White to move
Stockfish
5:00
You
5:00

How Coaching Mode Works

Coaching mode runs two chess engines at once. The first is your opponent — Stockfish set to whatever difficulty you choose, from a beginner-friendly bot up to full strength. The second is your coach: a separate Stockfish instance pinned to Skill Level 20, the maximum, which never plays a move and exists only to judge yours.

What Happens After Each Of Your Moves

The moment you release a piece, the coach analyses the position twice: once as it stood before your move to find the strongest continuation that was available to you, and once after your move to see where you actually landed. The difference between those two evaluations, measured in centipawns (hundredths of a pawn), is what your move cost you.

Anything from an inaccuracy upwards stops the game rather than just labelling it: the opponent waits while you decide whether to take the move back and try again, or play on from the damage. See Take It Back, Or Play On? below.

Missed Tactics

Losing evaluation is only half the story. The coach also inspects the move you could have played and works out what kind of chance it was: a forced mate, a piece hanging on an undefended square, a fork, a winning capture, or a promotion. When that opportunity was clearly better than what you chose, a cyan arrow is drawn on the board from the square the piece was on to the square it should have gone to. That arrow is the most useful thing on the page — the tactics you keep missing are the tactics you have not yet trained your eye to see.

Take It Back, Or Play On?

When a move costs you something, the opponent does not reply straight away. The board freezes on the error — red square, arrows and all — and you are asked whether you want the move back. Two arrows are drawn: the move you should have played, and the reply Stockfish had ready against the one you chose. Choose Take It Back and the move is undone so you can find something better; the error is struck from your record and the replacement move is graded in its place. Choose Play On and Stockfish answers as normal.

Both answers are useful. Taking it back turns the game into a tactics drill: you know something was there, and now you have to find it. Playing on is the harder discipline — defending a position you have just damaged is the skill that decides most games between club players, and you cannot practise it if you undo every error. A reasonable habit is to take back tactical misses and play on through positional ones.

By default you are offered the move back on inaccuracies as well as mistakes — every move that cost you something. If that interrupts too often, narrow it to mistakes only, or switch it off entirely, under Offer Take Back in the coach settings.

Reading The Accuracy Score

Accuracy summarises the whole game in a single number. It starts at 100 and falls as your average centipawn loss rises — a game where you rarely drop more than a fraction of a pawn per move will sit in the nineties, while a game with two or three serious mistakes lands well below that. It is calculated only on your moves, so beating a weak opponent does not inflate it.

Getting The Most From It

  1. Set the opponent near your own level. Difficulty 3–8 gives you positions you can actually play, which produces far more useful coaching than being crushed in fifteen moves. If you do not know your level yet, estimate your chess ELO first.
  2. Read the reason, not just the colour. The coach explains why a move failed — what was left hanging, what got forked. That sentence is the lesson.
  3. Give the endgame a moment. The coach always searches to depth 18, deep enough for the long forcing lines that decide thin positions — but on a slow device that verdict can take a few seconds to land. It is worth waiting for.
  4. Look for the pattern across games. If most of your mistakes are pieces left undefended, that is a scanning habit to fix, not a chess mystery.

Do You Still Need a Human Chess Coach?

A human coach builds you an opening repertoire, sets long-term goals and holds you accountable — an engine does none of that. What an AI chess coach does better is the part most improving players actually need: instant, objective feedback on every single move, in every game you play. Below about 1800, the fastest rating gains come from eliminating blunders and spotting simple tactics, and that is precisely what a chess blunder checker trains — every blunder flagged the moment you make it, while the position is still on the board. Many players use both: engine coaching for daily play, a human coach for occasional deeper lessons.

Who Is Coaching Mode For?

Anyone from a complete beginner to a club player around 2000. Beginners should set the opponent to difficulty 1–3 and treat every red square as the lesson; intermediate players get the most from the missed-tactic arrows; stronger players can face full strength and use the accuracy score to measure consistency. It is a chess coach app that runs in your browser — nothing to install — so it is there for every game you play, not just the ones you remember to review.

Everything runs in your browser — no upload, and no waiting for a server to analyse the game afterwards. The feedback arrives while the position is still fresh in your head, which is exactly when it is worth the most.

Already played the game elsewhere? Analyze a finished game instead: enter your Chess.com or Lichess username, or paste the PGN, and the same full-strength engine grades every move of it, scores both players for accuracy and turns your mistakes into practice positions.

Prefer an uninterrupted game? Play Stockfish online without the coach, at any difficulty.

Coaching FAQ

Is the coach really Stockfish at full strength?

Yes. Your opponent uses whatever difficulty you set on the slider, but the coach is a separate engine instance that always runs at Skill Level 20 with no strength limit applied. It only ever analyses; it never plays a move for either side. That is why you can beat a difficulty-2 opponent and still be told your winning move was a mistake — the two engines are not the same engine.

What is a centipawn, and why 150 for a mistake?

A centipawn is one hundredth of a pawn, the unit engines use to score a position. A move that drops 150 centipawns has handed over a pawn and a half — in practice an exchange, a piece for compensation, or a position that has gone from comfortable to difficult. Below that, at 60 centipawns, a move is called an inaccuracy: worth knowing about, but not usually what decides the game.

Why is there no separate "blunder" grade?

Because the distinction does not change what you should do about it. A 300-centipawn error and a 160-centipawn error are both moves you would want back, and both get the same treatment here: the red square, both arrows, and the offer to retry. Splitting them into two grades made the panel busier without telling you anything you could act on. The centipawn figure is still shown on every verdict, so the severity is never hidden — a move that drops eight pawns says so.

How deep does the coach search?

Depth 18, always, with nothing to set. A shallower search is quicker, but it misses tactics and then grades the move as though they were never there — which is worse than the couple of seconds it saves. There is one exception: the Hint and Best Move buttons cap the search at depth 14, because those answer while you are sitting waiting for them.

Why does the coach sometimes disagree with itself between moves?

Search depth is finite even at 18. A tactic that only appears a move or two beyond the horizon can shift an evaluation once the position is searched again from a different starting point, so a move called fine at the time may look worse in hindsight. Sharp tactical positions and endgames are where this shows up most.

Does the analysis slow the game down?

The coach searches while your opponent is thinking and while you are choosing your next move, so in practice you rarely wait for it. On a slow device a verdict can arrive a few seconds after you move; the board stays fully playable in the meantime and the feedback is filled in when it is ready.

Is my game sent anywhere?

No. Both engines are WebAssembly builds of Stockfish 18 running inside your browser tab. Your moves are never uploaded, and closing the tab discards everything.

What is the difference between the Hint button and the coach?

Hint looks forward: it asks the engine what to play in the position in front of you, before you commit. The coach looks backward: it grades the move you already played and shows you what was available instead. Hints are for when you are stuck; coaching is for learning why you were stuck.

Will this make me better at chess?

It will if you use the explanations rather than just the colours. Most improvement below club level comes from removing blunders, not from finding brilliancies, and the fastest way to remove them is to see each one immediately, in the position where it happened, with the reason attached. Play a few games, note which category keeps repeating — hanging pieces, missed forks, walking into checks — and train that one pattern with chess puzzles.