ChessGrader / Review Lichess games
Game review for Lichess players: grades, not just blunders.
Lichess already gives you free engine analysis, and it's genuinely good. ChessGrader adds the layer it deliberately leaves out: a full grade for every move, an estimated game rating, and drills for your mistakes.
First, the honest part
Lichess's server analysis is free and strong — Stockfish searching millions of nodes per move on their hardware, no paid tier, no ads, funded by donations. It is rate-limited rather than unlimited: their features page lists 40 requested game analyses per day, and the source caps it at 200 per week with one running at a time. In practice most players never touch that. Their in-browser analysis board genuinely is unlimited. If all you want is an eval graph and your blunders flagged, Lichess already does that well and we're not going to pretend otherwise. (We compared the big two properly in Lichess vs Chess.com analysis.)
So why point ChessGrader at your Lichess games at all? Because Lichess stops at three labels.
What ChessGrader adds on top
- The full grade ladder. Lichess marks Inaccuracy, Mistake, and Blunder — only what went wrong. ChessGrader grades every move: Brilliant, Great, Forced, Best, Excellent, Good, Book, Inaccuracy, Miss, Mistake, Blunder. You find out what you did right, and the thresholds and safeguards are published.
- Brilliant and Great detection. Sound sacrifices and only-moves get flagged — and only after a 600,000-node verification search with the engine's two best lines, so the label means something. Here's what actually triggers a Brilliant.
- Estimated game rating. Anchored to your listed rating, then adjusted by how the game's accuracy compares with what is expected at that level. Lichess does not offer one. Treat it as a vibe, not a measurement.
- Mistake drills. Replay each error until you'd find the right move at the board, instead of nodding at the graph and queuing the next blitz game.
- Plain-language explanations of what each mistake gave away, not just a centipawn number.
ChessGrader vs Lichess analysis
| ChessGrader | Lichess server analysis | |
|---|---|---|
| Price / limits | Free, unlimited, no account | Free; 40 requested analyses/day, 200/week, account required |
| Move labels | Full ladder: Brilliant to Blunder (11 labels) | Inaccuracy, Mistake, Blunder only |
| Positive-move recognition | Yes — Brilliant, Great, Best, Excellent | No |
| Accuracy + ACPL | Yes (CAPS2-inspired accuracy approximation) | Yes |
| Estimated game rating | Yes | No |
| Mistake drills | Yes | No (separate puzzle features) |
| Engine search | 100k nodes/move, key moments verified at 600k | ~2M+ nodes/move on their servers |
| Where it runs | Your browser — games never uploaded | Lichess servers |
Note the engine row: Lichess's server search is deeper than ChessGrader's first pass. That's a real difference and we won't spin it. ChessGrader's fixed 100,000-node budget exists so every device produces identical grades, and the moments that decide labels get re-searched at 600,000 nodes. For grading purposes the two agree on the overwhelming majority of moves — the full reasoning is on the methodology page.
How it works
Type your Lichess username. ChessGrader fetches your games from the Lichess public API — no login, no password, no account connection — and analyzes them with Stockfish 17.1 running as WebAssembly in your browser. Your games and analysis are never uploaded — the only thing published is a one-line summary on the public recently analysed rail, which you can switch off. Accuracy is calibrated against public Chess.com reviewed-game benchmarks rather than copied from Lichess, so compare it within ChessGrader. You can also paste any PGN directly.
Local Stockfish 17.1, a rating-aware expected-points model, and published grade boundaries. ChessGrader is not trying to out-engine Lichess — it is finishing the report Lichess starts.
Frequently asked questions
Isn't Lichess analysis already free and unlimited?
Free yes, unlimited not quite. Lichess gives every signed-in user server-side Stockfish analysis with an eval graph, accuracy, and centipawn loss, rate-limited to 40 requested game analyses a day and 200 a week — limits most players never reach. It is genuinely good. ChessGrader adds what Lichess leaves out: 11 labels including Brilliant, Great, Forced, Best, and Miss, an estimated game rating, mistake drills, and plain-language explanations — with no account at all.
What grades does ChessGrader give that Lichess does not?
Lichess labels only Inaccuracy, Mistake, and Blunder. ChessGrader uses 11 labels: Brilliant, Great, Forced, Best, Excellent, Good, Book, Inaccuracy, Miss, Mistake, and Blunder, with public expected-points boundaries plus chess-specific safeguards.
Is it safe to type my Lichess username?
Yes. ChessGrader reads your games through the Lichess public API — the same games anyone can already see on your profile. There is no login, no password, and no account connection, and the analysis itself runs entirely in your browser.
Will ChessGrader's accuracy match the number Lichess shows?
Not necessarily. ChessGrader uses a CAPS2-inspired approximation calibrated against public Chess.com reviewed-game benchmarks, while Lichess uses its own published method. The engines also search different node counts, so compare accuracy trends within one tool rather than expecting matching percentages.
Is ChessGrader's engine weaker than Lichess server analysis?
Lichess's servers search more nodes per move than ChessGrader's 100,000-node first pass, so on raw depth, yes. ChessGrader compensates where it matters: potential blunders, brilliancies, and big swings are re-searched at 600,000 nodes with the top two lines before any major label is awarded. For move grading the two agree on the vast majority of moves.
What is the estimated game rating and should I trust it?
When your listed rating is available, ChessGrader anchors the estimate there and adjusts it by the game's accuracy relative to what is expected at that level. A rating-less PGN falls back to a community ACPL fit. Treat either as a vibe, not a measurement—one game and one time control are noisy.
Can I review Lichess games from other players or from years ago?
Yes. Any standard-chess game available through the Lichess public API can be reviewed, however old, and you can look up any public username. Variants like Chess960, Atomic, and Crazyhouse are not supported.