Chess notation cheat-sheet (for new readers)
Pieces: K king · Q queen · R rook · B bishop · N knight · (a pawn move has no letter)
Actions: x takes · + check · # mate · O-O short castles · O-O-O long castles · e8=Q pawn promotes to queen
Annotations: ! good move · !! brilliant · ? mistake · ?? blunder
Pick a theme and start solving
Most people don't need "more chess puzzles." They need the right ones. Below are the four families that cover roughly 90% of the spots where a club player either wins the game or hands it back. Click into the one that matches what hurt you in your last loss.
- Tactics: forks, pins, skewers, deflections, discovered attacks, and the rest. Start here if you blunder pieces or miss your opponent's threats. Branch into forks, pins, skewers, discovered attacks, and sacrifice sets.
- Checkmate patterns: named mating nets that show up over and over once you know the shape. The pattern pages, like back-rank mate, smothered mate, Anastasia's mate, and the depth ladders from mate in 1 through mate in 5.
- Endgames: technique under reduced material. Drill the king-and-pawn race in pawn endgames, the cutting and shouldering work in rook endgames, plus knight, bishop, and queen sets.
- Opening puzzles: tactical shots that come straight out of the opening you actually play. Sicilian, Italian, Ruy Lopez, French, Caro-Kann, and more, sorted by opening from the openings puzzles hub.
Sort by difficulty or by phase
If you'd rather pick by how hard the line is than by motif, use the difficulty buckets. Easy puzzles are sub-1400 positions, mostly one or two moves deep, ideal for warm-ups and for rebuilding pattern recognition after a layoff. Hard puzzles sit at 1900+ and demand real calculation, candidate-move discipline, and willingness to look at quiet moves, not only checks and captures.
You can also drill by phase. Middlegame puzzles are about converting a position you've already navigated out of the opening: king-side attack timing, queenside breaks, and what to do with an isolated queen pawn. Endgame puzzles are about technique once the queens come off. Opening puzzles are about punishing the typical mistake your opponent's repertoire allows.
How to actually improve with chess puzzles
I've watched a lot of students grind 50 puzzles a day for months and stay the same rating. The grind isn't the problem. The lack of a verdict after each puzzle is. Here's how I run a session that moves the needle, written out so you can copy it.
- Pick a theme that matches a recent loss. If you hung a piece to a knight fork last night, go to forks. Don't open mixed-bag puzzles when you know what bit you.
- Set the rating band 100 to 150 points above your puzzle rating, not 400 above. The trainer's ±100 slider keeps you in the zone where you'll fail roughly one in three. That's the learning zone.
- Before you move, name the tactic out loud. "This is a pin on the d-file, the rook can't move." If you can't name it, you didn't see it. Take a hint.
- After a fail, you get a "Not the move." verdict and the board holds while a short countdown loads the next puzzle. Click the countdown to pause, sit with the position for a moment, then click Next. Do not redo the same puzzle ten times. Pattern recognition comes from breadth, not repetition.
- End the session with two mate-pattern puzzles from a named set like Anastasia's mate or smothered mate. Mating-net shapes stick best right before you stop.
Twenty minutes done this way beats ninety minutes of unsorted blitzing through random puzzles. I know because I tried both, badly, for years.
What the trainer actually does (and what it doesn't)
I get the same email a lot: "does this thing track my weaknesses, give me reports, schedule reviews?" No. The trainer is deliberately small. Here's the honest feature list.
- A rating-band slider with a ±100 window, so you pick your zone and stay there.
- Three free hints per puzzle: the first names the motif, the second points to the piece that moves, the third gives the move.
- Next and Prev buttons. You can rewind the last puzzle if the auto-advance pulled you off too fast.
- A clear verdict after every attempt: "Solved!" when you find the move, "Not the move." when you miss.
- Auto-advance a few seconds after you solve or fail, with a click-to-pause control on the countdown.
- A board styled in the chessdoctrine palette that scales clean on phones.
What it does not do: track you across sessions, send weakness reports, run spaced-repetition, or claim that every position has been hand-graded by an engine at depth 40. The puzzle pool comes from real games, sliced by theme and rating, and presented in the format I'd want as a player. If you want more than that, my 88 YouTube lessons are the place to go deeper.