Deblunder study notebook
How to analyze a chess game and turn mistakes into training
Good chess game analysis ends with a decision about what to practice. Start with your own explanation, use an engine to test it, and turn one recurring mistake into a position you can solve again.
1. Start with a completed game and your own notes
Choose a game you have finished, preferably one where you remember a difficult decision. Export its PGN or open its analysis board. Keep the player's side and time control clear. Comparing a rushed bullet loss with a thoughtful rapid game can hide the actual problem.
Before revealing engine moves, mark two or three moments where you were unsure. Write what you expected your move to achieve and the reply you considered. This preserves evidence about your thinking that an evaluation graph cannot reconstruct. Never use an analysis tool during a game in progress.
2. Use a blunder checker to verify consequences
An engine can show where a move sharply worsens a position. Treat that label as a starting point: play the opponent's strongest reply and explain what changes on the board. Does a defender disappear? Does a file open? Does a forcing check prevent your intended recapture?
A centipawn change is not an explanation, and different engine settings or platforms can classify the same move differently. Examine the continuation and the practical decision. A winning position can become harder to convert without becoming lost.
A concrete example: the bishop was also a blocker
In the real-game example on the Deblunder homepage, Black has a bishop on e5 between White's rook on e1 and Black's rook on e8. After 24.Nc7, Black plays 24...Bxc7. The capture wins a knight, but moving the bishop clears the e-file and permits 25.Rxe8+.
The useful lesson is not simply 'avoid Bxc7'. Before moving a piece, ask which line it blocks and what becomes exposed after it leaves. The homepage lets you inspect the position and alternative continuations. Its engine analysis supports this tactical consequence; it does not establish a universal rule that captures are bad.
3. Compare causes across games, then choose one
Keep a short record: position, intended idea, missed reply, and an actionable cause. 'Lost a rook' records the outcome. 'Moved the only blocker between two rooks' describes a relationship you can check next time. If the cause cannot be established from the position or your notes, leave it unknown.
Look for the same relationship in other games before calling it a recurring weakness. A single example suggests a practice target; it does not prove a stable habit. Keep rapid, blitz and bullet comparisons separate and record the sample you used.
4. Solve again before looking at the answer
Return to the position before the mistake. Choose a move and explain your opponent's strongest reply before revealing the engine line. Compare your reasoning with the verified consequence, then revisit the position later with the answer hidden. This tests recall rather than recognition of a move you just saw.
Deblunder imports completed Chess.com and Lichess games by public username, or PGN files, then connects diagnosis to guided practice from your positions. The sequence is Solve, Understand, Compare, Recall. Sessions contain 3, 5 or 10 positions. Your first valid attempt is scored; later practice does not rewrite that result.
What to measure in your next games
Choose a process you can observe: did you check what your moving piece was defending or blocking? After another comparable sample, review whether that same error recurs. Rating and puzzle scores can fluctuate, so neither alone proves that a specific skill improved.
Deblunder's initial diagnosis and first training are free without a card. Continued full access has paid conditions shown in the app, including a seven-day trial option. Game availability and analysis depth limit what can be concluded. No tool can guarantee a rating gain.
Frequently asked questions
Can I analyze games from Chess.com and Lichess together?
Deblunder supports both sources and PGN. Keep player perspective, duplicates and time controls clear when interpreting a combined sample. You do not provide either platform's password.
Should I analyze every move?
For a practical review, start with a few critical decisions and their consequences. A complete engine scan can locate candidates, but explaining and revisiting one useful position is a more manageable next action than memorizing every engine line.