Choose a position immediately before a useful mistake, hide the answer, and try it again later in saved-position practice. When you solve it, explain the opponent's threat and why your continuation works. That turns a past game into a specific exercise rather than another review you watched and forgot.
Your own games supply context: the move you wanted to play, the reply you missed, and the reason the position matters to you. They do not automatically supply a well-designed puzzle. Some positions are ambiguous, some depend on long analysis, and some have several good answers.
The practical task is to curate a small notebook that you can understand and revisit.
Key takeaways
- Save a position because it teaches a clear idea, not only because its evaluation dropped.
- Test your explanation as well as your selected move.
- Use unfamiliar positions to check whether the idea transfers beyond memory.
Select positions with a question you can answer
A good early exercise asks something concrete: can you win the loose piece, stop the mating threat, or find the forcing check? The answer should become visible through a line you can follow.
A position requiring a subtle long-term sacrifice can be valuable, but it may be a poor first item in a short daily notebook. If you cannot explain the recommendation after reviewing it, record it as an open question rather than grading yourself against a mysterious move.
Start with a few positions. Adding every engine warning creates a collection faster than it creates understanding. You can always return to the PGN for more material later.
Preserve the position before the decision
Store the board before your mistake, with the correct side to move. Saving the position after you lost the piece changes the task into damage control.
A FEN can record the setup, but it does not preserve the full game history. The PGN specification's FEN section describes position fields. Keep the original game as well when repetition history or the preceding plan matters.
Outside the tool, a compact note can hold the information the board cannot:
| Field | Example |
|---|---|
| Question | What forcing move attacks both king and rook? |
| Original thought | I looked for a direct rook capture. |
| Missed idea | A knight check wins a tempo. |
| Transfer cue | Inspect knight checks when a king and rook share vulnerable squares. |
This note is a suggested manual worksheet. ChessStep does not currently provide all these annotation fields in its notebook interface.
Rehearse the reason before the move
For the illustrative position below, White has a knight on e5 and king on g1; Black has a rook on d8 and king on h8:
3r3k/8/8/4N3/8/8/8/6K1 w - - 0 1
The continuation Nf7+ checks the king and attacks the rook. After ...Kg8, Nxd8 captures the rook. With only the remaining kings and White's knight, this simplified example is then drawn; winning the rook does not mean winning the game. On a repeat attempt, saying “I remember f7” is less informative than noticing both attacks and the fact that Black must respond to check.
Try reversing the task in your explanation. Why can Black not simply move the rook away? Because that would leave the king in check in this position. This shows you understand the forcing element rather than only the destination square.
You can also ask what would change if the king or rook occupied a different square. Do not assume the same move remains correct; set up the new position and verify it separately.
Schedule another attempt without treating the schedule as magic
Return after a gap long enough that you have to think. A next-day attempt is a practical place to start. If you miss the idea, review it and keep it available; if you solve it confidently, leave a longer gap.
This is a suggested practice arrangement, not a proven optimal interval for every player. Fatigue, familiarity, and the difficulty of the position affect the result. You do not need a complicated scheduling theory to begin keeping a manageable set of exercises.
ChessStep records the first attempt each time you open or restart an exercise and treats revealing the answer as needing another look. Correct saved-position attempts receive a later review date; missed ones remain due. The current scheduler is simple and is not a validated measure of chess mastery.
Add a transfer check
A repeated position asks whether you can reconstruct a familiar idea. A new position asks whether you can recognize when that idea applies. Both are useful, but they should not share one undifferentiated score in your own notes.
After revisiting a fork, inspect a different game for a knight check against two targets. After a loose-piece exercise, examine a fresh position and name the undefended pieces without knowing whether a tactic exists.
Sometimes the correct answer is “there is no safe capture.” That matters because real games do not label every position as a puzzle with a winning move.
The progress guide offers a manual way to distinguish repeated-position recall from unfamiliar decisions.
Know exactly what the tool saves
In ChessStep review, you can save a highlighted learning moment and revisit it in practice. The current product saves the selected position and continuation. It does not automatically generate a new family of related puzzles.
The practice checker compares your move with that stored answer. A different move can be legal and still be rejected; some alternatives may also be good. If you suspect that, analyze the alternative instead of concluding that the interface has proved it wrong.
Guest positions and practice attempts stay in this browser. When you sign in, those guest learning records merge into the first account you use and sync to that account. On another device, sign in and open My notebook to load them. Pending changes remain on the original device until synchronization succeeds; use Retry sync if an error appears.
The JSON export is an archive for keeping a separate copy. This version does not have a matching JSON restore interface. Full bot games follow a different rule: sign in to save them; an unsaved guest bot game is not part of the notebook merge.
Keep the collection small enough that each item has a purpose. The reward for reviewing a mistake is a question you can ask yourself at the board, not merely a larger archive.