You finish a tense game, paste it into the analysis board, and the engine slaps a verdict on every move: +1.8 here, -0.3 there, a sudden #-5 you never saw coming. Most adult improvers treat that number as a grade. It is not a grade. It is a forecast made by a player thousands of rating points stronger than the strongest human alive, assuming both sides keep playing at that level. Read it that way and the evaluation becomes one of the best teachers you have. Read it as a report card and it quietly makes you a worse player.
This guide explains what the number actually measures, why a “winning” evaluation often means nothing at your level, and a workflow for turning engine feedback into study themes instead of anxiety.
What the number actually measures
The evaluation is expressed in centipawns — hundredths of a pawn. An evaluation of +1.00 means the engine judges White to be ahead by the equivalent of one pawn, all else being equal. +3.00 is roughly a minor piece. The sign tells you who is better (positive favors White, negative favors Black), and the magnitude tells you by how much, measured in that abstract pawn currency.
Two things about this are easy to miss. First, the unit is not literal material — a position can read +1.5 with equal material because the engine is pricing in king safety, piece activity, pawn structure, and long-term pressure. Second, and more important, the evaluation assumes best play from both sides for the rest of the game. That assumption is doing enormous, invisible work. The engine is telling you what happens if both players are the engine. Neither of you is.
Centipawns to feelings: a rough translation
For practical purposes, here is how to read the eval bar without overreacting:
- 0.0 to plus or minus 0.5 — equal. Treat any move in this band as fine. Chasing a “more accurate” move here is wasted study energy.
- 0.5 to 1.5 — a real but small edge. The kind of advantage that decides games between strong players and evaporates constantly between improvers.
- 1.5 to 3.0 — a serious advantage. Worth understanding why it appeared.
- 3.0 and beyond — usually a material gain or a decisive attack. This is where your blunders live, and where your study should concentrate.
Why “+1.5, winning” is a lie at your level
Here is the single most useful idea in this article: the engine’s evaluation is only as real as the players’ ability to defend it. A +1.5 between two 2700s is nearly a win, because both sides will find the precise moves that keep the advantage alive. The same +1.5 between two 1200s is, statistically, close to a coin flip. The position will swing several times before move 40 because neither player can hold a slight edge under pressure.
This is why obsessing over small evaluation losses is a trap for improvers. If your move dropped the eval from +0.4 to -0.1, you did not blunder — you played a normal human move in a roughly equal position. The engine flags it because it sees a marginally better try, but that “mistake” has almost no bearing on whether you win the game. Your results are decided overwhelmingly by the large swings: the hung pieces, the missed tactics, the collapses in won positions. (We break that pattern down in our guide to why you keep hanging pieces and how to stop.)
The “best move” trap
Modern analysis tools highlight the engine’s top choice in green and beg you to play through it. The danger is subtle: you start learning moves instead of ideas. You memorize that the engine wanted Nd5 in this exact position, which will never recur, instead of understanding the recurring idea — a knight heading to a hole the opponent can no longer contest.
The best engine moves are frequently the least instructive for an improver, because they rely on calculation depth you cannot reproduce over the board. A line that is “winning” only if you find seven precise moves in a row is not a lesson you can use on Tuesday night. The move you should study is the human one: the clear plan that keeps the position simple and your pieces working. When the engine and a sensible plan disagree by a quarter of a pawn, follow the plan.
A workflow that builds skill instead of dependence
The way most improvers use an engine — turn it on, watch the bar, nod — trains nothing. Calculation is a muscle, and the engine is a forklift. Use this order instead.
Step 1: Analyze with the engine off
Before you turn anything on, go through the game yourself. Mark the moments you remember feeling lost, the moves you were unsure about, and your own guess at what went wrong. This is where the learning actually happens, because you are reconstructing your decision process. Our full method for this is in how to analyze your own chess games.
Step 2: Turn the engine on to check, not to lead
Now switch it on and compare its verdict to yours. You are not looking for every 0.3 it disagrees with. You are hunting for the two or three moments where the bar jumped by more than 1.5 — the real errors. For each one, cover the engine line and ask: could I have seen this? What would have warned me? A hanging piece, an undefended back rank, an in-between move. Name the pattern.
Step 3: Convert evaluations into themes, not corrections
One game tells you noise; ten games tell you your signal. If the engine keeps punishing you at the same moment — the transition out of the opening, positions with an isolated pawn, time-pressure endgames — that is your study theme. A single corrected move is forgotten by next week. A theme becomes a training plan. This is exactly how to build a study plan that actually works.
When the engine is “wrong” for you
The engine is never wrong about the position. It is frequently wrong about you. It will recommend a sharp piece sacrifice that is winning with perfect follow-up and losing with yours. It will rate a calm, safe move as half a pawn worse than a chaotic one. For an improver, the safe move is often the better practical choice, because your job is to win games against humans, not to satisfy a 3600-rated oracle.
This is also why raw engine work does not transfer the way people expect — the same reason your puzzle rating can climb while your game results stall. We unpack that gap in why your puzzle rating lies. The fix is the same in both cases: train the decision, not the answer.
The bottom line
The evaluation bar is a compass, not a scoreboard. Use it to find the few moments that actually decided your game, ignore the rounding-error fluctuations that decide nothing at your level, and always translate a number into an idea before you move on. Do that and the engine becomes what it should be: a sparring partner that points at your real weaknesses, instead of a judge handing out grades you cannot act on.
Turn your engine reports into a real plan
If your game reviews keep flagging the same weaknesses but you are not sure what to actually train, that is the gap MyChessPlan was built to close. Start with our free archetype report — answer a few questions and find out which type of improver you are and where your evaluation losses really come from. Ready to go further? The $14.99 personalized improvement plan turns those patterns into a week-by-week training schedule built around your actual games. Get your free archetype report at MyChessPlan.com.
Frequently Asked Questions
What does a +1.5 engine evaluation mean in chess?
It means the engine judges that side to be ahead by about one and a half pawns’ worth of advantage, assuming best play from both sides for the rest of the game. Between strong players that is a serious edge, but between improvers it is closer to a coin flip, because neither side can reliably hold a small advantage under pressure.
What is a centipawn?
A centipawn is one hundredth of a pawn, the unit chess engines use to express evaluations. An evaluation of +1.00 equals a one-pawn advantage, +3.00 is roughly a minor piece, and values inside half a pawn either way are effectively equal.
Should I always play the engine’s top move?
No. The engine’s best move often relies on calculation depth you cannot reproduce over the board, so it teaches you a specific move instead of a reusable idea. When a clear human plan is only a fraction of a pawn worse, the plan is usually the better practical choice for an improver.
How do I use a chess engine without becoming dependent on it?
Analyze the game yourself first with the engine off, then turn it on only to check your conclusions and find the large evaluation swings. Focus on the two or three real errors per game, name the pattern behind each one, and convert recurring patterns into study themes rather than memorizing individual corrections.