How it actually works
Import your games
Connect your Chess.com account or paste a PGN. New games sync automatically after that.
Record your thinking, before reveal
At each key moment, write what you were actually thinking and what you'd play with a second look — before the engine says a word.
See the verdict, and why
Only now does the engine's take appear — with a concrete explanation of the chess mechanics, not just an evaluation number.
See your patterns over time
The same kind of mistake, showing up across games, is easy to miss one at a time and hard to miss in aggregate.
A real explanation, not a mockup
This is the actual four-section card every key moment gets, with real sample content from a real game.
What your move did
Nd4 heads for a strong, centralized outpost, eyeing b3, c2, and f5 — a reasonable-looking plan on its own.
What it missed
The bishop on g7 already controls the long dark-square diagonal all the way to d4, and nothing White has covers that square. Bxd4 just wins the knight outright, for free.
Why the recommended move is stronger
Nc2 aims at the exact same outpost but gets there through a square the bishop doesn't touch. The engine's line runs Nc2 Rb8 Nd4 — the knight reaches d4 two moves later, completely safely. (engine: +0.3 vs -3.1)
What to remember
Before parking a piece on a square, trace every diagonal running through it — a long-diagonal bishop is exactly the kind of thing that's invisible if you're only looking near your own pieces.
Why this is different
Engine-only tools grade the move. ReflectChess grades the gap.
A best-move arrow and an eval bar tell you what was wrong. They never find out what you actually believed when you played it — which is the only thing you can actually fix.
The reveal comes second, not first.
If you see the verdict before you've committed to your own reasoning, you're not reflecting — you're just reading. The order is the whole point.
Explanations name the pieces and squares, not just the score.
"+1.4 vs -3.0" isn't a reason. A hanging piece on a named square, a pin along a named diagonal — that's something you can actually learn from next time.
A note from the person who built this
I got tired of finishing a game, running it through an engine, and staring at a red arrow that told me a move was bad — without ever being asked what I actually thought was happening on the board when I played it.
That gap between what you believed and what was true is where the actual learning lives. Most tools skip straight past it. ReflectChess is my attempt to slow that step down instead of skipping it.
This is a small, early product. It doesn't have a user base or a leaderboard to show you — just a mechanism I think is worth trying, built by someone who plays and loses plenty of games himself.