Most players do one of two things with a finished game. They close the tab, or they open an engine, scroll to the red moves, nod, and close the tab. Neither changes how they play next week. Analysis works when it compares what you thought during the game with what was actually true, and then writes the difference down. This article is a method for doing exactly that in twenty to thirty minutes per game.
Why analyze at all
Your own games are the only source of feedback that is about your chess rather than chess in general. A tactics trainer tells you that knight forks exist. Your games tell you that you miss knight forks when you are a pawn up and relaxing. Puzzles and opening videos are useful, but they treat every student the same. The errors in your games repeat, and they repeat in patterns you cannot see from the inside. Analysis is how you see them.
Step 1: Annotate before you switch the engine on
This is the step almost everyone skips, and it is the one that matters most. Before any engine runs, replay the game yourself and write short notes at three kinds of moments:
- Where you were unsure. You hesitated between two moves, or you were not sure what the plan was.
- Where the character of the game changed. A pawn structure locked, a piece was traded, a king walked.
- Where you calculated. Write down the line you saw and what you expected the opponent to do.
Keep the notes blunt: "thought ...Nd4 was strong, didn't see Bxf7+". You are not writing for anyone else. The purpose is to have a record of your thinking that the engine can then correct. If you look at the engine first, it replaces your thinking instead of testing it, and you will remember the engine's conclusions for about a day.
Step 2: Import the game
Chessblood's analysis board accepts a pasted PGN, an uploaded file, or a direct import of your recent games from Chess.com or Lichess by username. The engine runs inside your browser, so nothing is uploaded and there is no queue. Choose a depth and let the game review run through every move. Depth 18 to 20 is enough for a reliable picture of a club game; the engine can always think longer on individual positions later.
Step 3: Read the graph for turning points, not for red
The evaluation graph is the map of the game. Resist the urge to jump straight to the blunders. Instead, look for the places where the curve changes direction or jumps, and count them. Most games have three to five moments that decided the result, and that is where your time should go.
Chessblood's graph and move grades are built on win probability rather than raw centipawns, so a jump from +0.3 to -0.7 shows up as a real swing while a wobble from +8 to +7 barely registers. The articles on move classifications and accuracy explain the numbers in detail. For the workflow, the point is simple: a move only matters if it changed the expected result.
Step 4: Ask three questions at every turning point
At each of those three to five moments, put your own notes next to the engine's line and answer three questions in writing.
What did I think? Copy your annotation from step 1. If you had no note there, that is itself an answer: the critical moment passed without you noticing it was critical.
What was actually best, and why? Do not just read the engine's move. Play its line forward on the board until you can say in one sentence what the idea is: "the knight comes to d5 and the e7 bishop can never move". If you cannot explain the move in words, you have not learned it, and you will not find it next time.
What pattern did I miss? Try to name it. A tactical motif (pin, overloaded defender, back rank), a positional idea (the wrong piece trade, a pawn break you could not allow), or a process failure (moved fast in a position that needed time, stopped calculating one move too early). Process failures are the most common and the most fixable.
Step 5: Check the opening against theory
Find the first move of the game that was not a book move. The review marks these for you. Then ask whether your deviation was bad or merely unfamiliar. A move that scores +0.1 for the opponent is fine; you do not need to memorize a refutation of it. A move that hands over half a pawn is worth a note.
FEN
r1bqkb1r/ppp2ppp/2n5/3np1N1/2B5/8/PPPP1PPP/RNBQK2R w KQkq - 0 6The position above is a good example. Black's 5...Nxd5 looks natural and is played constantly at club level. The engine rates it at roughly +0.9 for White, and 6.Nxf7 (the Fried Liver) drags the black king into the center. If this were your game, the lesson is not "memorize twenty moves of the Fried Liver". It is "after 4.Ng5 d5 5.exd5, play 5...Na5", one line in your notes. Chessblood's opening guides give the main lines and the ideas behind them if you want to go further.
Step 6: Treat the endgame as its own game
Endgame mistakes hide well because the evaluation often stays at the same number for twenty moves and then collapses in one. Scroll through the endgame slowly and watch for the moment a winning evaluation turns into a draw, or a draw into a loss. Then identify the endgame type: rook and pawn, opposite-colored bishops, king and pawn. Those names are searchable. Knowing that you failed to hold a Philidor position is a lesson you can act on; "lost the endgame" is not. The article on endgame fundamentals covers the handful of positions that come up most.
Step 7: Keep a lessons file
Open a plain text file and add one or two lines after every game. Make them specific and about you:
- "Played ...h6 against Bg5 reflexively. Cost a tempo in two games this month."
- "Stopped calculating after the first check. Twice."
- "When up a pawn I trade queens automatically, even when my pieces are more active."
Read the file once a month. Themes will appear within ten games, and they will not be the themes you expected. That file is your training plan, and it is a better one than any general-purpose course because it is built from your actual mistakes.
How often to analyze
Every rapid and classical game deserves the full method. Blitz games are worth a quick pass in batches: run the review, look at the opening deviations and the single biggest swing, and move on. Bullet is not worth analyzing for chess content; if you want to review it at all, look only at the openings.
Common mistakes
Looking only at the blunders. The blunder is usually the last link in a chain. Two inaccuracies earlier put you in a position where the only good moves were hard to find. Fix the inaccuracies and the blunder never arrives.
Trusting shallow evaluations. At depth 10 the engine changes its mind constantly. For any position you are about to draw a conclusion from, let the engine reach depth 20 or more and see whether the assessment holds.
Analyzing every move. The eighteenth move of a known opening and the recapture on move 31 are not worth your attention. Spend the time at the three to five moments that mattered.
Accepting lines without understanding them. An engine line you cannot explain is trivia. Play it out until you can describe the idea.
Only analyzing losses. Your wins contain mistakes your opponent did not punish. They are the cheapest lessons you will ever get, because you did not have to lose to learn them.
If you want to start right now, open the analysis board, import your last serious game, and write your annotations before you press analyze. The blunder checker gives a faster, narrower pass when you only have five minutes.