Game analysis
Chess game analysis for coaches
A coach with thirty students cannot replay every game. ChessCore attaches each game to the student, has Stockfish flag the mistakes in the games the coach requests, drafts a plain explanation with AI, and leaves the decision to the coach. Import, engine analysis and AI explanation are separate steps, and you still decide which explanation is ready to share.
Game analysis is one part of the academy record. For how it sits next to enrolment, attendance and fees, see chess academy management software.
Request a demoSix steps
From imported game to reviewed feedback
You start from a game the student actually played. Stockfish analysis for chess academies flags the moves to inspect; you decide what to teach. Six steps carry a game from the board to feedback.
Games come in
Games sync from a linked username (automatic sync starts on the Pro plan and up), or you import a game link. A PGN import also accepts a file of up to 200 games, attached to the student you select. Importing a game does not itself request analysis.
The coach picks games to analyse
The coach requests Stockfish analysis for the games that matter. With the academy’s practice loop on, which is the default, ChessCore also queues analysis overnight for up to two recent games per student that have none.
Stockfish classifies each move
Stockfish runs on ChessCore’s servers and measures how many centipawns each move lost. A loss of 50 or more is an inaccuracy, 100 or more a mistake, 200 or more a blunder. Each flagged move gets the better move, and the game gets an average centipawn loss for the whole game and for the student’s moves alone.
The AI explains, the coach approves
The AI explains the engine’s findings in plain words for the coach. Per-move explanations start as drafts. The coach edits and approves each one before it can be shared.
Approved explanations feed skill progress
When the coach approves an explanation, the skills it covers count toward the student’s skill progress.
Mistakes become practice
The coach can turn the student’s own flagged mistakes into tactics puzzles assigned to that student, or add them to the student’s puzzle review queue.
Step through the game
The move the engine flagged, and what the coach says
Game 1 of the synthetic sample, move by move. The eval bar moves only where the page has a measured figure.
Game 1, 8 September 2026Synthetic sample
Amara vs Mateo S.
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
The pinned knight leaves f6 and opens the diagonal from g5 to d8. The evaluation goes from -0.24 to +8.96, a loss of 920 centipawns, well past the 200 that marks a blunder.
Illustrative AI draft
“On move 5 you moved your knight, and the engine evaluation dropped.”
After Coach Elena's editApprove to share
“On move 5 your knight was pinned to your queen, so moving it let White take your queen. Before you move a piece, check whether it is pinned.”
Import, queue and status
For a PGN file, preview it, import its valid games, then request analysis for the games you want reviewed. A file can hold 200 games, and importing that batch does not itself request analysis.
Each game has its own status: Pending, Analysing, Done or Error. Pending shows “Analysis queued”; Analysing shows “Analysing game”, not a percentage or a completion-time estimate. If a request cannot enter the queue, the message is “could not queue engine analysis”. You can use Request engine analysis again on that game after the problem is resolved. Each analysis worker processes one game at a time; that is a processing limit, not a promise of unlimited queue capacity.
One worked example, from file to feedback
Try the synthetic PGN sample. Choose a student, select Preview games and inspect each result before importing. A whitespace-only file returns “PGN file is empty or contains no game blocks”. For example, an illegal opening move 1. e5 produces “Invalid move in PGN: e5”. Invalid games appear as “Invalid game” with their error, while valid games can still be imported. Correct the file, choose it again and preview it again before importing.
Made-up example academy: Knight's Gate Chess Academy. The sample contains three legal, invented games with Mateo S. as Black; these are teaching fixtures, not customer results or actual online games. In game 1, dated 8 September 2026, 5...Nxe4 moves the knight from f6, opening the bishop's diagonal from g5 to the queen on d8. White replies 6.Bxd8. A local Stockfish 19 search at depth 16 measured the position before the knight move at -0.24 pawns and after it at +8.96 pawns, both from White's perspective. These are reproducible sample measurements, not output captured from a ChessCore customer account.
Here is one move explanation, in the same human-edit format as the AI review example. Illustrative draft: “On move 5 you moved your knight, and the engine evaluation dropped.” Illustrative feedback after Coach Grace's edit: “On move 5 your knight was pinned to your queen, so moving it let White take your queen. Before you move a piece, check whether it is pinned.” The draft and edit are authored examples, not a captured AI response or a real coach approval. Check AI chess move explanations for students against the actual board before you share them.
The same mistake across three games
Chess mistake pattern analysis for students compares more than one position. Compare game 1, 8 September, 5...Nxe4, game 2, 15 September, 6...Nxe4 and game 3, 22 September, 7...Nxe4 in the same download. Each dated synthetic game shows the f6 knight leaving the same pin against the d8 queen. A coach can record that repeated theme in student progress tracking and turn flagged mistakes into homework and student training. Three invented examples demonstrate the review method, not evidence of a real student's progress.
Engine settings
Stockfish runs on ChessCore's servers, not in the student's browser. Each analysed move is searched to depth 16 by default, and the coach can pick a depth from 8 to 22. The engine looks at the two best lines in each position, and analysis covers up to the first 120 half-moves of a game. Moves played from a position that was already lost (300 centipawns or worse for the side to move) are not flagged. ChessCore reports centipawn loss, not an accuracy percentage.
Engine evaluation and AI explanation
Engine-grounded AI chess analysis splits the work. Stockfish is a chess engine. It produces every evaluation and every flagged move. The AI does not calculate chess. It writes the explanation from the engine's output, and player names in the game file are masked before it does. A coach reads both and has the final say. Grounding in the engine line reduces AI chess hallucinations but does not remove them: Stockfish checks the moves, not the sentences, so a confident-sounding explanation can still be wrong. The coach checks the words against the engine line before approving them for sharing. For what the AI can and cannot do in ChessCore, see ChessCore AI, and for how engine output differs from written explanation, see AI chess analysis compared with engine analysis.
Can I just paste a game into ChatGPT?
You can paste a game into a general chat assistant or a standalone engine, and for one position that may be all you need. ChessCore is not a general-purpose engine and does not try to be one. It is academy software, and what it adds is the record around the game: whose game it is, which batch the student is in, what the coach approved and shared, and what homework followed. A pasted game leaves you to carry all of that by hand for every student, every week.
Limits coaches should review
Engine output is a strong signal about a position, not a lesson plan. An AI explanation can be wrong or too advanced for the student, which is why it stays a draft and why a coach edits it. Chess.com games are imported read-only. Flagged moves show where the evaluation changed, not why a student of this level played them. Treat the flagged list as where to look first, and use your own judgment on the rest of the game.
To see a game go through the review with your own students' level in mind, you can ask for a walkthrough.
Request a demoAt a glance
The engine settings in numbers
16
Default search depth
Each analysed move is searched to depth 16, and the coach can pick from 8 to 22.
2
Best lines per position
The engine looks at the two best lines in each position.
120
Half-moves covered
Analysis covers up to the first 120 half-moves of a game.
300
Centipawns, already lost
Moves from a position already 300 centipawns or worse are not flagged.
200
Games in one PGN file
Importing a file does not itself request analysis.
FAQ
Frequently asked questions
Which games can ChessCore analyse?
Games synced from a student’s Lichess or Chess.com username, games imported from a Lichess or Chess.com link, and games loaded from a PGN file of up to 200 games. A student’s games sit in one record whatever their source. The coach can request analysis for any of them. With the academy’s practice loop on, which is the default, ChessCore also queues analysis overnight for up to two recent games per student that have none.
Does the AI explanation reach a student or parent by itself?
No. Per-move explanations start as drafts. The coach edits and approves each one, and only approved explanations can be shared. Class and fee reminders and the automatic monthly progress report are separate: they go out on a schedule the academy turns on, and the monthly report carries no AI-written text and no game analysis.
What happens to a flagged mistake after the review?
A coach can send mistakes flagged in a student’s analysed games into that student’s practice queue as puzzles built from their own positions. See student training for how practice is assigned and reviewed.
Related features
Works with the rest of the academy
- Batches and attendanceRecurring batches, a single class moved without touching the rest, attendance and make-up credits.
- Fees and UPI AutoPayOne invoice per fee, part payments kept as part payments, fixed reminder templates, Razorpay and Stripe.
- Parent portalFamilies check attendance, homework, reports and invoices with their own login, nothing to install.
- White label and your brandYour name, logo and colours on portals, emails and certificates, your own domain and your own app.
- Student trainingPuzzle sets and homework from 6,095,563 rated puzzles, with completion and accuracy per assignment.
- Progress trackingMonthly reports built from attendance, games, ratings and coach notes, reviewed before AI text reaches a family.
- Live online classesBuilt-in video, a shared board the coach drives, screen sharing, a whiteboard, polls and recording.
- TournamentsSwiss, round robin, knockout, arena, team and simul events with live standings and FIDE reports.
Written by Abhishek Tiwari. See our editorial standards.
