What the Chess Analysis Board calculates
This calculator searches a legal chess position with Stockfish 19 Lite in your browser and returns candidate moves, principal variations, evaluation, and modelled win-draw-loss shares. Set up the board, paste FEN or PGN, import a public Lichess game, navigate every move, load a curated opening, or review a detected board image. Depth or time limits, one to five candidate lines, hash, node count, speed, search time, SAN, UCI and board arrows make each engine answer auditable.
For opening positions, the tool can request live Masters statistics. For seven-piece endgames, it can request exact Syzygy results. Those network-backed facts are labelled separately from finite local-engine estimates.
How to analyze a chess position
- Set the position by making legal moves, using the editor, loading FEN, importing PGN or a public Lichess URL, choosing a curated opening, or scanning a clear board image.
- Confirm the side to move. For a scan or custom position, also review every piece, castling right, and en-passant field.
- Choose Quick, Standard, or Deep. Advanced settings support a depth or time limit, one to five candidate lines, and 16-256 MB hash.
- Select Analyze position, compare candidate scores, WDL and material evidence, then replay or save any returned line.
- Use live position evidence for Masters statistics when a server token is configured and exact Syzygy results for seven-piece endgames.
- For PGN, navigate to any move or run the disclosed 250 ms-per-position full-game review. Export annotated PGN or the saved-variation report when finished.
Engine, rules, and score convention
Stockfish 19 Lite supplies the UCI search. Chess.js 1.4.0 supplies legal move generation, FEN and PGN handling, check, checkmate, stalemate, insufficient-material checks, and move history. CalculatorGeek independently counts occurrences of the current position for threefold repetition and reads the FEN halfmove clock for the fifty-move rule. The interface binds every result to the searched FEN; editing the board, changing a search setting, or navigating the game clears stale analysis.
| Output | Meaning |
|---|---|
| +0.80 | White is favored by about 0.8 pawns in the searched position. |
| -1.20 | Black is favored by about 1.2 pawns. |
| Mate in N | The engine found a forced mate at the current search horizon. |
| WDL | Stockfish's modelled outcome distribution, normalized to White win, draw, and Black win; it is not a personal win probability. |
| MultiPV 1-5 | Stockfish retains the requested number of candidate lines. More lines can reduce effort on line one. |
| Depth or time | The user chooses a fixed depth from 8 to 24 or a time limit from 0.25 to 10 seconds. |
The tool uses a fixed White-perspective score convention, even when Black is to move. Read how to read a chess engine evaluation before comparing positions.
Choose the right chess position input
FEN stores one position, including the side to move and rule-state fields. PGN stores a complete game or move sequence and can be replayed move by move. The board editor is best when you want to build or correct a position visually.
| What you have | Best input | What to verify |
|---|---|---|
| A position on screen | Move pieces on the board | Piece placement and side to move |
| A FEN code | Load position | Castling and en-passant fields |
| A complete game | PGN or public Lichess URL | Select the exact move to analyze |
| A screenshot or photo | Scan board | Every piece, orientation, turn and rule state |
The editor checks king count and adjacency, pawn ranks and counts, total pieces, castling consistency, en-passant state, and whether the side that just moved was left in check. A board image contains neither move history nor reliable castling and en-passant rights, so image recognition deliberately requires human review. The detector runs locally in the browser; an editable FEN and visual correction palette remain available when confidence is low. For more detail, see FEN vs PGN for chess analysis and the chessboard image-to-FEN guide.
Worked example: 1.d4
From the initial position, moving the white pawn from d2 to d4 produces rnbqkbnr/pppppppp/8/8/3P4/8/PPP1PPPP/RNBQKBNR b KQkq d3 0 1. The interface must now say Black to move and Custom position, clear the starting-position score, and analyze only that FEN. A best move such as Nf6 can change with engine build, depth, hash state, and MultiPV settings; the auditable facts are the FEN, engine version, depth, score convention, and returned line.
Accuracy, privacy, and limitations
- Legal moves are checked by Chess.js 1.4.0 and core move generation passes standard perft reference positions through depth three.
- Checkmate, stalemate, insufficient material, current-position repetition, and the fifty-move clock are handled before engine search.
- Full-game review uses the same 250 ms local search for every position. Displayed-evaluation loss is classified as near-best at ≤0.20, good at ≤0.50, inaccuracy at ≤1.00, mistake at ≤2.00, and blunder above 2.00. It is not an official platform accuracy score.
- Search strength depends on the selected limit, device speed, MultiPV, hash, and Lite build; it is not equivalent to unlimited cloud analysis.
- WDL is an engine model, not a guarantee or rating-specific personal forecast.
- Image recognition can confuse pieces or orientation. Review the reconstructed board before analysis; a photograph cannot reveal move history.
- Stockfish and image recognition run locally after assets load. Live Masters, tablebase, and Lichess-import requests send the current FEN or public game ID through CalculatorGeek to Lichess services.
- Use engine help for study and post-game review, not during a live rated game.
What this analysis workflow adds
The reviewed competitor set makes FEN/PGN import, engine lines, board editing, opening data, game review, or screenshot recognition available in different combinations. CalculatorGeek brings them into one auditable workflow: validated position editing, move-by-move PGN navigation, direct public Lichess import, review-gated local image detection, curated openings plus live Masters evidence, exact seven-piece tablebase results, keyboard board navigation, selectable MultiPV playback, saved variations, WDL, depth-or-time controls, hash controls, stop control, material evidence, full-game turning points, annotated PGN and report export, shareable FEN, and local play.
The tool does not label speculative tactical themes as facts, invent a “brilliant” move, or present its finite review as official accuracy. The explanation names calculated evidence and the opponent's best returned response.
Sources and version record
- Stockfish: official engine project and current release information
- Stockfish WDL model documentation
- Stockfish.js: WebAssembly/browser port and license
- Chess.js releases and rules fixes
- Lichess opening explorer
- Lichess API and tablebase specification
- FIDE Laws of Chess
Calculation implementation reviewed 2026-09-25. Engine: Stockfish 19 Lite single-thread WebAssembly. Rules layer: Chess.js 1.4.0. One schema owner should emit WebApplication, WebPage, and BreadcrumbList; no rating or FAQ rich-result claims are included.
Frequently asked questions
Why did the score disappear after I moved a piece?
The previous score belonged to a different FEN. Clearing it prevents a stale engine result from being presented as analysis of the new position.
Is a positive score always good for the player to move?
No. CalculatorGeek displays the score from White's perspective. Positive favors White; negative favors Black.
Can a board photo identify whose turn it is?
Not reliably. A still image also cannot prove castling or en-passant rights, so those fields require user review.
Does deeper analysis guarantee the perfect move?
No. More depth usually improves the search, but finite engine analysis remains dependent on the engine build, settings, hardware, and position.