once, or not at all?
3. Threats: what would each side like to play next?

It is a search method, not a rule that forcing moves are best. If the forcing patterns themselves are unfamiliar, the beginner's guide to chess tactics covers forks, pins and skewers with examples.

Step 4: Use the engine as a referee, not a teacher

Only now open the engine, and look at the moments you marked, not every move. For each one, compare your move with the engine's choice and ask why yours is worse. Play out the engine's line a few moves until you can say the reason in one sentence, without quoting a number.

White to move

87654321

♜︎♝︎♛︎♚︎♝︎♞︎♜︎♟︎♟︎♟︎♟︎♟︎♟︎♟︎♟︎♗︎♞︎♙︎♘︎♙︎♙︎♙︎♙︎♙︎♙︎♙︎♖︎♘︎♗︎♕︎♔︎♖︎

abcdefgh

Here Black's knight has left c6, so the pawn on e5 looks free. 4.Nxe5? takes it, but Black has one strong reply, 4...Qg5!, attacking the knight on e5 and the pawn on g2 at once. Neither is defended. If White now plays the tempting 5.Nxf7, forking the queen and the rook on h8, Black does not save the rook. The queen takes on g2 with a new attack on the rook on h1: 5...Qxg2 6.Rf1 Qxe4+, and 7.Be2 is met by 7...Nf3 checkmate. White's only other way out of the check, 7.Qe2, gives up the queen.

What do engine evaluations actually mean?

An evaluation is the engine's estimate of who is better, shown in pawn units from White's side: +1.3 means White is better by roughly a pawn and a bit, −0.5 means Black is slightly better, and a score like M3 or #3 means a forced mate in three. These are the numbers Stockfish 17.1 gave for the example above:

Position	Engine evaluation	What happened
After 3...Nd4, with a calm move such as 4.Nxd4	about +1.3	White is better
After 4.Nxe5? Qg5	about −0.5	the "free" pawn turned White's edge into a worse position
After 5.Nxf7? Qxg2	about −5	a lost position

Read the swings, not the decimals. As a rough guide, a drop of about a pawn marks a mistake and a drop of several pawns marks a blunder. The exact figures change with search depth and engine version, so do not argue over 0.2.

Three things to keep in mind:

• Do not copy the engine's move. Find the reason. Here it is: before grabbing material, check your opponent's forcing replies (a queen move that attacks two undefended targets is a forcing reply).
• Look at the opponent's best answer to your move, not only at the engine's replacement for it. 4...Qg5 is the move that explains the mistake.
• Notice "only moves". Stockfish found that 4...Qg5 is Black's only reply that keeps an advantage; every other reply leaves White better. If your opponent would have needed a hard-to-see only move, the mistake was real but may matter less than one any opponent punishes.

How to analyze games on Chess.com and Lichess

Chess.com: open the finished game from your game archive and choose Analysis (or Game Review, which adds a summary and labels such as "mistake" and "blunder"). How many full Game Reviews you get depends on your membership, but the analysis board itself is enough for this method. Turn the evaluation bar and suggestions off for Step 1.

Lichess: open the game and press Request a computer analysis; it is free. You get an evaluation graph, the engine's lines and a Learn from your mistakes mode that replays the errors as short exercises. Lichess also lets you save positions into a free study, which is a convenient place for your journal positions.

Both sites let you export the game as a PGN file, the standard text format for chess games, so you can keep an archive or analyze it elsewhere.

Menus and limits on both sites change from time to time; the method does not depend on any one button.

Tactical oversight or positional mistake?

After the engine check, name the kind of mistake. The two you will meet most often call for different training.

• A tactical oversight is a short, forcing sequence you or your opponent missed: a hanging piece, a fork, a pin, a mate. A two- or three-move calculation would have found it. Both examples above are tactical.
• A positional mistake leaves nothing hanging. The move is legal and safe for now, but the position it leads to is worse: a bad trade, a weakened king, a passive piece, a lost ending.

White to move

87654321

♜︎♚︎♟︎♔︎♙︎♖︎

abcdefgh

Material is level: a rook and a pawn each. White's rook is attacked, and 1.Rxd8? Kxd8 looks like a sensible simplification. It loses. In the pawn ending both pawns race, for example 2.h4 a4 3.h5 a3 4.h6 a2 5.h7 a1=Q 6.h8=Q+ Qxh8: Black queens first, and the new queen on a1 covers h8 along the long diagonal. Keeping the rooks with 1.Ra1, which also attacks the a-pawn, holds the draw. An endgame tablebase (a database of perfect play for positions with seven pieces or fewer) confirms both verdicts: Ra1 and five other rook moves draw, the trade and every other move lose.

Nothing was left hanging here, and no tactic was missed. The mistake was judging the position after the trade. That is a positional (evaluation) mistake, and "solve more puzzles" would not fix it; knowing a few basic pawn endings would.

A quick test when you classify:

Question	If yes, it is probably
Was a piece left undefended, or a check, capture or fork missed?	Tactical oversight
Did your opponent's last move create a threat you did not answer?	Missed threat (tactical)
Was nothing lost at once, but the resulting position clearly worse?	Positional or endgame mistake
Did you know the right idea but move in a hurry?	Time management
Did you leave known opening moves and get lost?	Opening knowledge

Look for patterns across several games

One mistake is an incident. A weakness is a mistake you keep making. Review at least ten games of the same time control before deciding what your weaknesses are, and more is better: ten games can show an obvious habit, but not a reliable ranking. Keep Rapid and Blitz separate at first; faster games add mistakes that come from the clock rather than from understanding.

For each game, record the mistakes you classified in the step above. After ten games you will usually see the same two or three labels coming back.

Choose your three weaknesses

Rank each type of mistake by three questions:

1. How often does it happen? Count games in which it appeared.
2. How much does it cost? A blunder that loses a piece costs more than a small inaccuracy.
3. Can you train it in short sessions? "Check the threat before moving" is trainable in minutes a day. "Learn a whole opening" is not a 15-minute problem.

Pick the top three, not ten. Three is enough to vary your practice and few enough to see whether anything changes. This hypothetical tally, with invented numbers for illustration, shows the idea:

Mistake type	Games it appeared in (of 12)	Typical cost	Trainable daily?
Missed opponent's threat	7	piece or more	yes
Took material without checking the reply	4	piece	yes
Traded into a lost ending	3	the game	yes
Slow opening development	2	small	partly
Time trouble	1	varies	partly

Here the three weaknesses are the first three rows. Re-check the ranking every few weeks with new games.

We are building an Android app, MindGrid Chess Trainer, that is meant to do this sorting across your games and turn the top three into short daily practice. It is in development: you can join the Chess Trainer beta waitlist to hear when testing opens. The method in this guide works without it.

Keep a simple mistake journal

A notebook or a spreadsheet is enough. Keep one row per mistake, not per game:

Game and move	What happened	Type	Lesson, then practice
Oct 3 Rapid, 3...Nf6??	Allowed Qxf7 mate	Missed threat	Ask what their last move attacks. Practise: 5 "name the threat" positions
Oct 5 Rapid, 4.Nxe5?	Grabbed a pawn, lost to ...Qg5	Took material without checking the reply	Checks, captures, threats for my opponent first. Practise: retry this position next week
Oct 6 Rapid, 1.Rxd8?	Traded into a lost pawn ending	Positional / endgame	Picture the position after the trade. Practise: pawn-race endings

The rows above reuse this guide's examples; your journal holds your own games. The lesson column matters most. Write it as an instruction to your future self, not a verdict: "check what their last move attacks", not "blundered".

Turn mistakes into training exercises

A position you got wrong in a real game is the best exercise you have, because it is exactly your kind of mistake.

• Save the position, not just the move. Store it in a Lichess study or write down its FEN (the one-line text code for a position that both sites can copy and paste). Try it again cold a week later.
• Make a threat drill. Play through a game and stop after each of your opponent's moves. Say out loud what that move attacks before you look at your reply.
• Mix in quiet positions. Practise some positions where the right answer is a calm move, not a combination. A game never tells you a tactic exists.
• Play out endings. From a saved ending, play both sides against a friend or the engine until you know why it is won, drawn or lost.
• Match puzzles to the weakness. If your journal says "missed forks", solve fork puzzles, such as those explained in our forks, pins and skewers guide, rather than a random mix.

A sample week of 15-minute training

An example for a player whose three weaknesses are missed threats, grabbing material without checking the reply, and trading into bad endings. Your games themselves are played outside these fifteen minutes.

Day	5 minutes	5 minutes	5 minutes
Mon	Threat drill on one of your games	Retry 2 saved positions	One pawn-race ending
Tue	10 short tactics puzzles on double attacks	Threat drill	Review this week's journal lessons
Wed	Replay your latest game, engine off: mark critical moments	Engine check of those moments	Journal entry
Thu	Retry 2 saved positions	Puzzles: "is the pawn really free?"	Trade-or-keep decisions in 3 positions
Fri	Threat drill	10 puzzles mixed with quiet positions	One rook-and-pawn ending
Sat	Replay your latest game, engine off	Engine check of the marked moments	Journal entry
Sun	Re-read the week's lessons	Retry the hardest saved position	Rest, or a free game

Progress shows up first in your journal, as the same label appearing less often in new games. Rating changes take many more games to show and move up and down for many reasons, so judge the habit, not the number.

Questions beginners ask about game analysis

Should I analyze every chess game?

No. Review the games where you were unsure or made a mistake you can name; a single careful review is worth more than skimming ten. Look at a win now and then too, since wins hide mistakes your opponent did not punish.

Do I need Chess.com Premium to analyze my games?

No. Everything in this guide works with free tools. Lichess computer analysis is free, and Chess.com's analysis board works without a full Game Review.

Should I memorize the engine's moves?

No. Memorize the reason: the threat you missed, the reply you did not check, the ending you misjudged. The reason transfers to new positions; a specific engine line rarely does.

What accuracy score is good?

Accuracy percentages summarize a whole game in one number and depend on how complicated the game was. Two games with the same accuracy can contain very different mistakes. Use the score as a hint of where to look, not as the goal.

Should I analyze Blitz or Rapid games?

Start with Rapid if you play it: you had time to think, so your mistakes show how you think. If you mostly play Blitz, analyze those too, but keep them in separate journal columns.

Can solving puzzles replace game analysis?

No. Puzzles train recognition once you know a tactic is there. Game analysis shows when you failed to look in the first place, and which patterns you need. Use both: let the journal decide which puzzles you solve.

After your next game

1. Replay it with the engine off and mark two or three critical moments.
2. At each one, write your opponent's threat and your candidate moves.
3. Check only those moments with the engine; find the reason, not just the move.
4. Name each mistake, add one journal row and save one position.
5. Practise that position, and others like it, for the next few days.

If you want to try tactics from real positions first, our free chess puzzles for kids are short warm-ups with explained answers, and the mate in one guide covers the checkmate patterns that decide many beginner games.