How ajedrez.dev works

A session is a short set of exercises, the ones due for review first. Miss one and you'll see it again soon, in another session.

Everything you train goes through three stages:

• New: not tried yet.
• Learning: each time you solve it, it takes longer to come back; miss it and you'll see it more often.
• Fixed: once you've solved it twice, two weeks apart: the pattern has stuck.

Solving it means no wrong moves and no hints.

Pick a side and play. You'll face an opponent who answers the way players at your rating usually do. Every move you make becomes part of your repertoire.

If your move is dubious (the engine prefers another, masters almost never play it, or it isn't the one in your repertoire), the game stops and shows you why. You can keep your move or change it: the one you choose stays in your repertoire.

Note that ajedrez.dev is designed for you to build a fixed repertoire, with a single prepared move in each position. We believe this is the best way to take openings seriously.

If the engine accepts your move, it lights up ; if it was also the only good one, it gets a ; if it's the move masters play most here (or, without masters' games, the engine's first choice), a . Turn the marks off in Board settings.

While the Book or Stockfish is open, you're exploring: nothing you play is saved. Close them and the game goes on where it was.

When the position is no longer theory, because hardly any master games reach it, the game ends.

A position comes to Train when you change your move after a stop, or save it with the bookmark. It comes back until it sticks.

After a right answer the line may go on: your opponent replies, and where only one move is good, you have to find it too. It stops where several are good. A miss anywhere counts for the position.

Answering a position that isn't due is practice and changes nothing.

Every line you've played, as a tree on the board: play a move or pick one from the list. Green replies have your answer beside them, red ones don't yet; the list says how often masters play each.

Each session gives you a real position from Lichess games, for one idea. You only see whose turn it is: whether to win or hold is part of it.

It ends when you reach a position that's easy to carry on from. When you're winning, for example: once you promote or leave your opponent without material, in good shape. When you're holding a draw: once you've shown you can hold it, or you reach a position where players at your rating rarely go wrong. Play on lets you play it to the end.

A move that throws the result away is taken back, saying why: try another from the same position. A session is clean without mistakes; a hint, or the tablebase before the end, counts as one.

When your move was the only one that kept the result, a marks it. When you're winning, a marks a move that mates soonest with best play: other moves may keep the win without getting you closer to it. Turn the marks off in Board settings.

A move that threw the result away, when most players at your rating make it too, becomes a position of its own: a later session gives it to you from the other side, against the tablebase's best defence, so you see why it was a mistake. The same mistake is saved once.

Learn gives you the next idea you haven't played; it opens when you solve the one before. The ones you've played come back in Train. An idea can have several positions (as White or Black, to win or to hold), each with its own schedule: after 3 clean sessions in a row a position is learned and comes back after a week, three weeks, two months, half a year; a mistake brings it back at once. Each time, a new setup. In the catalog, each idea's line says where you stand on its positions; opened, it shows them all.

Everything follows your rating (your Lichess account's, or the one in the settings): the order of the ideas, how often each comes up and is missed, and how the defender plays.

Tablebase, on the right: it gives the result, with perfect play, of any position with 7 pieces or fewer. It also shows how players at your rating play from there (how often they choose each move).

Hint, on the left: a move that keeps the result.

Stockfish replays your games in your browser and turns your mistakes into puzzles. Pick from your last games on Lichess or Chess.com (the accounts in the settings), or paste PGN. A mistake is a move that:

• gives away 10% or more of your winning chances, or
• misses a mate in one or two.

It becomes a puzzle when it has a clear lesson:

• one best move,
• a position that wasn't already lost,
• and a line to play out.

A game without one still gives a puzzle from its biggest mistake, where any move nearly as good as the best is right. The same idea missed twice in a few moves is one puzzle. A pasted PGN can also be taken as it is: its own moves become the solution.

You can also set up a puzzle by hand (New puzzle, in My puzzles) or start with the starter puzzles. Report a bad puzzle from its card: it's archived.

We also find opening and endgame exercises in your games.

Each analysed game's row shows what it gave, by icon. In Train, a position from one of your games links to the game at that move.

A hint shows the piece to move, then where it goes (you can turn it off in the settings). It counts as a miss.

So does opening the engine before the puzzle is solved or missed. After that, look all you like.

If your piece lands next to the right square, you're asked whether it slipped. Say so and it doesn't count; be honest.

Under the board you'll find these buttons:

• Hint: the first click shows you which piece to move; the second, the move. Using hints counts as a miss.
• Back and Forward, to go over the last moves (the arrow keys work too).
• Book, Stockfish and Tablebase, to analyse the position; threat vision, which marks the pieces with more attackers than defenders and, with a red arrow, your opponent's threat, when they have a move much better than the rest; and Déjà vu, the positions of your repertoire that look like this one, what sets them apart, and the move you prepared in each (free: it never counts as a hint).
• ⋯: more options, like flipping the board.