ChessGrader / Guides / Brilliant moves explained
What is a brilliant move in chess?
A brilliant move (!!) is the engine's best move that is also a genuine sacrifice — you give up material, the engine confirms the position holds anyway, and you weren't already completely winning. Best move alone isn't enough; the sacrifice is the point.
That definition is why brilliants feel special: they mark the moments you deliberately gave something up and were right. Here's exactly what triggers the label, why different sites disagree about the same move, and why you shouldn't try to farm them.
The ingredients
For ChessGrader to award a Brilliant, a move has to clear four bars at once:
- It's the engine's best move. A spectacular queen sac that loses isn't brilliant, it's a blunder with confidence.
- It's a genuine sacrifice. Measured by static exchange evaluation: the material you put en prise, counting all captures and recaptures on the square, comes out negative. Trading a knight for three pawns and a fork doesn't count — you have to actually be down material by the raw arithmetic.
- You weren't already totally winning. Sacrificing your queen when any move mates in 3 is showboating, not brilliance. The label is reserved for positions where the sacrifice carried real risk.
- It survives deep verification. The position must still be sound after the sacrifice, confirmed by a deeper search — not just the quick first pass.
Brilliant vs Great vs Best
These three labels get mixed up constantly. Side by side:
| Brilliant (!!) | Great (!) | Best (★) | |
|---|---|---|---|
| Engine's top move? | Yes | Yes | Yes (or ties the second line within 10cp) |
| Extra requirement | Genuine material sacrifice (static exchange < 0) | The only good move — second line ≥18% worse — or the precise punish of a game-turning error | None |
| Position constraint | Not already totally winning; still sound after | — | — |
| Deep verification required? | Yes, 600,000 nodes | Yes, 600,000 nodes | No |
| How often you'll see it | Rarely — many games have none | Occasionally | Many per game |
In short: Best is accuracy, Great is precision under pressure, Brilliant is precision plus courage. The full grade ladder is on the move classifications page.
Why sites disagree on the same move
Paste the same game into Chess.com and ChessGrader and you may get a brilliant on one and not the other. That's expected. Chess.com's Brilliant detection is proprietary and has changed over time — early versions were famously generous, and the criteria have been tightened more than once since. Nobody outside the company can reproduce it exactly.
ChessGrader took the opposite approach: the rules above are published in full on the methodology page, so when you get (or don't get) a brilliant, you can see exactly why. Both approaches are defensible. What isn't defensible is treating either site's label as objective truth — "brilliant" is a judgment call that each tool encodes differently, on top of the same engine.
Depth matters too. A sacrifice that looks sound at a shallow search can fall apart at a deeper one, and vice versa. Different node budgets alone can flip a verdict.
How ChessGrader verifies a brilliant
ChessGrader's first pass grades every move at a fixed 100,000 nodes. That's enough to grade ordinary moves, but not enough to stake a !! on. So candidate brilliancies — along with potential blunders and other critical moments — are re-searched at 600,000 nodes with MultiPV 2, meaning the engine works out its two best lines in full. Brilliant and Great can only be awarded by this verification pass, never by the quick one. If the sacrifice doesn't hold up under six times the search effort, the label is withheld.
Why brilliants are rare — and why that's the point
Most games contain zero brilliant moves, including most games by grandmasters. Sound sacrifices in genuinely unclear positions are just rare events in chess. A label you get every game is a participation trophy; a label you get a few times a year actually tells you something. If a site hands you brilliants constantly, it has loosened the definition until the word means "nice move."
How to get more of them
You don't — not directly. Hunting for sacrifices produces unsound sacrifices, which are graded as blunders, because they are blunders. Brilliant moves are a byproduct of two skills: calculating deeply enough to see that a sacrifice works, and being comfortable enough in sharp positions to enter them. Train tactics, play open positions, and review your games to find the sacrifices you didn't see. The brilliants show up on their own, eventually, and mean more when they do.
Want the full picture first? Start with what a free game review includes.
Frequently asked questions
Is a queen sacrifice always a brilliant move?
No. A queen sacrifice is only brilliant if it is also the engine's best move, the position remains sound afterwards, and you were not already completely winning. An unsound queen sacrifice is graded as a mistake or blunder, and a flashy sacrifice in an already-won position is just a best move.
Why did chess.com give me a brilliant but ChessGrader did not, or vice versa?
The two sites use different definitions. Chess.com's Brilliant criteria are proprietary and have changed over time, while ChessGrader publishes its exact rules: engine's best move, genuine sacrifice by static exchange, not already winning, and verified sound at 600,000 nodes. The same move can pass one definition and fail the other, especially near the boundaries.
Do engines play brilliant moves?
By the mechanical definition, yes — engines regularly play sound sacrifices that would earn the label. But the label exists for humans: it marks moves that required courage and calculation beyond what the position forced. For an engine there is no risk, so 'brilliant' is really a human-facing annotation, not an engine concept.
What is the difference between a brilliant move and a great move?
Both are engine-best moves verified by a deep search. A brilliant additionally requires a genuine material sacrifice. A great move requires no sacrifice but must be the only good move — the second-best option at least 18% worse in win probability — or the precise move that punished an opponent error that turned the game.
How rare are brilliant moves?
Genuinely rare. Most games contain none, including games between strong players, because sound sacrifices in unclear positions are uncommon events. If you see brilliants in most of your games, the tool grading them is using a loose definition.
Can a losing move ever be brilliant?
No. A brilliant must be the engine's best move and the position must remain sound after the sacrifice. A sacrifice that objectively loses is graded by how much win probability it threw away, no matter how spectacular it looks.