Skip to content

Game review

How ChessOS finds a brilliant move

A brilliant move gives something up that you can see being given up, and it still turns out to be right. Here is the exact rule we use, three classic examples to step through, and what the label can and cannot tell you.

The rule

Four checks, all four have to pass

The definitions are ours, built from the engine's winning chances and the line it gives. Nothing here depends on a second-choice move.

  1. Check 1

    The move is the best one, or very nearly

    A brilliant move is a good move first. If it loses more than one percentage point of win chance compared with the engine's choice, it is graded by what it lost, however pretty it looks. Players rated under 1000 get more room, as chess.com gives them: any move that loses under eight percentage points of win chance can count.

    Loses under 1 percentage point of win chance (under 8 for players rated under 1000)

  2. Check 2

    It gives up real material, and this move is what gives it up

    A piece, or the exchange: two pawns' worth is the smallest sacrifice that counts, measured by what the opponent could win by capturing (every capture and recapture on that square counted), net of anything the move itself took. The piece has to be offered by this move: moved or uncovered into danger, or left where it stands after the opponent has just attacked it, while the move threatens nothing of its own in return. A piece that was already hanging does not count.

    The opponent could win 2 pawns' worth or more, at least 1 net of what the move took

  3. Check 3

    The material is really given up

    If taking the piece is won straight back with one capture, it was a trade in disguise, not a sacrifice (a queen offer excepted; players rated under 1000 are judged more generously here too). The opponent does not have to take it: the classic brilliancy is a piece that cannot safely be taken.

    Not won straight back with one capture

  4. Check 4

    It is a fight, not a victory lap, and it holds

    A move that gives up a piece when you are already crushing the game is not brave (unless the sacrifice forces mate), and a move that leaves you worse did not work. It also has to be a choice: never the only legal move or a reply to check.

    Win chance under 90% before, 40% or more before and after

Try it

Three classic sacrifices

Step through each one. The move to look at is already on the board. These are famous games and positions, not ChessOS reviews; they show the idea the rule is looking for.

Paul Morphy – Duke of Brunswick and Count Isouard · The Opera game, Round ? · 1-0

Move 16

16. Qb8+

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
Download PGN

Morphy against two amateurs, Paris, 1858

A queen given up for a back-rank mate

Black has just one knight guarding the back rank. White's queen goes to b8 with check, the knight has to take, and the rook mates on d8. The queen is gone in order to take away the only defender.

The sacrifice removes the defender.

White to move – Black · Smothered mate, Round ? · 1-0

Before the first move

Step through the game with the arrows or your keyboard.

1.
2.
Download PGN

A study position (the smothered-mate pattern)

A queen offered to smother the king

The black king is boxed in by its own pieces. The queen steps next to it with check, protected by the knight, and the rook has to take. That rook was the last flight square for the king: the knight then gives mate from f7.

The sacrifice drags a defender onto the square that smothers the king.

Honest limits

What the label cannot tell you

  • It counts material. A quiet move that wins by an idea alone, with nothing given up, can be the best move of the game and still not be called brilliant.
  • It trusts the engine's line. A sacrifice that works only because the opponent has to find a very hard defense still passes, if the engine's line holds.
  • It knows the players only by their ratings, and only when the game records them. Players rated under 1000 get the more generous reading; when no rating is known, everyone is judged as an established player. A beginner who stumbles into a sound sacrifice gets the same badge as a master who calculated it.
  • Other tools use other definitions. Two reviews of one game can disagree about a brilliant move without either being wrong.

Comparing tools

What we have checked, and what we have not

We tuned and checked this rule on a published set of 100 chess.com games, each with one move chess.com called brilliant. At the depth our review uses, the rule finds 81 of the 100 and calls about 2 moves in every 1,000 brilliant that chess.com did not (about one wrong badge every six games). We have not run other tools ourselves, so there is no league table here and we will not make one up. If we publish one, it will be done the way it should be: a fixed set of real games anyone can download, the answer key named and its limits stated, how often each tool calls a brilliant move that the key does not, and the date each tool was run.

Think a call is wrong? Send us the game from the feedback link at the bottom of any page, and we will look at it.

See it in a real review

The demo account has 35 analyzed games with every move tagged.