What ChessClerk does

Everything in ChessClerk and how to use it, 186 entries, in the same words as the Manual inside the app (Help > Manual).

Board · Help · Learn · Clerk · Play · Tools · Openings · Explorer · Puzzles · Drills · Library · Repertoire · Endgames · Review and tracking · Motifs · Databases · My study · Research · Chat · Lines

Board

Board only

Board only, in the row with the ⚙ and Arrange: everything but the board goes, and the board fills the screen, still square. You can still move, play a game against the computer or do a puzzle; only the boxes the game waits on show (promotion, the game's end). Esc, or the faint ✕ in the top-right corner, brings everything back. It stays on after a reload until you end it.

Moving pieces

Click a piece, then the square it goes to, or drag it there. To castle, move the king two squares, or drop it on its own rook (as in Chess960). A pawn reaching the last rank opens a box: pick queen, rook, knight or bishop. An illegal try flashes the square orange; so does a king in check.

Nav buttons

Beside the board in Original and Focus, under it in the other layouts: ⏮ start, ← back, Play (a game at once; ⏸ and ▶ during a game), 960 (a Chess960 game), → forward, ⏭ end, ↶ undo (takes the move back off the game), ↷ redo (puts it back), ⇅ flip the board, ⟳ reset to the start, ½ offer or claim a draw, ⚑ resign, Tools (the tool buttons; Nav brings these back). The arrow keys, Home and End step through the game too.

Eval line and captured pieces

Under the board: the pieces each side has taken, with the side ahead marked like "white +3", and the eval in gold, like "position: +0.40 (depth 18)". + is good for White, - for Black; M3 means mate in 3. Every eval is worked out by our own Stockfish 19 on your computer.

SAN panel

The moves of the game under the board. The move on the board is lit gold. Side-lines are in brackets, comments in italics. Click any move to see it on the board.

Arrows and colored squares

Right-click a square for a colored ring; right-drag from one square to another for an arrow. Teal, or crimson with Shift, violet with Ctrl, gold with Alt. The same mark again takes it off. On a phone or tablet, press and hold a square for a ring; hold again for the next color (teal, crimson, violet, gold, then off). Marks belong to the position they were drawn on and are saved in PGN. Tools > Clear wipes them from the position shown.

Annot

The Annot button over the chat shows or hides every arrow and colored square, and the symbols and comments in the move list. It remembers your choice.

Symbols, comments and side-lines

Right-click a move in the SAN panel: give it ! ? !! ?? !? or ?! (again to take it off), write a comment, make a side-line the main line, or delete the move and everything after it. Playing a different move from an earlier position makes a side-line.

Double-click the eval

Double-click the eval under the board to watch it solved. A mate plays out, the defender's replies too, about three moves a second, then the board snaps back; do it a few times and play it from memory. Any other eval plays its line only when the last move swung it by 3 pawns or more and the best move is far ahead of the rest; otherwise it says why there's nothing to watch.

Move notes

Notes, at the top of Current game: pick White, Black or Both, and every move of that side gets a short note under the moves as it is played or stepped to, in any line or game on the board: its comment, the opening book's idea, the motifs it brings and what our engine thinks of it. Off turns them off. Nothing is made up: a move with nothing true to say gets no note. During a game you are playing, only the comment and the book show until the game is over. It remembers your choice.

Record

Record, in the corner row with the ⚙ (on a phone, at the end of the Board · Panel · AI row): press it and ChessClerk records what happens on the board for up to a minute: the moves, stepping through a game, the arrows and colored squares, a motif's red box, and what the clerk says, as captions. Press it again to stop. Then cut the clip to the part you want, watch it, and make the video, Phone (tall) or Wide, drawn by ChessClerk itself. On a phone, Share sends it with the phone's own share menu; on a computer, Download saves it. Nothing is uploaded anywhere.

Screenshot

Screenshot, beside Record (on a phone, the small camera): one press makes a clean picture of the board, the moves around it and the clerk's last post, with chessclerk.com small in the corner. On a phone, Share sends it with the phone's own share menu; Download saves it. Nothing is uploaded.

Help

Font, flash color and panel widths

Settings (⚙) > The screen: Font picks the screen's words' font (Georgia, Palatino, Times, Sans, Verdana; Georgia unless you change it). Flash color picks the color an illegal try and a king in check flash on the board (orange, red, yellow, magenta, green; orange unless you change it). The text panel and the chat start the same width: drag the thin divider between them (or Tab to it and use the arrow keys) to give one more room; double-click it, or press Even under Panel widths, to make them the same again. Each choice is kept on this computer.

Memory

Help > Memory: what the clerk keeps about you, one line each, from what you told it or corrected it on. It reads them on every turn. ✕ takes a line out; Forget everything empties it. Under them, the topics you work on with where each stands, and how you work and what you miss: what the clerk writes for itself as you go, used quietly and never shared; ✕ takes any out. Kept on this computer only.

Old sessions

Help > Old sessions: every session of the chat (+ New Session starts one), with when it was, how many posts and its first words. Read shows its posts, read-only. Ask the clerk about any post by its number, like #92.

Introduction and Feedback

The first row of the rail. Tell us what's wrong, what's good and what you want: write it, press Send. Report a bug keeps your words with a snapshot of the board and what you last pressed. Until the server is connected, everything waits on this computer; Copy all puts it on your clipboard. Your data lists what ChessClerk keeps and has Delete my data.

Today

Today, at the top of the rail: things to pick from, never a schedule. Your plan for today (the steps of the weekly plan you make on My study > Tracking that you put on today and haven't ticked; shown only when there are some), your own mistakes as puzzles (Look for new ones checks your newest; missed plunders are marked), a puzzle to stretch you (a little above your puzzle rating), your next King's ladder rung, and your last game with its review.

Ratings

Help > Ratings: ChessClerk shows ratings in USCF only. Your rating is worked out from your current play here (the median cp gap of your last 200 moves that count), never typed in. Wherever ChessClerk gives a rating from your play (the free review, Tracking, your goal), it is the USCF one. Ask the clerk what it is on Chess.com, Lichess, FIDE or Glicko, or use the converter there, which turns a rating from any system into all the others, from the ChessGoals survey.

Easy for everyone

Help > Easy for everyone: make the words bigger (the board keeps its size), switch to color-blind safe colors for the highlights, arrows and rings, and play from the keyboard: Tab to the board or press B, move the gold square with the arrow keys, Enter or Space to pick up and put down, Escape to leave the board.

Computers and phones

Desktop recommended but not required; phones are still adequate. Every screen works on a phone, with typing or the 🎤 (your choice); the full set of panels side by side needs a bigger screen.

Arrange

Arrange, at the end of the row with Board only and the ⚙: every box gets a gold bar to drag it anywhere and a corner to stretch or shorten it. The board's corner changes the size of its squares, so it always stays square. Done ends it; Put back gives the layout its own arrangement. Each device and layout keeps its own.

Ask the clerk

Help > Ask the clerk lists what only the clerk does, with no tab of its own: teach a whole collection of motifs, serious business (a great game, annotated), remember something about you, and keep what you think of ChessClerk. Each button brings up the chat with the message started; nothing is sent until you press Send.

Settings (⚙)

The ⚙ button, at the start of the row with Board only and Arrange, opens Settings: every switch in ChessClerk on one page (the clerk's, the board's, practice and the screen). Each switch is also still where it always was, and the two are the same switch: change it in either place and both show it.

Keyboard shortcuts

Help > Keyboard shortcuts, or press ? anywhere: every button, tab, group, subtab and tool has a keyboard shortcut, with a box to find one. F flips the board, U undoes, R redoes, P is ▶ Play; ← → Home End step through the moves and B goes to the board, as always. Everything else is a first letter and then one or two more: G, a group's letter, a tab's letter opens the tab; S and a letter opens a subtab of the open tab; I and two letters presses anything inside the open tab; T for the tool row, N for the nav row, C for the chat bar, A for the corner row, O for a popup that is open. A small box at the bottom shows what comes next; Esc stops. Nothing fires while you type in a box. Each one keeps its keys on this computer, so they stay the same every visit. The clerk can tell you the shortcut for anything. The switch at the top of the sheet turns the letter shortcuts off (and on again), for anyone they get in the way of, like someone talking to a speech program.

Work offline

⚙ Settings > Offline > Make it work offline downloads one pack (about 90 MB) while you are online, so ChessClerk works with no connection for everything that runs on your device: engine games, puzzles (about 6,500 of them, every theme and rating; your puzzle rating keeps working), endgames and the King's ladder, drills, the motif cards and their example games, their opening trees and your Library, the showcase and famous games, the tutorial, and Review of your own games with our Stockfish. It says the size first and whether your device has room, shows how far it has got (Cancel stops it), and keeps itself up to date whenever you open ChessClerk online. Remove the offline pack frees the space. With no connection, the clerk and imports from Lichess and Chess.com say they need one. On a phone, add ChessClerk to the home screen too (Install as an app) so it opens like an app.

Show and tell

Under the clerk's first post are three buttons, all optional. Show and tell me all about this site is a guided tour: each Next opens a feature (a tab, a subtab, something on the board) while the clerk says what it does, in the chat; Stop ends it, and you can ask the clerk anything at any point. Look at my own games goes through your games. Watch a game plays one out for you to watch. Or skip the buttons and just talk to the clerk.

Language

ChessClerk opens in your browser's language, through Google's page translator, and switches by itself when you write to the clerk in another language. If you mix two languages, it asks which you want. English, in the chat's top strip, puts the screen back in English (it shows only while the screen is in another language); ⚙ Settings > Language picks any language Google offers. The board, its coordinates, the moves (Nf3 in every language), game files, names and the chat are never translated: the clerk answers in your language itself, and nothing you type goes to Google. In the move box you can also type the piece letters of your language (Sf3, Cf3, Кf3); the move shows as Nf3. A browser set to British English reads British spelling.

Coming soon

Help > Coming soon lists what is built or planned but left out until after the release: Coaching, Find a coach, Tournaments, Moderator, Share live, Playing other people. Each has a line on what it is and its projected release date.

Install as an app

On a phone or tablet, open ChessClerk's website and use the browser's Add to Home Screen (or the Install ChessClerk button when it appears). ChessClerk then opens like an app, and the board, puzzles, drills, the King's ladder and the engine work without a connection.

Accessibility

⚙ Settings > Accessibility. Always on: a screen reader reads each square with its piece, each move in words, check and mate, and the hidden Read the position button reads every piece; from the keyboard, Tab to the board (or press B), the arrow keys move the gold square, Enter picks up and puts down, Escape puts a piece back, and typing a move like Nf3 on the board goes to the move box. Choices, each On or Off: Screen reader announcements (on to start), High-contrast pieces, Larger pieces, Shapes as well as colors, Reduce motion (follows your device until you choose), Easier reading, Bigger buttons and Strong focus ring. Text size and color-blind safe colors are under The screen on the same page. Ask the clerk to turn any of them on.

Learn

Beginner tutorial

Learn > Beginner tutorial: how every piece moves, capturing, promotion and castling, check, getting out of check, checkmate, and the fork; then pins (and Légal's trap), skewers, discovered check and attack, deflection, the open file, a rook on the seventh, the pawn center, three strategies (trade when ahead, push the passed pawn, the rule of the square) and scanning the board both ways; each a task on the board (a wrong move is taken back with a pointer). Practice along the way: mating with king and queen against perfect defense, and games against the Stockfish that adapts to you. A lesson with a motif behind it offers See it in real games. Finished lessons get a ✓.

Serious business

When you're bored or flat, the clerk may say "let me teach you some serious business" and play one of the great players' great games on the board, from Morphy's Opera Game to Kasparov's Immortal, while the text panel shows its notes move by move. ⏮ ◀ ⏸ ▶| and the speed control it; the nav buttons work too. You can ask for one any time. Serious business on (Panel ▾ > Serious business at the top of the chat, or say "serious business on"; "serious business off" to stop; asking about it never switches it) keeps the clerk plain and businesslike, with no jokes and no great-game offers.

Guess the move

Learn > Guess the move, or Guess on any game in Research > Databases. The greats' games come first; in those you guess only at the moments picked, and the other moves are played for you. Pick a side; the game plays up to each of your moves and you guess the master's. The master's move scores 3; otherwise our Stockfish measures yours: a gap under 0.30 scores 2, under 1.00 scores 1. Then the master's move is played and the game goes on. At the end you see your points out of the most possible.

Puzzle Builder

Learn > Puzzle Builder: build your own puzzles for a motif, one step harder at a time. Pick the motif and how hard a puzzle you're aiming at (0 to 10), start from a real example one step easier, a blank board, the position on the board or a Lichess puzzle; Edit the board to change it, Mark the start, then play the whole solution (both sides, ending on your move). Check it: our Stockfish checks the line wins at least 1.50 or mates and that every move you must find is the one clear answer, and grades how hard it is, out of 10 (0: the motif is already on the board; 1-4: that many moves away; 5: three or more defenses; 6 or higher). Keep it puts it in your series on the motif's card.

Chess thinking

Learn > Chess thinking: moves of thought for the board, drilled by writing. Three lists: Thinking moves, by category; Drill the head-work, every head-work motif as a drill; and Grandmaster habits. Pick a move or drill to see its card: what it is, its shape (the sentence pattern), a model sentence and when it's clean. Drill it starts your message to the clerk; write the sentence about the position on the board or one of your games, and the clerk says whether it's clean and why, then takes it a step further. When you have it on more than one position, the clerk offers a button that marks it mastered. Grandmaster habits: ideas from Jacob Aagaard's Grandmaster Preparation books, in four groups (Calculating, Reading the position, Attack and defense, Endgames), each a lesson: what it is, why it works, how to do it at the board, positions from the greats' games (click one to put the game on the board) and a drill on a real position, judged by our Stockfish (Check it) and by the clerk (Ask the clerk). A drill you get right puts a ✓ on its idea.

Kotov

Learn > Kotov: Kotov's method on your own worst positions (or the board's). First the three questions (where is my worst piece, where are the weaknesses, what is my opponent's idea), then name every candidate move before analyzing any; then each candidate's branch once, in order, with no going back; then pick your move. Scored out of 100: 60 for naming the engine's best move, 10 each for up to two other candidates within 0.30 of it, 20 for the move you chose, and 5 off for each clearly bad candidate (1.00 or worse) you spent a minute or more on. Afterwards: the engine's best, what each candidate costs, whether your lines expected the opponent's best reply, the three questions answered with their moves graded by our Stockfish, and Ask the clerk about my tree.

Tagmemics

Learn > Tagmemics: look at one thing (a position, a move, a plan, an opening, a game) nine ways. The rows are Particle (it as a fixed thing), Wave (it moving through the game) and Field (it inside the web of relations); the columns ask Contrast, Variation and Distribution. Write a few words a box for a sketch or a full outline; it saves as you type, and the board's position can be kept with the grid. Ask the clerk puts your grid in a message; the clerk pushes on the thin boxes and doesn't fill them in. How and Why teach it; Your grids keeps them all.

Motif drills

Learn > Motif drills. Blindfold: a real position where a motif happened, given only as a list of pieces; see it in your head and type the move that was played. The board holds it under the cover; click to peek (counted). Recall: for the motifs no real game shows yet (or every motif), say what a motif is from its name, or pick its name from its write-up; the next card is one you haven't got yet, at random; nothing is scheduled. Made-up positions are always flagged.

Sacrifice or not

Learn > Sacrifice or not: positions from the greats' games where a sacrifice nobody could calculate to the end was there: some the master played, some the master turned down. Write why if you like, then choose Sacrifice or Don't, with the engine locked until you choose. Then you see whether you were right, our Stockfish's deep look (depth 20) and its quick look (depth 8), which can get it wrong, and what the master did.

Convert it, hold it

Learn > Convert it, hold it. Positions from the greats' games, checked by the tablebase. Bring it home: the great won from here; win it against perfect defense. Hold it: the great was worse and drew; hold it. A move that gives the result away is taken back and counted. The master's way puts the great's game on the board at that position.

Read the line

Learn > Read the line. A position from a great's game stays still on the board, and the moves that followed are written out. Read them, see the end in your head, and answer: where a piece stands, whether a square is attacked, whether it is check or mate, and what is hanging. All four right and the next line is a move longer. Show the end plays the line on the board.

Plans by structure

Learn > Plans by structure. Pick a structure the greats played in most (it goes on the board), or keep the board's own position, and pick your side. Say your plan starts your message in the chat: finish it in your own words, typed or spoken, and send; the clerk compares it with what the greats actually did in that structure (their piece moves, pawn breaks and where the kings went), plainly, with no score. That is a clerk turn and costs credit like any other. What the greats did shows their plan for your side, free, once you've said yours.

Chunking test

Learn > Chunking test: how much of a position you can take in at a glance (de Groot's test, and Simon and Chase's, 1973). Press From a master game: the test opens in a box, and a real position from the greats' games flashes on its board for 5 seconds (change the seconds in the tab if you like), then the board clears. Rebuild it from memory: pick a piece in the tray and tap squares (tap a piece again to lift it, or pick the eraser), drag a piece from the tray onto a square, drag a piece on the board to another square or off the board to take it away, or type the pieces (White Kg1 Qd1 a2, Black Kg8 a7: a square alone is a pawn; or a FEN's board). Done counts the pieces right (the same piece on the same square), wrong (a piece where the position had another or none) and missing (a real piece not on its square), and shows them on the board: yours ringed green or red, the ones you left out faint; Show the real position shows the real one. The same pieces at random puts those very pieces on random squares, the experiment's other half: real positions come in chunks a player knows, random ones don't. Every try is kept on this computer, as numbers by kind of position, in the tab.

Clerk

The chat

Type in "Ask me anything about chess" and press Enter or Send (Shift+Enter for a new line). Ask for a game and the clerk starts one at once, a random color against the Stockfish that adapts to you. Ask for a position, a line, a puzzle or a lesson and it offers a button under its post that does it when you press it. Every post is numbered. On each post: ✕ takes it out of the conversation, 📌 pins it so the clerk keeps reading it in every session, ✎ edits it. + New Session starts fresh: the clerk stops reading the posts above, except pinned ones. ☰ hides or shows the sections on the left.

Play

Cover (blindfold on any board)

Press Cover on the tool row to hide the pieces and the coordinates on whatever is on the board: a game, a line you're studying, a puzzle. Type moves in the box under the board, where both sides' moves so far are written; click the board to peek (peeks are counted). In a game under the cover you can also type your move in the chat. Press Cover again to uncover. A game started as Blindfold in the New game popup keeps its own gap; covering an ordinary game doesn't make it a blindfold game.

Play

Press ▶ under the board: a game starts at once, no popup, against the Stockfish that adapts to you, your color random, no clock. If the board shows a position, the game starts from it and you play the side to move. During a game the same button is ⏸ (pause) and ▶ (carry on).

New game

Spar in the opening lists starts a game at once from the line it puts on the board, against the Stockfish that adapts to you, and Play them in Opponent's pieces starts one at once against that player at the strength pressed: no popup, your color random, no clock. The opponent loads before the first move.

Opponents

Play and 960 are always against Stockfish 19 adapting to you: it plays slightly better than the median cp gap of your last 200 moves. In the New game popup, Stockfish can also play its best, or at that same strength in the style of Tal, Nezhmetdinov, Petrosian or Botvinnik.

Blunder popup

If your move costs more than your line, a popup says what it cost. The line follows your play: 2.0 pawns at 1200 and below, rising evenly to 2.5 at 2200 and above, from the rating your own games show; strong players give up material for the position on purpose more often, so their line sits higher. A move past the line is searched again, deeper, and if the material comes back it was a sacrifice and no popup opens. What's wrong shows the best move and what the opponent can now do; Show me on the board plays it out. Take it back to try again, or keep it.

Blunder check

In the New game popup, against the computer: Off, or 1, 3, 5, 7 or 9 plies. When your move costs more than your line and the opponent's punishment lies within that many plies, the punishment plays out three times, and the move is taken back: play another (the same move is refused that turn). A punishment deeper than the check reaches stands, so set it to 1 or 3 to leave the long tactics to you. With it on, the blunder popup doesn't ask.

During a game

You can step back through the game and try side-lines at any time; the opponent only answers at the newest move. ↶ takes back your move and the reply. ½ offers a draw (taken when the position is even or worse for the opponent). ⚑ resigns.

Pause

Play turns into ⏸ during a game: pause on your turn (the board is covered) and ▶ to carry on.

A game is not a mode

During a game you can still open a puzzle, a drill, autoplay, the Endgame tray or Solve. The game goes on hold; Back to the game (above the board, or ▶) brings it back exactly as it was.

Chess960

960, in Tools, starts a Chess960 game against the Stockfish that adapts to you, your color random. Its box asks two things: who sets it up (the computer, you, or each side its own), and whether the two sides are symmetrical or each its own. When you set a side, type its back rank, like RNBQKBNR. Castle by moving the king onto its rook.

Blindfold

Tick Blindfold in the New game popup. The pieces are hidden; type your moves in the box under the board, where the opponent's moves are written too. Click the board to peek; peeks are counted.

Games

Play > Games on the left lists your finished games. Click one to load it on the board. Cleanest first puts them in the order of your cp gap (book moves, forced moves and recaptures left out); a game ranks with about 20 moves out of book. In Add an over-the-board game, 📷 Scoresheet photo reads a photographed scoresheet: fix any move square by square, then Save it. A game marked not finished waits an hour before any tool works on its last position (fair play, as for 📷).

Read a scoresheet

📷 Scoresheet photo (Play > Games, in Add an over-the-board game; and Research > Databases, in Import) reads a photo of a handwritten scoresheet into a game. The AI reads the moves as written, the players, event, date and result; a sheet with one block of moves or two blocks side by side is read either way. Our rules check every move from the start, and only the moves that need you are asked about: To look at lists the moves that aren't legal, that two pieces could make, or that the reader wasn't sure of, and each shows its line of handwriting cut from the photo. Pick one of the moves offered, click the piece and the square it went to, or type it; Insert a move before puts in one the sheet missed, Delete takes out one written twice. In Play > Games the moves fill the box for Save it; in Databases the game goes into My games and Scoresheets and onto the board. A read uses AI credit (on the free trial, one free question).

Blank scoresheet

🖨 Blank scoresheet, beside 📷 Scoresheet photo in Play > Games, prints an empty scoresheet for an over-the-board game: event, date, round, board, both players with their ratings, 60 numbered moves in two blocks, the result to circle and both signatures. Any game prints with its notes and diagrams: 🖨 on a game in Research > Databases, or Print sheet in the Diagram popup for the game on the board.

Game folders

In Play > Games, Add a folder of games picks a folder on your computer: every PGN (or zip of PGNs) in it is read, and each folder becomes a headed list (a subfolder is its own list). Click a list's heading for its games, and a game to load it. Adding the folder again reads it fresh; Remove takes the list away (your files stay on the disk). Each list is also a database in Research > Databases.

Engine matches

Play > Engine matches, a bench for experiments: pick two to six players (Stockfish, the human-like engine at a USCF strength, the adaptive engine, or Tal, Nezhmetdinov, Petrosian or Botvinnik), and open each one's Settings to set it on its own: Stockfish's time, depth or nodes a move, its Skill and a rating limit (on Stockfish's own scale), the adaptive engine's target (so many cp stronger than the other player's measured gap, or than a gap you type), variety, its opening book, threads and hash. Match settings: how many games, where they start (the start, a FEN, the line on the board, Chess960 or a set of openings, each played with both colors), ending early by the score, games at once, how deep each move's gap is measured, and whether the board follows the game. Every setting has a default, so Start works at once. The results show two players' score with its margin in rating points, the crosstable, and every game with each move's gap; Download PGN saves them all with each player's settings in the headers. Save setup keeps a setup by name; the experiments ship with it; Past matches runs any match again exactly. It all runs in your browser: no AI credit. These games never count in your record.

Take over

Take over, in Tools, plays the position on the board out against the engine: from any moment in any game (a great game, Watch a game, a database game, a Library line, your own game, or the board now) you take one side and play on, to see if you can hold the win or save the draw. Go to the move, press Take over, pick your side (the side to move comes first) and the engine: At its best (the default) or Adapts to you. It is an ordinary game from there: the blunder popup, ↶, ½ and ⚑ work as in Play, it is saved in Play > Games and reviewed when it ends. Not while a game is running or on hold, and not when the position is already over. The clerk can start it too.

Eyes-free chess

Play > Eyes-free: speak your moves and ChessClerk speaks back, for walking, driving or blindfold training. Press Start listening (the browser asks once for the microphone). Say moves the way you'd say them: knight f3, knight takes e5, pawn to e4, castles, castle long, queen h5 check, e8 queen. When more than one piece fits, it asks ("Which knight, b1 or g1?"): say the square, or the first or the second. The other side's moves are said back ("Black plays knight to c6."). Ask: where's my queen (or their queen), what's attacked, read the position, whose move. Say: take back, resign (then yes or no), repeat, new game, hide the board, show the board, help, stop. Any popup that asks something is read out; say yes or no. Hide the board covers the pieces while Eyes-free is on, and with it ticked a new game is a Blindfold game (its own gap). It works in a game against the engine, Play vs in the endgames, a line or a puzzle, and on a board with no game. The microphone is closed while ChessClerk speaks. In a browser that can't listen, type what you'd say in the tab's box. The clerk can start it for you.

Tools

The tool row

Press Tools (the last nav button) to swap the nav buttons for the tools; Nav brings the nav buttons back.

✦ Highlights

Three states, a press apiece. Off (it starts off). Blue: the last move's two squares are ringed blue. Turquoise: that too, and touch any piece, either side's, to ring it and the squares it can go to. A third press turns it off.

Σ Engine move

Plays Stockfish's best move. Take it back (↶) and press Σ again for the second best, again for the third, and so on.

⚠️ Threats

Marks in orange every piece, of either side, that is hanging (attacked and undefended, or attacked by something cheaper), pinned to its king, or forked.

💡 Hint

Lights the piece to move for a decent move that isn't a blunder. In puzzles and line drills it gives that hint instead. Free.

Worst piece

Lights, for four seconds, the side to move's worst piece: its knight, bishop, rook or queen with the fewest moves. 💡 is the other half: a decent move that isn't a blunder. Free.

Solve

Plays the position out from here at two plies a second, then puts the position back.

Endgame (set up a position)

Opens a tray under the board with an empty board. Pick a piece and click squares; click a piece again to lift it off; ✕ is the eraser. Choose who is to move, or paste a FEN or PGN in the box. Done checks the position and puts it on the board.

W/S weak and strong squares

A small mark in the corner of every weak and strong square, for both sides: a ring for weak (no pawn of that side can ever attack it again), a dot for strong (an outpost no enemy pawn can ever attack, guarded by a pawn). Cream for White, dark for Black.

Diagram

A picture of the position with its arrows and colored squares, and the moves so far: download it, print it, or copy the FEN or PGN. Print sheet prints the whole annotated game: comments, symbols and side-lines, with diagrams two to a row.

Paste a position or game

Copy a FEN or a PGN from anywhere and press Ctrl+V on ChessClerk (outside a text box) to put it on the board.

Eval

Press Eval on the tool row to turn the eval under the board on or off (off: no searching). It says the position's score and depth, and how much the last move changed it, for the side that played it. Your choice is remembered.

Evaluations as

Under the board, Evaluations as picks how every evaluation in ChessClerk is written. Moves: cp gap (how much worse than the best move, in pawns: 0.35), accuracy % (the win chance before the move against after it, by the formula Lichess publishes; for a single move the position before it counts as even unless the page knows its score), or estimated USCF rating (the USCF rating of a player whose median gap equals this move's gap; a move with no gap shows the top of the scale). Positions: + or - pawns, or the win chance % for the side named. A median gap in Tracking is written the same way. Mates always show as mate. The choice is kept on this computer.

💬 Explain each move

Press 💬 on the tool row and every move made on the board gets one plain line from the clerk in the chat, with what it cost in AI credit (about a tenth of a cent a move; Haiku writes them). Our Stockfish checks the move first and the clerk is told what it found, so it never guesses the eval. Off during a game against the engine. Press again to stop.

📷 Photo to board

Press 📷 on the tool row and pick a picture of a board, a book diagram or a screenshot (on a phone the camera opens too). The clerk reads the position and answers with a Show the position button; nothing changes on the board until you press it. A position that can't happen is refused. Fair play: for an hour after the picture is sent, no tool works on that position or on the game played on from it (the engine, the eval, the human-like engine, the tool row, the analysis tabs, the clerk's look at the board); a countdown under the board says how long is left.

Heat maps

Heat, in the tool row, draws a heat map on the board itself. Three states, a press apiece: off (it starts off), Control, Blind spots, then off. Point at a square (or tap it) for its count. Turning a map on puts its short line in the chat and along the foot of the board, until you point at a square.

Heat: square control

For the position on the board now, and it follows every move: each square is tinted by who has more pieces attacking or defending it, blue for White, red for Black, stronger the bigger the lead (up to three). A dashed ring marks a contested square (both sides bear on it). A gold corner marks a hole: a square on a side's own third or fourth rank that none of its pawns can ever attack again. Pointing at a square gives both counts.

Heat: blind spots

Across your own games (your games here and your measured imported games, as the skill radar counts them): the squares where you missed a tactic (where the deciding piece would have landed, and what it would have hit) and where you allowed one (where your opponent's reply landed and what it hit). Read from your side: your Black games are turned round, as if you had White, and the board turns White-up while it shows. The line names your hottest squares, and says plainly when there are too few tactics on record to call it a pattern (it needs 10).

Greats

Opens Research > Greats in your position for the board: the closest pawn structures in the greats' games, who had them, and what they played.

Best practical move

Tools > Practical (and in My study > Review and at the top of My study > Analysis): the engine's top move assumes a perfect opponent; this shows which move gives a real opponent the most chances to go wrong. Set the opponent's strength in USCF (it starts from the game on the board: the human-like engine's strength, or a rating in the game; otherwise your own, or 1200 before you have games here) and press Find. For each of the engine's top 5 moves, in the engine's order: its engine score in cp, the expected cp against that player (their likely replies at that strength, each scored by our Stockfish), how often they find a good reply (within half a pawn of their best) and their likeliest reply. The practical choice is the one with the highest expected cp; the line on top says where it differs from the engine's first choice. Not during a game being played. Our Stockfish and the human-like engine on this computer: no AI credit.

Human chances

Tools > Human (and in My study > Review, under Best practical move): how real players move here, at a strength you set in USCF (it starts from the game on the board, or your own). Spread: how many likely moves a player of that strength has here, from the human-like engine's chances: one obvious move, a few, or scattered. For each of the engine's top moves: its trap value (the cp the opponent gives back on average: each likely reply's chance times what it costs them against their best), how often they go wrong, and how scattered their replies are after it, so you can see where they're most likely to go wrong. Swindle mode: when even the engine's best leaves you 3 pawns or more down, the move whose likely replies most often let you back within 1 pawn, instead of the engine's least-bad move. Not during a game being played. Our Stockfish and the human-like engine on this computer: no AI credit.

Zugzwang finder

Zugzwang, in the tool row, finds mutual zugzwangs: positions where whoever must move loses or does worse. Press it once: it reads the position on the board and says in the chat what it is; with 7 pieces or fewer the tablebase is asked with each side to move, so a mutual zugzwang is proven, and both kings' squares are ringed in gold. With more than 7 pieces it says plainly it can't be proven. It follows the board as you move or step through a game and speaks up when a position is a mutual zugzwang. The first press also reads your own games (My games) and lists the mutual zugzwangs their endings reached, each with Look. Press again to turn it off. It is grayed during a game being played.

Your win chances from here

Tools > Chances (and in My study > Review): how often players of your strength win, draw and lose from the position on the board against an opponent of a chosen strength. It plays many quick games out on this computer: your side by the human-like engine at your strength, theirs at the opponent's, each move drawn by how likely a player of that strength is to play it. A playout runs up to 15 moves a side; one that isn't over by then, or that the human-like engine already calls decided, is scored by the human-like engine's own win, draw and loss odds there. It stops after 200 playouts or about 90 seconds, or when you press Stop, and says how many it played and its margin of error (95 times out of 100 the real score is within it). Your side starts from the game on the board (the side called You, else the side at the bottom of the board), your strength from your games here (USCF 1200 before you have any), the opponent's from the game on the board, else your own; change any of them. It is an estimate. Not during a game being played. No AI credit.

Watch the engine change its mind

Tools > Depths (and in My study > Review): the engine searches deeper and deeper (depth 1, then 2, then 3 ...), and its best move can change on the way. Pick how deep to go (12, 16, 20, 24; 20 to start) and press Search: the table shows its best move, score and line at depths 1, 2, 4, 6, 8, 12, 16, 20, 24 (as deep as you asked), the depths where its choice flipped marked "changed". The line on top says each change of mind and names the horizon effect when a move that looked good at a shallow depth fails deeper: the shallow search stopped before it saw the refutation. Pick a row to see that depth's line on the board as arrows (the first three moves, the first boldest); it comes off when the board moves on. Not during a game being played. Our Stockfish on this computer: no AI credit.

Openings

The opening's name

Under the board, in gold: the name of the opening (with its ECO code), deepening as the moves go on. It knows every one of the 3,865 Lichess opening names, found by position, so a line reached by a different move order is still named.

Openings tab

Learn > Openings on the left. On the board: the opening on the board with its idea, each side's plans and the structure, and Read on Wikipedia (opened at the right section). Find an opening: type any name, like najdorf poison pawn, and click it to put it on the board, two moves past its defining move. Opening ideas: 13 chapters of ideas, each with an intro and example lines to load, and In their opening trees: the lines of their opening trees that go with that idea, each with its own idea under it when it has one.

Opening book

In games, the opponent plays from their opening trees first: lines in proportion to their quality, within the opponent's rating band, never repeating the last two games' openings straight away, and never in Chess960. Untick Opening book in the new-game popup to turn it off.

Explorer

The cp-gap explorer

Tools > Explorer shows it under the board, in place of the eval and the captured pieces. Moves are in cp-gap order: the best move first, each move's gap (how much it gives away, in pawns; under 0.10 counts as 0), never by popularity. Choose the side you play: on your side only moves within 0.4 of the best are listed; on the opponent's side every move, with its full plunder. Click a move to play it.

Games and scores

Each database the crawl has counted gets its own column: how many games had the move, and the side to move's score. They are never added together.

Marks

✦ chessquirk: a move played less, and scoring worse, than a move that is objectively worse (at least 30 games each). ⇅ inversion: a worse move that scores better than a better one. ◇ neglected: a sound move (within 1.2 pawns) nobody in that database has played, though the position has 200 games or more.

What's wrong

Every move with a real gap has it: what the move gives away and the line that punishes it, with Show me on the board.

☠️ Plunder

Its own button on the tool row (and at the top of the explorer), in analysis and in games against the engine: goes through the line on the board and lists every mistake the opponent made (half a pawn or more), with the best move and how to punish it. Click one to go there; Show the punishment plays it out.

Where the evals come from

Our own Stockfish 19 only. Positions the crawl on this PC has reached come from its gapbook at depth 24; the rest are searched here at depth 18 and remembered on this computer. tools/sync-explorer-data.js brings the crawl's newest work over.

By any order and by this order

Under the explorer's status line: how many of a database's games reached this position by any order, and under it how many came by exactly the moves on the board (counted from the TWIC game files by tools/makeorders.js, the first 24 plies; an order in fewer than 5 games shows as under 5). Shown when the line starts from the normal position.

Ends

Where each move ends with perfect play, for the side making it: wins, draw or loses. With 7 pieces or fewer it comes from the Lichess tablebase straight away. From 8 to 12 pieces, press Ends ▸: our Stockfish plays each of the first six moves listed down to 7 pieces (at most 60 plies, depth 16), then the tablebase says how it ends; ? when best play doesn't get down to 7.

Puzzles

Puzzle

Tools > Puzzle (or Puzzle on the library row) starts a puzzle at once, nothing to choose. The dropdown beside it picks a puzzle type, every motif and theme there are puzzles of; with none picked the puzzles are random. During a game, the game goes on hold.

Solving

The opponent's move that sets the puzzle up is played for you, with you at the bottom of the board. Find the best move; the replies are played for you. A move is right if it is the puzzle's move, gives mate, or is just as good by our Stockfish. A wrong move is taken back and you choose Try again or Show the solution. 💡 lights the piece to move; Solve shows the solution. The live eval and the explorer are off while you solve.

Puzzle sets

Learn > Puzzles > Puzzle sets: pick a motif or theme, how hard (How hard: 0 to 4 of 10, or 6 or more, as the Puzzle Builder counts it, each a little above your puzzle rating; or any rating) and how many, name it and Save. It is kept on this computer; come back and work through it any time. Each set shows how many are done and how many solved. Nothing is scheduled.

Worksheets

Learn > Puzzles > Worksheets: pick one of your sets or a puzzle type, 6, 9 or 12 to a page and how many pages, and Print. Each diagram says who is to move; the answers are on the last page. In the print box you can also save it as PDF.

How hard

ChessClerk works out how hard your puzzles should be from how you do on puzzles alone: each one aims a little above what you've been solving, higher when you solve every one, lower when you miss a run of them. Repeats don't count.

Drills

Name the square

Learn > Drills. Find it: a square's name is shown, click that square. Name it: a square lights up, type its name and press Enter. Light or dark: a square's name is shown with no board; say its color (the Light and Dark buttons, or L and D). From White's side or Black's. It counts right and wrong. Not timed.

Knight tours

Learn > Blindfold. Pick a board size (3x3 to 8x8) and a tour: the 3x3 loop by the phone keypad (1-6-7-2-9-4-3-8), both ways; the longest 4x4 walk; 5x5 to 7x7 tours and their mirrors; and the sixteen 8x8 tours (each corner, each first move, outside in and inside out). Watch it walked; Play it by clicking the squares in order (💡 lights the next one); or do it Blindfold, typing each square's name, with Peek for two seconds.

Library

Your lines

My study > Library. Play a line on the board, step to where it should end, and press Save the line on the board: it keeps the moves from the start to the move shown, named after the opening (or what you type), with you playing the side at the bottom of the board. Lines are listed in SAN alphabetical order; click any move to see the line up to it, and double-click a move to mark it gold (double-click again to take the gold off); the gold stays with the line.

The clerk's lines

Ask the clerk for an opening's important lines (like 'put the Najdorf lines in my library, I play Black'): it offers a button that fetches the lines of their opening trees and the named lines inside that opening, builds lines from its position off our Stockfish and the databases, or both, into your Library, as deep as you choose (Main road, Solid or Everything).

Autoplay

▶ Autoplay plays a line through on the board; Autoplay all plays every line shown, in order. − and + set the speed, from 0.4 to 7.4 moves a second.

Play vs line

Play vs: you play your side, the line's replies come by themselves. Only the line's move is taken; another is taken back and counted as a slip. Press 💡 when you want a hint: it lights the square of the piece to move. Play vs comes at every size, smallest to biggest: a line (its own Play vs), a bundle (Play vs bundle, the lines shown), a variation (Play vs variation, every line through the one picked, like Sicilian Defense: Najdorf Variation, English Attack), an opening (Play vs opening, every line of it, like Sicilian Defense) and all (Play vs all, every line in your library), each shuffled. The eval is off meanwhile.

Move orders

Move orders on a line lists the other ways into its position that the opening data knows: the named Lichess lines and their opening trees, each with its name and moves. Click one to put it on the board. Press again to close the list.

Bundles

Type part of a name or some moves in the box (like najdorf, or Nf3) to narrow the lines; Autoplay all and Play vs bundle then run over just those. Stop ends either one.

How deep

How deep you want to know an opening is yours to choose: Main road (the main line and the most-played replies, about 10 moves), Solid (the main sidelines too, about 15 moves) or Everything (the whole tree). In the Library it sets how far into each line Play vs goes and how deep Check coverage looks; in My study > Lines for an opening, and when you ask the clerk, how long and how many the fetched lines are and how far Build grows.

Practice what's due

Every line you play through in Play vs, at any size, gets a time it is due again: played cleanly, the wait grows (a day, then about 2.6 days, a week, three weeks and on); a slip or 💡 shortens it. In the Library, Practice what's due picks Bundle (the lines shown), Variation or Opening (the ones picked) or All, and plays the due lines there, the ones you miss most first, as deep as How deep says. A quiet count says how many are due, and each line says when it is due. Nothing reminds you or nags: you practice when you choose. It is kept on this computer.

Map of your lines

Map, in the Library (the lines shown) and in My study > Repertoire (the repertoire shown), draws your lines as a map, left to right by move number: each position a box, your moves gold and theirs cream. Two move orders that reach the same position meet at one box, marked ⇄ (its tip lists the orders). Holes marks every common reply and trap you have no answer to with a red !. Click a box to put the position on the board; the one on the board is lit. The map scrolls inside its box; − and + zoom, and so do a pinch on the phone and Ctrl with the wheel.

Coverage check

Pick an opening of your lines and press Check coverage: every opponent reply played in 1% of games or more that your repertoire has no answer to, most played first, like "Your repertoire has no answer to 6.h3, which comes up in 4% of games", as deep as How deep goes. Click the sentence to see the position; Build the answer grows lines from it into your Library (your best moves, their most played replies).

Repertoire

Your repertoire

My study > Repertoire, one for White and one for Black. Play your line on the board, step to where it ends and press Add the line on the board: your moves become your answers in those positions. A position can have several answers; adding never replaces one, and ✕ takes one out. Transpositions find the same answers. On the board shows your answers wherever you are.

Gaps

Find gaps walks your repertoire and lists every opponent reply played in 2% of games or more (in the database with the most games there) that you have no answer to, most played first. Click one to see it on the board.

Map

Map draws the repertoire shown as a map: positions as boxes left to right by move number, your answers gold and their replies cream. Where two move orders reach the same position they meet at one box marked ⇄. Holes marks the common replies and traps with no answer in red. Click a box to put it on the board. It scrolls and zooms inside its own box.

Leaving your repertoire

When a game against the engine ends, ChessClerk checks it against your repertoire. If you left it, the chat says where, what your line was, and what it cost by our evals.

Holes and traps

My study > Repertoire > Holes and traps: checks the repertoire shown (White or Black) against the replies the databases reach most and the traps at each of your positions, and lists every reply you have no answer to, most reached first: a common one (2% of games or more) or a trap (the answer people usually play there gives away 1.50 or more by our gapbook; it names that answer). Click a row to put the line on the board and answer it yourself, or press Add and the move named (the gapbook's best answer) goes into your repertoire.

Novelty alert

When a database game on disk reaches your repertoire and the opponent plays a reply your repertoire has no answer to, it shows the next time you open the site: a quiet line in Help > Today (only when there's something new) and the list under Holes and traps in My study > Repertoire, new ones marked New. Click one and that game goes on the board at the reply, to step through to the end. It reads the crawl's over-the-board games on this computer (signed in), so it grows as the crawl reads more; nothing is sent anywhere.

Endgames

Endgames tab

Learn > Endgames judges the position on the board: who wins with best play, from the Lichess tablebase (exact, up to 7 pieces) or, for 8 to 10 pieces, our Stockfish at depth 22 (press Ask Stockfish). Set up any position with the set-up tray, or press Random endgame for a position of one of 12 kinds.

Play vs, Autoplay, Memorize

Play vs: you play the side to move and perfect defense answers. Keep the result takes any move that keeps the win (or the draw); Zero gap takes only the fastest. A move that doesn't count is taken back; 💡 lights a right move, Solve plays the rest. Autoplay plays it out perfectly at the library's speed; Memorize does the same and then goes back to the start for you to play it from memory.

Why each move, and Play it down

Autoplay and Memorize say why each move is played: the only move that keeps the win, the fastest of several, one of several that keep the draw, or the longest defense, with the result. With 8 to 12 pieces on the board, Play it down to 7 has our Stockfish play both sides until 7 pieces are left, then the tablebase says exactly how it ends; Put the 7-piece position on the board takes you there.

Famous positions and 100 Endgames

Lasker-Reichhelm, Lucena, Philidor, Réti and Saavedra, and the 100 Endgames chapters (100 positions, each with what it should come to). Click one to put it on the board.

Teach me

In Learn > Endgames, Teach me teaches the position on the board (set up, random, or from a game), or one your own games reached (From your games): who wins with best play, which ending it is (its King's ladder rung, and the named position best play reaches, like the Lucena Position, as the motif detectors name it), the perfect line in three to five steps with what each step does, Watch (the board plays it step by step, each move with why), and Test me at the moments where only one move keeps the result. Seven pieces or fewer are taught from the tablebase, every word proven; 8 to 12 are first played down to 7 by our Stockfish. Ask the clerk to explain puts the facts in the chat box.

Trade drill

In Learn > Endgames, the trade drill reads your own games (My games) for the moments where a trade decided the ending: the trades that threw it away, the won endings you missed, and the ones you got right, all judged by the tablebase. The drill puts the ones that went wrong back on the board (Trade, or not? Which trade?); a moment comes back until you find it cleanly twice in a row. Prove it plays the ending out; Teach me this ending teaches it. Practice trades gives random trade exercises when your games run short.

King's ladder

Learn > King's ladder: the whole endgame from the end, in 35 rungs: the kings and the opposition (close, distant, diagonal and at every distance), the kings' battle (one king dominating), king and pawn (every file, and the opposition from far away), a pawn each (the same file, next-door files, far apart), more pawns, counting tempi (who has to move, as a sum), then the queen and rook mates the fast way and the textbook way, then each piece in turn as it joins, with the king always in, the two minor-piece mates, the rook and queen endings, the surprises, six and seven pieces, eight to ten pieces, every endgame, watching any position solved, trading into a won ending, and choosing your first move by the ending you want. Each rung has its lesson, its ideas, its themes from your corpus (all 221 endgame themes are placed), its tabiyas with Look, Tree (the tablebase's perfect line with the other defenses as side-lines) and Watch, and a test of the right kind; the mates are tested zero gap. A clean test puts a ✓ on the rung. A theme's motif card opens its rung. Every rung has real lessons: the explanation, Watch (the board plays the idea move by move with why), See it (each claim on the board) and Solve (positions against perfect defense; solving them all cleanly masters the lesson). Every rung is open to everyone from the start; nothing is locked.

Counting tempi

Rung 10b of the King's ladder counts who has to move, the way combinatorial game theory does (Berlekamp, Conway and Guy, Winning Ways; Noam Elkies, On numbers and endgames). A position where whoever moves loses is worth 0; each spare tempo (a waiting move that changes nothing else) is worth 1 to its side; parts of the board that don't touch add up. Read the sum: more than 0, White wins; less than 0, Black wins; 0, whoever must move loses. A square either side can step into is a star, and goes to the side to move. The opposition and corresponding squares are the same question, and Nim shows how separate parts add. Every position has 7 pieces or fewer and is proven by the tablebase. The test: find the waiting move.

Your own endings

The King's ladder's last row, 34. Your own endings, is made from your own games (My games: the games you play here and the ones you import). It finds the endings you misplayed: with 7 pieces or fewer, a move that threw away the result the tablebase knows (a win that became a draw or a loss, a draw that became a loss), proven; with 8 to 10 pieces, in a game our Stockfish has already checked, a move that lost at least a pawn and the result with it, marked our engine's best, not proven. Each game is read once and only its first misplay is kept: the moment it went wrong. Each one sits under the rung that teaches that ending (king and pawn against king under rung 5, rook endings under the rook rungs), with Look, Your game (the game on the board at that move) and Play it from here (play it again against perfect defense). With none, it says so.

Review and tracking

Review

My study > Review measures every move of the game on the board with our own Stockfish (depth 12): each side's median gap and what it stands for in USCF, how many moves gave away 0.50, 1.00 and 2.00 or more, the accuracy %, the turning point (the first move giving away a pawn), and the moves where the game turned with the best move: Go shows it, Show me plays what didn't happen. What did they want? asks, at each side's three worst moves, which move the opponent wanted if they could move again. Mark the game writes ?!, ? and ?? and the gaps into the game. Every game against the engine is reviewed when it ends, with the summary in the chat.

Audit

The deep review: depth 22 on every position, and depth 26 again wherever a move gave away 0.20 or more or the best two moves were within 0.15.

Tracking

My study > Tracking: your median gap over the last 200 moves, blindfold apart, and, once you've set a goal with a rating (Your goal), the rating it stands for; the trend, each 50 moves; by phase; results against weaker, equal and stronger opponents; your gap when up, even or down 1.5 pawns; your next five moves after a blunder; your imported games (My games not played here), measured by our Stockfish in the background when nothing else is thinking, with their median gap; the skill radar; thinking patterns by opening (the thinking motifs the detectors find for you in the first 12 moves, opening by opening); and your 15 worst moves, each clickable. During a game, once after a blunder, if three of your next five moves each cost a pawn or more, a note says so.

Skill radar

On Tracking: one axis per kind of move, each your median gap there turned into a USCF rating (the number shows once you've set a goal with a rating): Openings (your first 15 moves), Middlegame, Endgames, Converting (1.5 pawns or more up), Defending (1.5 or more down), and Tactics (the motif detectors: tactics found, missed and allowed). Your moves in your imported games count too, once our Stockfish has measured them. An axis is drawn once 10 games have moves on it; click its name for its 10 costliest moves.

Compare with the human-like engine

On Tracking: the human-like engine, at your strength on our curve, looks at the same positions from your last 100 moves. It says how often you played our Stockfish's best move, how often it would have, how often you played what it would, and where one of you found the best move and the other didn't. It runs in the page; nothing is sent.

Analysis

My study > Analysis keeps games and lines you're working on, with their side-lines and comments: Save the board here, Open to carry on. A library line's Analysis button puts it here too.

What about my move?

In My study > Review, on any game on the board: go to the moment, press What about my move?, and type the move you wanted there or play it on the board. Our Stockfish gives its cp gap against the best move (and the game's own move's gap when the game went another way), and either plays out its refutation or says it was fine (under half a pawn). Each line has a small button after it that plays it out on the board: your move and what follows (✗ when it gave away half a pawn or more, ✓ when fine) and the best line (✓). Try another asks again at the same moment. The game itself is never changed. Not during a game; there the blunder popup does it.

Motifs

Motifs

Learn > Motifs: every motif of the corpus. Chapters (in gold) open to their groups and motifs; or see them in order of how often they decide games, counted by the crawl, for people, for GMs, or composite; or How common: how often each motif comes up in 1,000 games, how often players see it and miss it, and the share missed, by band (GMs and IMs over the board their own bands), the most common first, for the kinds you tick; or open one of the two categories, Attacking the King and Defense; or type any words of a name.

Attacking the King and Defense

Learn > Motifs, the Attacking the King and Defense buttons: two motif categories, each a way of playing in sections (for the attack: when to attack and when not, the sacrifices that strip the shelter, the order of operations, the pawn storms, the heavy pieces, the motifs that lead you in, the defense the attacker must foresee; for defense, the Petrosian school: prophylaxis, overprotection, the exchange sacrifice, restraint, the king to safety early, simplifying, counterattack, the squeeze). Every idea opens its motif's card and has three games: how a grandmaster did it, how a good 2000 did it and how a good 1600 did it. Click a game to put it on the board just before the move.

A motif's card

Its intro; Try it (solve its position on the board; 💡 lights the piece); real-game examples from the crawl, one at a time on the board, by how hard (0 to 10) or all, with who played it and our gap, and Autoplay all; a Lichess puzzle on its theme; the families it belongs to; and your own puzzles for it, kept as hard as you say.

Motif map and mastering

The Study path's top line is your motif map: mastered, recognized (you solved a puzzle on its theme or built one for it), named (you opened its card) and unseen. On each motif: Example plays the next real-game example as hard as this pass; Puzzle starts a Lichess puzzle on its theme; Build opens the Puzzle Builder; Explain starts your message to the clerk with Explain-back: explain the motif in your own words, and when it holds the clerk offers a button that marks it mastered and finished in this pass.

Study path

My study > Study path: every motif in the crawl's order of importance, in passes. Choose how hard a pass goes (say 0 to 1 of 10), mark motifs Finished, move the ones you see in your games to the front with I see this, and press Pointless from here when the rest of the pass isn't worth it: the next pass starts from the top, a step harder.

Families

My study > Families: 46 families of openings and structures with their signals, idea, plan, motifs, endgames and antidote. Side by side compares any two.

Databases

Databases

Research > Databases keeps game databases on this computer. Import PGN files or zip files (TWIC comes zipped), or your own games: My Lichess games (you sign in to Lichess) and My Chess.com games (you prove the account is yours by putting the code shown in your Chess.com profile's Location for a minute). Only your own account's games can be imported, never someone else's. Once one is set up, your new games there come in by themselves each time you open ChessClerk (in the background, only the games since last time, never one twice), and our Stockfish measures them like any imported game. Every game is read by our own rules; games are never cut short. Remove duplicates (same players, year and first 30 moves), Export PGN, Delete. 📷 Scoresheet photo in Import reads a photographed handwritten scoresheet, checks it by the rules and files it into My games and Scoresheets (see Read a scoresheet). 🖨 on a game in the list prints it with its notes and diagrams.

Check with the engine

Pick one database and press Check with the engine. Our Stockfish searches every position of every game (depth 10) and each game keeps both sides' median cp gap and blunders (1.5 pawns or more), shown in the games list. It can take a while: Stop keeps what is done, and Check again carries on with the games not yet checked. Each checked game shows its cp gap (book moves, forced moves and recaptures left out, both players' average), and sort: cp gap lists the cleanest first.

Editing a database

Pick one database. Rename a player: type the name as it is and as it should be, and every game with that name (either color, capitals not minding) gets the new one. ✎ on a game in the list opens its tags (event, site, date, round, players, result, ECO, ratings) to change. Undo the last edit puts the database back as it was before the last rename or tag change.

Finding games

Filter by player names (one name finds either color; two names find their games against each other), event, years, result, opening code, lowest and highest rating (both players), and length. Games with the position on the board finds every game reaching it in its first 30 moves, by any move order. Click a game to put it on the board.

In the crawl's games

Beside Find structure: the same pawns (exact or at least) in the crawl's games of every database and rating band: how many games reached them, how they ended, how many from each database, the piece moves most played while they stood, and up to three examples (click one to put it on the board). Plans for this structure adds those piece moves after their opening trees' lines.

My study

Studies

My study > Studies: name a study, then Add the board as a chapter (its moves, side-lines and comments come too). Find this marks the move shown as one to find. Quiz me plays the chapter and stops before each marked move for you to find it. Load puts a chapter back on the board. Everything stays on this computer.

Calculation

My study > Calculation, for the position on the board. Write your candidate lines before moving (in SAN, like Nxf7 Kxf7 Qh5+), each with a note; they stay with the position. Check them: each is played out and our Stockfish (or the tablebase, at 7 pieces or fewer) says where it ends. Solve it cold: write the whole line without moving a piece; each of your moves is checked (it must keep the result, or be within 0.30 of the best) and the first wrong one is named.

Progress card

My study > Progress card: a square picture of your last seven days (puzzles solved and your puzzle rating, games played, your moves' median cp gap against the week before, King's ladder rungs passed) to post anywhere. Save picture downloads it; Share opens your phone's share sheet. Only what is true goes on it.

Composite

My study > Composite: the block's family on one page. On day one, sketch the nine-box grid on it, a few words a box. At the end of the block, the push: the full grid with your sketch above each box, each box opening a step of the whoosh, and last of all your rule, one sentence. Finish the whoosh keeps it in Your rules.

What's costing games

My study > What's costing games: from thousands of real games checked with our Stockfish at depth 25 and read by every motif detector, the motifs costing players at your strength the most: how often the engine's best move made one, how often it was found and missed, what a miss gave away, and what the misses cost a hundred games. Every band is on the over-the-board scale, with its own heading, and you can pick any of them; beside each motif, how often it came up for you in your own games.

Puzzles from real games

Puzzle, then its type list: From real games (or Learn > Puzzles, under Games from), when the crawl has published some: positions from real games, from your strength up to grandmasters, where the other side's move handed the player a winning chance. Each is graded by the player who faced it (found by a 1700: 1700; missed by a 1700: 1900), and you're dealt the ones players 100 to 200 points above you handle.

Openings by rating

The explorer's By rating box: for the position on the board, what players of each rating (over-the-board ratings) actually play here and how it scores for them, and the named traps: how often each is walked into at each rating and how it scores for the side that sets it.

Ask why

My study > Ask why, or just ask the clerk: on any position of any game, why a move was played, or why another wasn't. Our Stockfish looks at the move and its best three, and the motif detectors say what each one makes, what it allows the other side and what it sets up, along its own line. In a master's game, tick the moves players rated 1100, 1500 and 1900 pick most (the human-like engine) to see why the master didn't play the move that looks good. The tab shows the facts for free; the clerk tells them in words.

Explain your moves

My study > Ask why > Explain your moves, or its switch in the New game popup. On: in Play, after each of your moves the game waits (the eval hidden until you've given your reason) and the clerk asks why you played it. Book moves, recaptures and only moves aren't asked about. Your reason is kept on the move as a comment, so it is in the game's PGN and in later reviews. Answer in the chat, typed or with 🎤; the clerk says back your reason, whether the move was the best and, if not, why not, and gives the better moves with the idea behind each, from our Stockfish and the motifs. Back to the game carries on. Go through the game does the same for any game on the board, one move of a side at a time; say "skip" or "stop".

Your stats

My study > Your stats shows your record from the games you played here: your median cp gap overall and over your last 50 moves, by phase, by kind of endgame and blindfold, your share of blunders, your results with each color, and, once you've done a puzzle, your puzzle rating with your strongest and weakest themes. A stat that isn't there can be asked for in the box under them; it is kept so it can be added.

Lines for an opening

My study > Lines for an opening puts an opening's important lines into your Library: type its name, pick your side, and Fetch (the lines of their opening trees and the named lines inside it), Build (lines grown from it off our evals and the databases) or Both, and how deep you want to know it: Main road (the main line and the most-played replies, about 10 moves), Solid (the main sidelines too, about 15 moves) or Everything (the whole tree). The clerk does the same when you ask it for an opening's lines.

Model games

My study > Model games. Pick one of the 46 families; its model games are greats' games where that structure stood on the board. Play through takes the great's side and stops where the plan is carried out: guess the move there (scored as in Guess the move); every other move is played for you. Show the plan reveals the family's plan.

Classic studies

My study > Classic studies. The great composed studies (Réti, Saavedra, Troitzky, Kubbel, Kasparyan, Grigoriev and others), each checked by the tablebase or our Stockfish. Solve it cold puts one on the board and opens Calculation: write the whole solution without moving a piece, and each of your moves is checked. Done marks the ones you've solved.

Engine lines

My study > Engine lines shows the engine's top three lines for the position on the board, each with its score and its moves, and follows the board as it changes. It is off during a game against the engine, as the eval is.

Look at my own games

Look at my own games (a button in the clerk's first post, or ☰ Menu > My own games): once your Lichess or Chess.com account is proved yours (Lichess sign-in, or the code in your Chess.com Location), our Stockfish reads your newest 20 games, and the motif detectors find the two or three motifs you missed most, each shown in your own games with the red box, a click stepping through it on the board. Then the clerk talks it through in the chat. The rest of your games come in and get measured in the background.

Event folders

My study > Event folders: type an event's name (your own name for it; a folder you have is offered as you type), say which side you had, and press File the game on the board, or File beside one of your finished games. Press a folder to see its games (Put it on the board, Measure, ✕ to take one out) and Report: across those games, the cp gaps by phase (your moves, the median gap, how many gave away a pawn or more), the motifs you missed and the ones you allowed (the same reading as the skill radar's Tactics), and your costliest moves. A game your record doesn't hold (played somewhere else and not measured yet) is counted once you press Measure. Only what you put in; nothing is filed for you.

Your patterns

My study > Your patterns. How easy you are to prepare for: from your own games (My games, and your Lichess and Chess.com games once your account is set up here), for each color: how often you play the same move in the same position in your first 10 moves (positions you've had 3 times or more), where an opponent preparing for you would aim (the deepest line 25% or more of your games reach, with your score there), your set replies (one move 80% of the time or more) and where you score worst. Click a line to put it on the board. Under pressure: from your moves on record, your median gap and share of blunders (1.5 pawns or more) as usual, in the 5 moves after a blunder, short of time (under 10% of your clock, in games that kept clock times) and when clearly winning (3 pawns or more up by the engine).

Test my lines

My study > Repertoire (and Test my lines in Tools > Chances): plays the end position of each of your lines out many times, the same way as Your win chances from here (your side at your strength, the opponent at the strength you choose), and ranks the lines by how well a player of your strength scores from there, with the count of playouts and the margin of error for each. Pick your White or Black repertoire, all your Library lines or one opening of them. Tick "with the engine for my side" to also see each line's ceiling: the score when your side plays Stockfish's moves. About 60 playouts or 25 seconds a line (twice that with the ceiling), with progress as it goes; Stop keeps the lines already done. The last ranking is kept on this computer. Estimates, not verdicts. No AI credit.

Research

Novelties

Research > Novelties finds good moves nobody has played: within 0.30 of the best by our Stockfish, in a position the databases know (50 games or more), with no game in them. Novelties here looks at the position on the board; Deep novelties looks three moves out, down the two most played moves at each step. Each is followed 6 plies and filed in the tab; Open puts it on the board.

Deconstruct

Research > Deconstruct finds realistic move orders from the start that reach the position on the board in the same number of moves. The human-like engine, at the rating you pick, ranks them by how likely people are to play them; our Stockfish checks every move, and an order with any move giving away more than 0.50 is left out. Open puts an order on the board.

Deep analysis

Research > Deep analysis searches the position on the board for 1, 5, 15 or 30 minutes on an engine of its own, so the board never stalls, and shows our Stockfish's three best lines as they deepen. Stop ends it early and keeps what it found; Put it on the board adds a line to your game.

GM analysis

Research > GM analysis: grandmasters' games, the great games of Serious business and the greats' famous games, each audited move by move by our Stockfish (depth 22, again at 26 where it matters; depth 12 while the deeper audit is still being worked on), with notes only on the deciding moves, each naming its motifs, and the fantasy line: the playable move that was there and nobody found. Walk into it puts that line on the board so you can play through it with the nav buttons; Show me plays it out. Worked out ahead, so nothing waits, and free to read.

Crawl reports

Research > Crawl reports: what hundreds of thousands of real games say, checked with our own Stockfish at depth 25. Pick who played (online players, strong online players, grandmasters, the classics, other over-the-board games) and see, by rating band, what each phase costs, how games end, upsets, how often an advantage is converted and how often the human-like engine guesses the move; the openings: the ones that suit you (where your moves after book are most like those of players 100-200 points above you), every ECO code with the centipawns a move gives away after book, smallest first, and the opening tree by band for the position on the board, cheapest move first; and search any named player's style.

Opponent's pieces

Pick one of your databases and a player in it, or a player from the crawl's list (their games in your databases, or else the ones that come with ChessClerk, their newest and their best-known): their families as White and as Black (each with its antidote first), the pawn structures they reach, and their tree, move by move, with what each move gives away against our Stockfish. Prepare for them as tomorrow's opponent: Between rounds (fast: their newest 75 games) or The night before (full: every game), recent games counting the most; the sheet shows what a move of theirs gives away by phase, where it jumps, the motifs behind their mistakes, and each of their mistakes (most often times biggest first) with the move that punishes it, ready to put on the board. Play them opens a game against them at their strength or yours, or 15 cp better. Only their chess: no results, no win rates, no ratings, no profile. Titled players' families lists the players the titled-player crawl brought into your databases, with their favorite families. Nothing here fetches anyone's games from the internet by name: it reads your databases and the games that come with ChessClerk.

Players

Research > Players: every named over-the-board player with 50 or more games in the crawl, their names put together. A player's page shows their games and score, their style beside the average player's, their games in yours and your databases, and Read about them (Wikipedia). The eras: each decade's average gap, Morphy's time to today, and what it stands for on today's scale; it measures the moves, not the players.

Traps you can set

Research > Traps you can set looks at your sound moves in the position on the board (within 0.50 of the best) and the replies the human-like engine expects at the rating you pick. A trap is a reply it gives a 15% chance or more that loses 1.50 or more. Show me plays your move and the reply.

Shallow vs deep

Research > Shallow vs deep rates every move at depth 8 (a glance) and depth 22 (a deep look). Deep finds look bad at a glance but are among the best three deep: novelties worth a look. Shallow traps look fine at a glance but lose 0.60 or more deep: where a quick player goes wrong.

Board tools

Research > Board tools lists every button by the board: the nav buttons (first, back, Play, 960, forward, last, undo, redo, flip, reset, draw, resign) and every tool behind Tools (Explain each move, highlights, engine move, threats, hint, worst piece, pawns, Puzzle and the puzzle type dropdown, solve, endgame, weak and strong squares, clear, diagram, photo, plunder, explorer, lines, eval, shield and the rest), each with what it does and a button that presses it.

Greats in your position

Research > Greats in your position (or Greats on the tool row) finds, for the position on the board, the closest positions by pawn structure and material in the greats' games and the titled players' games: who had this structure most (the greats first: "Karpov had this structure 14 times") and, under The plan, what they did with it (piece moves, pawn breaks, where the king went, how those games went). Then the closest games, closest and strongest first: Open puts one on the board at that position, then Watch it plays it with notes and Take over lets you play on from there against the engine; Next shows the next ten moves. Both sides' plans adds up each side's plan across the closest games. Search the board searches again after the board has changed; nothing is shown during a game being played.

Know your opponent

Research > Know your opponent: type a player's name and press Look them up (or Enter). It reads only their over-the-board tournament games that come with ChessClerk (The Week in Chess, the masters' files and the famous sets the crawl read), never an online username. When the name could be several players, each is a button: pick one. Their page shows As White and As Black: their results (games, won, drawn, lost, score) and their repertoire as a tree from the start, each move with how often they played it and how those games ended for them; click a move to go deeper, ◂ Back to step back, Put it on the board to play the line. Under it, Where the crawl graded their games: the motifs on their side in the crawl's examples and how often their move there gave away a pawn or more, their costliest positions (Put it on the board, their move next), and from Research > GM analysis's audits their average gap in each phase and their costliest moves (Open the game). It grows as the crawl reads more games.

What to play against them

Research > Know your opponent, on a player's page: What to play against them. Pick You play White or You play Black. For each of your repertoire's moves their over-the-board games reached (with no repertoire for that color, the main options played against them at each branch), it shows how many of their games got there and what share of their games with that color that is, their score there, and their average cp gap in those games where Research > GM analysis audited them. Where they score worst lists the lines (3 games or more) where they score lowest; click one to put it on the board. The mix: at each branch where you have two or more choices, a suggested share for each (more often where they score worse, never always one, so they can't prepare for one line), and a box for your own share, kept on this computer. It is only a suggestion; nothing holds you to it.

Under pressure (an opponent)

Research > Opponent's pieces, on a player's page under the prep box: from the moves of theirs the prep measured, their median gap and share of blunders as usual, in the 5 moves after a blunder of theirs, short of time (in their games that kept clock times) and when clearly winning (3 pawns or more up in material). Measure more with Between rounds or The night before, then Read again.

Chat

Full screen

Panel ▾ in the chat's title bar, then Full screen: the open panel (the tab you have open, or the chat) fills the screen. Escape brings it back.

Read it to the clerk

Panel ▾, then Read it to the clerk: the open panel's words go into the typing box with its name at the top. Add your question and press Send; nothing is sent before that.

Outside sites

Panel ▾ lists YouTube, Claude, GPT, 365Chess and Lichess. Each opens in a new browser tab; YouTube, Claude and GPT start with the opening on the board (or its position when it has no name); Lichess opens its front page (nothing of Lichess plays or signs in inside ChessClerk; your own Lichess games come in through Research > Databases).

Chat font and size

Panel ▾ > Font and size, in the chat's title bar, picks the chat's font (As designed, Serif, Sans or Mono) and size (Smaller to Largest), kept on this computer. Help > Easy for everyone makes the whole column bigger as well.

Drop a text file on the chat

Drag a text file (.txt, .pgn, .md, .fen and the like, up to 200 KB) onto the chat and its words go into the typing box, ready to read over and send. Nothing is sent until you press Send.

Lines

Lines pulled out of games

My study > Lines is a shelf of lines pulled out of games. Step to where the line should end and press Pull out the line on the board: it keeps the moves from the start of the game (or from the move number you type) to the move shown, with the game it came from. Write a note on any line; it is kept as you type. Send to library saves it as a library line (named after your note); Send to repertoire makes your moves in it your answers. Click a move to put the line on the board. A double-click on Lines in the rail pulls the line on the board from the start.

ChessClerk · For creators · Terms · Privacy · Licenses · How the clerk works