Deblunder study notebook
How to turn your own chess games into puzzles
The best puzzles for you are the positions where you actually went wrong. Pick the position before the mistake, hide the engine answer, solve it cold, then review it again days later. You can do this by hand or let a tool do the selection and scheduling.
Why your own positions train better than random puzzles
Puzzle collections train pattern recognition across thousands of positions chosen by someone else. Your own mistakes show which patterns you personally miss, in the openings and structures you actually play. Solving them again closes the gap between knowing a tactic and seeing it in your games.
There is a trade-off: your games give fewer positions than a puzzle database, and many of your mistakes are not clean tactics. That is fine. A position where you moved too fast or missed a defensive resource is still worth practicing, as long as the question is clear.
Step 1: choose the right position
Use the position immediately before your mistake, with you to move. That is the decision you need to repeat. Avoid positions where the engine prefers a long, quiet line with a tiny advantage: they make poor puzzles because there is no clear right answer.
Good candidates have one move or idea that clearly changes the result: a capture that loses a defender, a check you ignored, a recapture you could not make. Keep the time control in your notes. A bullet blunder and a rapid blunder may need different training.
Step 2: hide the answer and write the question
A puzzle only trains you if you cannot see the solution. Save the position without the engine line, then write a one line question: 'Black to move. What does 24...Bxc7 allow?' or 'White to move. Find the move that keeps the e-file closed.'
On Lichess you can save positions in a study chapter with the analysis hidden until you check it. On Chess.com, Game Review includes options to retry mistakes in the reviewed game. A simple spreadsheet with the FEN and your question also works.
Step 3: solve, explain, then review on a schedule
Solve the position without moving pieces first. Say your move and the opponent's best reply. Only then reveal the engine line and compare. If you were wrong, write the relation you missed in one sentence.
Come back to the same position after a few days, then after a longer gap. Spaced review is what turns a one time correction into recognition. Positions you solve easily can wait longer; the ones you miss again come back sooner.
Automating the whole loop
Doing this by hand takes discipline: selecting positions, hiding answers and scheduling reviews. Deblunder does it from your completed Chess.com, Lichess or PGN games. It selects decisive positions, groups them by recurring cause and trains them in four steps: Solve, Understand, Compare and Recall.
Only your first valid attempt is scored, engine answers stay hidden while you solve, and Recall hides them again in later reviews. Practice attempts do not change accuracy or review dates. The first diagnosis and training are free without a card.
Frequently asked questions
Does Lichess make puzzles from my games?
Lichess's puzzle database is built from real games played on the site, but it is a shared database, not a personal collection of your mistakes. To practice your own positions on Lichess, use 'Learn from your mistakes' in the analysis board or save positions in a study.
How many positions should I train per day?
A small, consistent number works better than a long session you cannot keep. Three to ten positions per day, with older ones coming back for review, is a sustainable routine for most players.
Should I include positions from blitz and bullet?
Yes, but keep them separate from rapid and classical. Fast games add time pressure mistakes, and mixing them can hide the cause that matters in your main time control.