How Game Review annotates your moves

When you import a game or play one on TacoChess, our analysis pipeline runs Stockfish over every move and tags it with a quality label.

The labels

• Brilliant (gold) — a move that is the best, hard to find, and either tactical or non-obvious. Most games have zero or one of these.
• Best (green) — the engine's top choice
• Good — a reasonable move that does not lose the position
• Inaccuracy — small drop in evaluation
• Mistake — meaningful drop, often turning equal into worse
• Blunder (red) — a move that loses material or the game

What "engine eval" means

The number next to each move (+1.2, -0.4, #3) is Stockfish's evaluation in pawn units — positive means White is better, negative Black. #3 means mate in 3.

Why labels are sometimes wrong

Two reasons:

1. Time-control mismatch. A "blunder" in a 60-second bullet game may be the only practical option. We weight evaluations by time control where we can.
2. Engine depth limits. Our default analysis depth is 18 plies. For complex tactical sequences this sometimes misses lines a deeper search would find.

Pro accounts can request "deep" re-analysis (depth 32) on any game.