A modern chess engine on a phone is stronger than any human who has ever lived. Open an analysis board and you get an evaluation for every position and a "best move" for every situation. It sounds like having a perfect coach on call.
Yet plenty of people have used engines for years without improving much. The reason is simple: they let the engine think for them. This guide covers five ways engine output can mislead you, and a routine that keeps your brain doing the work.
Five ways engine output misleads
1. "+2" doesn't mean you can win it
An evaluation of +2 means that with perfect play, one side has a large advantage. But perfect play might be a ten-move sequence no human would find. A +2 position that requires precise calculation can be harder to play than a +1 position with a clear, simple plan.
Takeaway: when choosing between lines, also ask: can I actually understand and execute this?
2. The "weird" engine move
Engines calculate very deeply. Some moves they choose rely on reasons buried twenty moves into a line. Copying such a move without understanding teaches nothing.
Takeaway: if a few minutes of playing through the line still doesn't reveal why the move is good, move on. Focus on moves you can understand.
3. Tiny differences
The engine might rate one move +0.4 and another +0.2. For practical purposes, that difference is meaningless. Both are good moves. Don't beat yourself up for choosing "the second-best move".
Takeaway: pay attention only to big drops in evaluation — those are the real mistakes. Our guide to blunders, mistakes and inaccuracies explains the usual thresholds.
4. Quick analysis can be unstable
Review tools typically analyze each move for a short time. In complex positions, a short analysis can change its mind with more depth. The review on our play page is designed the same way: fast enough for phones, good enough to catch real mistakes, but not deep analysis of every subtle choice. See understanding accuracy scores.
Takeaway: for critical positions, play out the lines on the board and check them yourself rather than trusting a single number.
5. Engines don't know your clock
An engine evaluates the position, not the situation. With ten seconds left, a safe simple move is worth more than a "best" move that needs five minutes of calculation.
Takeaway: when reviewing, remember the circumstances — time, pressure — to judge your decisions fairly.
A routine that keeps you thinking
Step 1: Analyze first, alone
Before opening the engine, go through the game yourself. Where was the turning point? Which moves were you unsure about? What were you thinking? Write a few lines.
Step 2: Open the engine and find the big mistakes
Look for moves where the evaluation drops sharply. Ignore small fluctuations.
Step 3: Compare with your own analysis
Did the engine agree with your turning point? If it flags a mistake you didn't notice at all, that's a blind spot — the most valuable thing you can learn from the game.
Step 4: Understand, don't just look
For each big mistake, play through the engine's suggestion and the next few moves. Put the reason into words: "This move is better because it stops the knight from landing on d5." If you can't explain it, you haven't really understood it yet.
Step 5: Write down one lesson
One line per game: "Missed a fork on c7 after moving my queen." After a few weeks, these notes show your most common mistakes. More on this in how to analyze your games.
Habits to avoid
Opening the engine immediately after the game. You skip the step where most learning happens.
Checking every single move. Focus on the few that mattered.
Memorising engine moves without understanding. When the position changes slightly, you won't know what to do.
Staring at the evaluation number. The number explains nothing. The reason is what matters.
Using an engine during a game against a person. That's cheating — and it stops you from improving. See fair play and cheating.
Using engines for specific training
Beyond reviewing games, engines are excellent for targeted practice:
- Endgame drills. Set up a basic endgame, like king and rook against king or the Lucena position, and play it out against the engine. The engine defends perfectly, so you only succeed with correct technique.
- Opening checks. After a game, check the first few moves of your opening against the engine to see where you left good play. Then read about why that deviation mattered.
- Position play-outs. Stop at a critical moment from your game and play it out against the engine from both sides. You'll quickly see which plans work.
In each case, the principle is the same: the engine sets the problem or checks the answer, but you do the thinking.
The engine is still your best training partner
None of this means avoiding engines. They find tactical mistakes others miss, practice endgames with you tirelessly, and are always available. Just use them in the right order: you think first, the engine checks after.
More in the game review section.
Frequently asked questions
Are engines ever wrong?
With enough time, engines very rarely get the best move wrong. But short analyses can be unstable in complex positions, and the evaluation number doesn't reflect how hard a position is for a human to play.
How big a difference matters?
There's no absolute threshold, but for club players, differences of a few tenths of a pawn usually don't matter. Focus on large drops.
Should I learn openings with an engine?
Engines tell you which moves are good but not why. Combine them with explanations from books or articles.
How do I know I understand an engine move?
If you can explain in words why it's better than the move you played, you understand it.
Should I analyze every game with an engine?
Analyze your losses and any game with an important turning point. You don't need to go through every game in detail.
Is it bad to check the engine during a casual game with a friend?
Only if you both agree to it as an analysis session. In any game where you're competing, using an engine removes the point of playing and, in rated games, counts as cheating.
What engine settings should I use for review?
The defaults of most review tools are fine for finding big mistakes. For important positions, let the engine think longer, and compare several candidate moves rather than just the top line.
