Practice

Solve the position you actually blundered — not a stranger's puzzle.

Every puzzle is generated from a real mistake in one of your own games, scheduled with spaced repetition so the ones you keep missing come back sooner. Plus single-attempt exams and short survival drills for when you want a real verdict, not just practice.

Puzzles built from your own blunders

Not a generic tactics database — the exact position you misplayed.

Every analyzed game that contains a blunder above your own threshold turns those moments into puzzles automatically. You get the position right before the mistake, replayed against you with a live evaluation bar, so the lesson sticks to something you actually did rather than a pattern lifted from a stranger's game. Misses are scheduled with a Leitner-box spaced-repetition system: a puzzle you get wrong comes back sooner, one you keep getting right drifts further out, and the newest, most-missed puzzles are reviewed daily.

Exams & Survival Challenges

A different kind of test: single-attempt exams that grade you against theory, and short survival drills that grade you against the engine from an unscripted position.

Opening Coach (see the Openings page) is for learning a repertoire — mistakes get taken back with feedback, and you can retry. The Exams below are graded differently: each is a one-shot test with a clear pass/fail verdict, for when you want to know where you actually stand rather than drill toward a line you already half-know.

Exam 4B / 4W / 5B / 5W — a single-attempt test of your own opening theory. Each of several tasks walks through a randomly chosen named opening's own moves for your color, typed exactly, with the computer playing the other side's book moves in between. You need a minimum number of correct moves to pass a task — more for a long-book opening that hands the other side several moves of its own — and once the book runs out, any further moves needed are graded against the engine's own best move instead. There are no retries: the first move you play is the one that's graded, and one wrong move fails that task and moves straight on to the next.
Hold The Advantage — starts from a position where you already have a real edge, pulled from an actual game, and tests whether you can keep it. Play accurately enough to stay above a pass threshold and you hold the advantage; slip too far and you don't. It's the mirror image of the puzzle set above: instead of fixing a mistake you made, it checks whether you can convert a good position into a full point.

Five Moves No Mistakes and Six Moves No Mistakes are shorter, repeatable survival drills rather than a one-shot exam. The opponent opens with two fully random legal moves — genuinely anything technically possible, not just sensible opening theory, so the position you're handed is different every time — then switches to the engine's own best move for every move after that. Your job is to find your own best reply, five or six times in a row, while staying under a single running budget for how much you're allowed to lose along the way.

That budget isn't a flat number of pawns. It's measured in win probability rather than raw centipawns, so the same size slip counts for less once the position is already decisive — say, the opponent's random opening already blundered a pawn or more — and counts for more in a close position, where every centipawn is doing real work. It's the same win-probability curve behind ChessMend's own move-accuracy grading elsewhere in the app, not a separate invented scale.

RANDOM OPENING

A genuinely fresh start each time

The opponent's first two moves are picked uniformly at random from every legal move — not a weighted "reasonable" opening — so no two attempts start alike.

RELATIVE BUDGET

Graded on win probability, not raw pawns

A single running budget across all your moves, scaled to how much is actually at stake in the position — not a flat cutoff regardless of context.

KEEP EXPLORING

Play on after the verdict

Passing or failing doesn't end the game — the board stays live afterward as free play against the engine, no longer graded.

HONEST HELP

Show Best Move & Undo Last Move

Both are available mid-drill for when you want to explore rather than test yourself — using either during a graded move fails that attempt, on the spot.

TWO WAYS TO REPLAY

Try Again vs. Next Challenge

Try Again replays the exact same random opening so you can attempt it again with a different plan; Next Challenge rolls a completely fresh one.

PLAY THE OTHER SIDE

Automatic Opponent Move, off

Also available in Opening Coach: turn this off to play a move for the opponent yourself — say, an unexpected capture — instead of the coach always finding its own best reply.

Daily Puzzle Challenge

A fresh themed set every day, with its own streak history.

Alongside open-ended practice, a Daily Puzzle Challenge assembles a themed set each calendar day — separate from your main blunder backlog — with its own history view so you can track how many days in a row you've kept it up, not just a vague feeling that you've been "pretty consistent lately."

SPACED REPETITION

A Leitner box, not a fixed queue

Misses come back sooner; mastered puzzles drift further out automatically.

DAILY

Daily Puzzle Challenge

A fresh themed set every day, with its own streak history separate from your main backlog.

THEMED

Practice by tactical theme

Filter down to one theme — forks, pins, back-rank — and grind it until the pattern is automatic.

MULTI-PLAYER

Puzzles by player

Track more than one person's games — yourself, a student, a training partner — with puzzles grouped per player.

EXPORT

Blunder screenshots & video

Export a blunder as a screenshot or short video clip — handy for sharing a lesson or a "how did I miss this" moment.

PROGRESS

Streaks & a progress timeline

A chronological view of auto-detected milestones — first analyzed game, puzzle-count round numbers, biggest rating jump.

Explore the rest of the app

← Back to the app

Drill the mistake you actually made.

Import a few games and your first puzzle set is ready in minutes.