An evaluation is a clue, not the explanation.

Traditional engine displays express evaluation in pawns or centipawns: 100 centipawns equals one pawn unit. The sign is usually from White's point of view, so +1.0 favors White and -1.0 favors Black. Some interfaces instead show the score from the side to move; check the interface before comparing values.

The unit is not a material receipt. Position, king safety, initiative, pawn structure, and long-term piece activity all influence the score. Modern Stockfish also normalizes its evaluation so the scale relates to expected winning chances in Stockfish self-play. That makes “one pawn” a convenient unit, not a claim that a pawn is physically missing.

A display such as M3 means the engine has found a forced mate within its convention for counting moves. It is not merely a very large positional advantage. The defending side may have several legal replies, but none avoids mate against the engine's calculated continuation.

Mate scores can change as search deepens, especially in composed positions or long forced sequences. If a lesson depends on a distant mate, verify the line at sufficient depth and focus on the first forcing idea a player could realistically recognize.

Suppose the position is roughly equal before your move and clearly worse afterward. The swing tells you that the move changed something important. It does not yet say whether the cause was a hanging piece, a tactical sequence, a bad exchange, or a strategic concession.

Compare the played move with one or two alternatives and inspect the opponent's reply. If every reasonable reply keeps the score stable except one forcing move, the missed forcing move is the lesson. If the score drifts because your pieces become passive over several moves, the lesson may be a plan rather than a tactic.

An engine score is the result of a particular search. More time and depth can reveal tactics that were initially invisible or reorder close candidate moves. A shallow browser analysis is useful for triage, but it is not a proof that every quiet position has been solved.

Positions often contain several moves with nearly equal evaluations. Treat a tiny difference as a preference, not a moral grade. For learning, the move whose purpose you understand and can execute may be more useful to study than a narrow engine edge that depends on a difficult sequence.

First state the score change in plain terms: equal became worse, winning became unclear, or a forced mate appeared. Next name the opponent's critical reply. Finally, connect that reply to a visible fact on the board.

A complete explanation sounds like: “After the pawn push, the bishop can take the knight because its only defender is pinned.” The evaluation confirms the consequence, while the sentence supplies the pattern you can notice again.