How Noctie's AI Chess Helper Grades Every Move
An AI chess helper is only useful if its advice arrives when you can still act on it. Noctie’s helper grades every move the moment you play it, on a seven-step scale from Blunder to Excellent, and the grade is based on how good the move looks from a human perspective rather than on the objective evaluation of an engine like Stockfish. That single design choice is why the feedback feels like a coach tapping the table instead of a computer printing “-1.37”.
The seven colors, in plain words
| Grade | Meaning |
|---|---|
| Excellent | A great move that’s hard to find |
| Great | A strong and non-obvious move |
| Good | One of the best moves in the position |
| OK | A somewhat unambitious move |
| Dubious | An inferior move but not a clear mistake |
| Mistake | A clear mistake or missed opportunity |
| Blunder | A critical mistake harming your chances |
Purple and blue at the top, greens in the middle, brown, orange and red at the bottom — the same colors appear on the move list, on the board, and later on the flashcards Noctie builds from your mistakes.
Why “looks good to a human” beats “centipawn loss”
Engines evaluate positions with brute force: millions of positions per second, and a number. Grade your moves by that number and two annoying things happen. First, a perfectly reasonable move gets punished because of an obscure engine line no one at your level would find. Second, a genuinely clever move that isn’t quite the engine’s first choice gets no credit at all.
Noctie is trained on over a billion human games, so it can ask a different question: compared with what humans at this level actually play here, how does this move rate? Reasonable moves that fail only to a computer refutation are not punished; clever moves that are strong and non-obvious are rewarded with “Great” and “Excellent”. In short: you “won’t be punished for reasonable moves that fail due to some obscure engine line” and you “get rewarded for moves which aren’t just best, but also really clever”.
What instant feedback does to your habits
The value of instant feedback is timing. Because the color appears while the position is still on the board and the idea is still in your head, your brain connects the pattern with the outcome without a separate analysis session. Noctie’s roadmap even lists work on making the colors more accurate and less harsh depending on your level — the aim is a signal you can trust and keep reading, game after game.
Three habits the feedback builds quickly:
- Noticing risk before you commit. After a few games you start predicting the color before you release the piece — that hesitation is exactly the calculation habit stronger players have.
- Taking moves back on purpose. Made a mistake? Noctie lets you take it back and ask for a hint. Used sparingly, this turns a lost game into a lesson instead of a resignation.
- Reviewing by color, not by move number. Scan the move list for the first orange or red; that is where the game was decided.
From colors to flashcards
Feedback and flashcards are two halves of the same loop. After each game Noctie picks the most instructive mistakes — not the biggest engine swings — and turns them into puzzles that come back on a spaced-repetition schedule. The color tells you a mistake happened; the flashcard makes sure you would find the better move next time.
Trying the helper for yourself
The free rating test — a calibration game — finds your level with no trial or subscription; the full feedback loop is part of every plan. Every plan (monthly €14/month, yearly €8/month billed annually) starts with a free trial that lasts 7 days, with unlimited instant move feedback across web, iOS and Android.