Most players meet the brilliant move as a !! icon in a game review, then spend weeks chasing another. Here is what the label measures, why Chess.com, Lichess and Chess DNA award it so differently, and how to find more without wrecking your results. Our free chess game review marks brilliant moves on any pasted PGN or username, no login.
A brilliant move, written !!, is a sacrifice that turns out correct: real material given up, and the engine still rates the move best or nearly best. Chess.com awards it, more generously at lower ratings, plus a separate Great label for only-moves. Lichess computer analysis never awards it. Chess DNA requires at least a minor piece given up after the whole exchange, under 2% winning chance lost, and a position still at least equal afterwards. You get more by learning sacrifice patterns, not by hunting them.
Interactive lesson — The brilliant sacrifice: give the queen, deliver smothered mate: Play it on the board below, and make the key move yourself. White to move, and behind on material: two rooks against a knight. Look at the black king instead. Its own rook and pawns box it in, and your knight on h6 already covers g8. Qg8+. The queen steps onto a square where she can simply be taken, and that is the whole idea. The king cannot take her, because the knight on h6 guards g8. 1...Rxg8. Forced, and fatal. The rook takes the queen and in doing so fills the last square the king had. 2.Nf7#. Smothered mate: the king is walled in by its own pieces and nothing can reach the knight. This is what earns a brilliant label: real material given up, the best move on the board, and a winning position after it.
The !! comes from written annotation, long before engines: ! for a good move, !! for a brilliant one, ? for a mistake, ?? for a blunder, !? for an interesting gamble and ?! for a dubious one. Wikipedia catalogues the full set of chess annotation symbols.
In a book, !! was a human judgement. An engine cannot measure how hard a move was to find, so automated reviews use a proxy: the sacrifice. You give up real material and the engine still rates the move best or close to it. Material down, evaluation intact.
That is also why the label is rare: losing material by accident is far more common than sacrificing it correctly. In Chess DNA's dataset of 6,288 real amateur games from 103 players, analyzed with Stockfish at depth 18, players rated 1200 to 1499 blundered 1.72 times per 100 moves in rapid, and that rate drops only about 30% past 1500. Mistakes barely move, from 4.8 to 4.5 per 100 moves. Club games are full of dropped pieces, and short on correct sacrifices.
The same game can show two brilliant moves on one site and none on another, because each site measures something different.
| Chess.com Game Review | Lichess computer analysis | Chess DNA | |
|---|---|---|---|
| Brilliant label? | Yes, marked !! | No. Only errors are labeled. !! is a manual annotation in Studies | Yes |
| Definition | A good piece sacrifice, best or nearly best, that does not leave you in a bad position, played when you were not already completely winning | None | At least a minor piece given up after the whole exchange sequence, under 2% winning chance lost, position afterwards at least equal |
| Rating-adjusted? | Yes, more generous for lower-rated players | Not applicable | No, one rule at every rating |
| Separate "great" or only-move label? | Yes, Great, for a critical move that is often the only good one | No | No |
Chess.com. The table paraphrases its help center. Great needs no sacrifice at all, and the accuracy model is rating-aware too, as covered in how chess accuracy is calculated.
Lichess. Its computer analysis labels errors only and stays silent about good moves. Our Lichess game review guide covers what its free analysis does well.
Chess DNA. Three conditions, the same at every rating:
The third condition does the real work. Winning chance saturates near zero in a lost position, so without it every desperate hang in a hopeless game would score as brilliant. The full threshold table is in chess move classifications explained, and our free Chess.com Game Review alternative covers second opinions.
The board above shows this position: r2q1r1k/pp4pp/7N/8/2Q5/8/PP3PPP/6K1 w - - 0 1. White, down material, has a queen on c4 and a knight on h6. Black's king on h8 is walled in by its own pawns on g7 and h7 and its rook on f8.
The move is 1.Qg8+! The king cannot take, because the h6 knight covers g8, and Black's queen cannot take, because the f8 rook is in the way. The only legal reply is 1...Rxg8, and 2.Nf7# is mate: the rook now blocks g8, the pawns block g7 and h7, and nothing can capture the knight. This is the classic smothered mate.
Against the three conditions: a whole queen given up, a forced mate in two, and a final position far above equal. White starts down material, so it was not already winning either. This is the textbook case the label exists for.
A brilliant count is not a skill measure: it depends as much on what your opponent allowed as on what you saw. Your blunder rate matters more, and why you keep blundering in chess explains it.
No trick produces them on demand, but you can see more of the sacrifices already there.
Sound sacrifices cluster in a few families: the smothered mate above; the Greek gift, Bxh7+ against a castled king when a knight can follow to g5; deflection and decoy sacrifices that pull a defender away or lure the king onto a mating square; back rank sacrifices; and clearance sacrifices that open a line for another piece. Our guide to recognizing chess patterns shows how to drill them.
Gambits trade material for time and open lines, which is where sacrifices grow. In the Fried Liver Attack, after 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 d5 5.exd5 Nxd5 6.Nxf7, White gives up a knight to drag the king into the open. See the Fried Liver guide and the chess traps index.
Every brilliant sacrifice is a forcing line. Before each move, list checks, then captures, then threats, and follow each until it stops forcing. That habit finds 1.Qg8+, a check a non-forcing search never looks at. Chess calculation training and chess tactics training lay out the routine.
Count the replies. In the smothered mate Black had exactly one. If you cannot see the end of your sacrifice, you are gambling, and the review grades the gamble on its result.
Run recent games through a free game review and look for positions where the engine's best move was a sacrifice you did not play. Those teach more than the brilliant moves you found.
Anderssen vs Kieseritzky, London 1851, "The Immortal Game". Anderssen gave up both rooks, a bishop and his queen, then mated with his minor pieces: 22.Qf6+ Nxf6 23.Be7#. Engines have since found improvements, a reminder that the old !! was a human verdict.
Donald Byrne vs Bobby Fischer, New York 1956, "The Game of the Century". Thirteen-year-old Fischer played 17...Be6!!, allowing 18.Bxb6 to win his queen. His pieces took back more than enough material and hunted White's king, ending 41...Rc2#.
Kasparov vs Topalov, Wijk aan Zee 1999. 24.Rxd4!! gave up a rook to start a king hunt that drove Black's king from a7 to White's first rank. Topalov resigned after 44.Qa7.
Find the brilliant moves in my games free →
A brilliant move is a sacrifice that turns out correct. You give up real material, usually at least a minor piece, and the engine still rates the move best or nearly best. Annotators mark it !!. Reviews cannot measure how hard a move was to find, so they use the sacrifice as the proxy. Each site adds its own conditions, such as not already being completely winning, or staying at least equal afterwards.
Find a piece sacrifice that is the best or nearly best move, in a position you were not already completely winning, that does not leave you worse. That is the shape of the rule Chess.com's help center describes, more generous at lower ratings. These come from tactics: a queen sacrifice into smothered mate, a bishop sacrifice on h7, a decoy onto a mating square. Unsound sacrifices are graded mistakes or blunders instead.
!! is the annotation symbol for a brilliant move: very strong, hard to find, and often a sacrifice. The standard set also has ! for a good move, ? for a mistake, ?? for a blunder, !? for an interesting but risky move and ?! for a dubious one. The symbols predate engines and were the annotator's own opinion. Game review tools borrowed !! for their brilliant label, each with its own engine-based definition.
Because the rule is deliberately rating-adjusted. Chess.com says it is more generous with brilliant moves for lower-rated players, and its Expected Points model reads the same evaluation differently by rating. Lower-rated games also have more loose pieces and exposed kings, so working sacrifices appear more often. The label therefore means less at 800 than at 2000. Chess DNA applies one fixed rule at every rating instead.
Not in its computer analysis. Lichess grades moves with a fixed win chance formula and labels only errors: inaccuracies, mistakes and blunders. No move gets a brilliant or great label. The !! symbol exists on Lichess only as a manual annotation you add to a move in a Study. To see brilliant moves marked on a Lichess game, paste its PGN into a game review tool that awards the label.
On Chess.com they reward different things. Brilliant is for a good piece sacrifice: material given up, and the engine still approves. Great is for a critical move, often the only good one, with no sacrifice required, such as the one defence that holds. Lichess uses neither label. Chess DNA awards brilliant but has no great label, because one number for how much winning chance a move cost is easier to trust than two overlapping scores.
Further reading on Wikipedia: chess annotation symbols and The Game of the Century.