How to Analyze Your Chess Games (A Repeatable System)
The engine says -2.3 and moves on. You still don't know why your plan fell apart, so next game you do it again. Sound familiar?
How to analyze chess games, in five steps that work on any site: (1) replay it once with the engine off and mark where you felt lost, (2) run an engine and find the biggest eval swings, (3) for each swing, dig out WHY the better move was better, the plan behind it, not just the square, (4) name the recurring pattern, and (5) log it so it carries into your next game. They run the same on chess.com's Game Review, on Lichess, out of a book, wherever you happen to play. Lichess's free "Learn from your mistakes" review even cross-checks against known opening theory, so a well-known book move rarely gets flagged as a mistake. Step three is the one nobody does properly, and it's the one that actually raises your rating.
One boundary first, because it matters. This is about reviewing a game that's already over. Not live help mid-game. Using an engine or a coach during a rated game is cheating, and we don't build for it. Everything below is what you do after the clock stops.
The 5-step method for analyzing any chess game
Most people open the engine, scroll to the red bars, nod at the ones they already suspected, close the tab. That's not analysis. That's just confirmation. The method below is slower on purpose. The slow part is where the learning hides.
- Replay the game first, engine off. Click through your own moves before any computer touches it. At each critical point, ask what you were thinking, and mark where you felt lost or rushed. Almost nobody does this step. It's the one that shows you where your judgment failed, not just where the eval did.
- Now run the engine. Read the swings as questions. Turn the analysis on and hunt the big jumps, the spots the graph fell off a cliff. Each one's a question, not a verdict. Why was this so much worse than what the engine wanted?
- For every swing, find the WHY. This is the real work. Don't just note that g6 was better. What did your move let happen? What did it stop defending? What plan did it abandon? Can't reconstruct the reason? That's your gap, and it's exactly when to ask a coach that explains.
- Name the recurring pattern. Three or four games in, you'll see the same mistake wearing different clothes. "I trade my good bishop too early." "I stop calculating once I'm ahead." Name it. A named mistake is one you can catch mid-game.
- Log it, one line. Write the pattern somewhere you'll actually see it. A note, a spreadsheet, the back of a napkin, whatever works. That log is the difference between a one-off review and a habit that sticks. Skip it, and next month you're paying tuition on the same lesson all over again.
Notice what's missing from that list: memorizing engine lines. You don't need to. Knowing the best move in one dead position does nothing for you. Understanding the pattern, though, means you find it yourself next time, over the board, no engine anywhere near you.
The engine is a metal detector, not a teacher. Stockfish, the open-source engine behind the analysis on every major site, searches many millions of positions a second, often over 100 million on modern hardware, per Stockfish's own benchmarks, though the exact rate varies and reports each one in centipawns, where 100 equals a full pawn of advantage. It's brilliant at finding WHERE you went wrong. It says nothing about WHY. Steps one and three are the human half of the loop, and they're where your rating actually moves.
What to look for at each phase: opening, middlegame, endgame
A blunder on move 8 and a blunder on move 48? Not the same animal. Different phase, different question to ask. What to check in each one, plus the repeat-mistake that likes to hide there. Most of the short "3-step" guides skip this part entirely.
| Phase | What to check | The question to ask | Common repeat-mistake |
|---|---|---|---|
| Opening | Did you develop with a plan, or just make moves? Where did you leave book, and was your next move sensible? | "What was my piece actually doing here?" | Moving the same piece twice, or memorizing lines without knowing the idea behind them. |
| Middlegame | The eval swings almost always live here. Check the moment the position turned: a missed tactic, a bad trade, a plan with no follow-up. | "What was my opponent threatening that I ignored?" | Playing your own plan while blind to theirs. Trading off your best piece for their worst. |
| Endgame | King activity, passed pawns, and whether you converted a winning edge or let it slip. Count tempi on the pawn races. | "Did I know the right technique, or did I guess?" | Shuffling in a won position, missing the one accurate plan, drawing games you should win. |
One honest note. Most improving players lose in the middlegame, not the opening. Review time tight? Spend it there. The opening feels important because it's the part you can study cleanly from a book. The eval graph usually says otherwise. Look at where your own games actually fall apart, and you'll know exactly where to point your attention.
The tools, honestly: your own guesses first, then an engine, then a coach
There's no single right tool. There's a right order. Guess first, verify with an engine, then get the why when the engine line doesn't click. Each stage does a job the others can't, and skipping straight to the engine is exactly why so many reviews teach nothing.
| Tool | What it gives you | What it cannot do | Best for |
|---|---|---|---|
| Your own eyes (engine off) | Builds real judgment. Forces you to commit to a read before the machine spoils it. | Will not catch the deep tactic you genuinely could not see. | Step one, every game. Non-negotiable. |
| An engine (Stockfish, on Lichess or chess.com) | Ruthlessly accurate. Finds exactly WHERE the eval dropped and what the best move was. | Says nothing about WHY in words. Hands you a number and a line, no reasoning. | Finding the moment it went wrong: not to memorize the line, but to locate the swing. |
| A coach that explains | Turns the engine eval into a plain-language reason you can actually use, and you can ask a follow-up. | Not a replacement for the raw engine numbers. It sits on top of them. | Step three, the why, when the better move makes no sense to you. |
Let me be fair to the free options, because honestly they're good. Want raw engine numbers for nothing? Lichess. Full stop. Free Stockfish on every game, eval graph, best lines, all free forever because it's a nonprofit. We use it ourselves. And doing it by hand, engine off? That's real analysis too, maybe the most valuable step of the lot, and it costs you nothing but attention. Neither one is a lesser choice. They're the first two rungs of the ladder.
The gap opens on the third rung. The engine tells you g6 was better. It won't tell you why your move lost. That's where a coach that explains earns its place, and it's the job Chessvia's AI chess coach is built for. Reads the same finished game, hands you the reason in words.
The step everyone skips: understanding WHY, not just labeling the blunder
This is the exact moment analysis goes wrong for most people. The engine flags move 23 as a blunder, drops a red "-3.5" next to it, shows you the better move, and calls it done. Cool. But why? What did you actually miss?
That gap isn't small. Label a move a blunder and you've learned nothing you can reuse. Understand why it was a blunder and you've learned a pattern you'll spot in a hundred future games. So picture the difference. The engine says "Blunder. Better was g6." A review built to explain would say something closer to this:
"This dropped the game. You played Nf6 to hit the queen on h5, but f7 was defended only by your king, and White's queen on h5 and bishop on c4 were both already aimed at it. That let Qxf7 land, and it's checkmate on the spot. The saving move was g6 or Qe7, covering f7 before the threat arrived."
Read those two again. Same engine underneath. Completely different lesson. The first hands you a verdict. The second hands you a reason you can carry into your next game, and it invites the follow-up an engine report can never answer. "Why not the other rook?" Ask it. You get an answer on the same position.
That's what Chessvia's free review at /analyze is built to do. You import a finished game, and instead of only grading each move, it explains the reason in words. One honest note on how it works, because we won't oversell it. The coach's read is grounded in a real engine. It checks the position against Stockfish first, then explains, rather than making up a reason from memory. It won't be right every single time, no honest tool is, and we measure how often it stays grounded rather than pretend it's perfect.
But when the goal is understanding why your move failed, a grounded explanation in plain words beats a number you have to decode yourself. Run it on one of your own losses. You'll feel the difference in about a minute.
Review One of Your Games Free: See the WhyHow often, and which games: wins or losses?
Two questions people always ask, and the honest answers are simpler than you'd think.
Which games? Deep-dive your losses, obviously. But don't stop there. The sneaky ones are the lucky wins, the games where your opponent blundered back and bailed you out. Those hide mistakes the scoreboard rewarded, so your brain files them under "good game" and you make the same error again next week. A close loss and a lucky win teach you the same amount. Review both.
How often? You don't need to deep-dive everything. That road leads to burnout and an untouched backlog. A lighter rule works better:
- Every serious game gets a 60-second glance. Engine on, find the single biggest swing, note it, move on.
- Deep-dive the games that hurt. The painful losses and the lucky escapes get the full 5-step treatment.
- Once a week, read your log. Pull up the patterns you named and ask: am I still doing this? That weekly loop is what turns notes into fixed habits.
Consistency wins here, not intensity. One deep review a week that you honestly log will out-improve a five-hour binge you do once and never touch again. The players who climb fastest? Not the ones grinding the most games. The ones who actually close the loop on the handful they do review.
Which platform? Link down to the specifics
This method is deliberately platform-agnostic. The five steps don't care where you played. But the exact buttons do differ, and if you want the click-by-click version for your site, we wrote it up separately:
- On chess.com and hitting the paywall? How to review a chess.com game for free covers the PGN-export trick that gets you past the one-review-a-day limit, and whether the ~$12/mo Diamond plan is worth it.
- On Lichess? The full Lichess game review walkthrough covers what's free and unlimited, and the one thing it won't explain.
- Stuck on why one specific move was so bad? We dug into what actually makes a move a blunder on its own.
Same system either way, just different buttons. Grab the guide for your platform, then circle back to the five steps.
Frequently Asked Questions
How do I analyze my chess games for free?
You have solid free options. Lichess runs free Stockfish computer analysis on every game, with an eval graph and Inaccuracy/Mistake/Blunder labels, no cost. Chess.com gives one full Game Review a day on a free account, and you can export the PGN for the rest. Chessvia's free review at /analyze reads a finished game and adds the plain-language reason each move helped or hurt. Start by replaying the game engine-off yourself, which costs nothing and teaches the most, then bring in a free engine to check where you went wrong.
Should I analyze my games with or without the engine first?
Without, first. Always. Click through your own game and commit to a read at each critical moment before the engine spoils it. This is what builds judgment: you're testing your own thinking against the truth, not just being handed the answer. Then turn the engine on to find the swings you missed. Do it the other way around and you'll nod along at every red bar and learn almost nothing.
How do I analyze my games with Stockfish?
Run Stockfish over the whole game, on Lichess or chess.com or a local board, and watch the eval graph. The moments it jumps sharply are your mistakes. For each swing, don't just copy the better move, work out why it was better: what your move hung, what plan it abandoned, what your opponent was threatening. Stockfish tells you WHERE with remarkable accuracy. The WHY is on you, or on a coach that explains the line in plain words.
Should I analyze my wins or my losses?
Both, for different reasons. Losses show your mistakes plainly, so they're the obvious deep-dive. The trap is your wins, mainly the lucky ones where your opponent blundered back. Those hide errors the result rewarded, and your brain marks them "fine" and repeats them. So review the painful losses and the games you barely escaped. You'll catch more than losses alone ever reveal.
How often should I analyze my chess games?
Skip the deep-dive on most games. Give each serious one a 60-second glance to catch the biggest swing, and save the full treatment for the games that hurt: close losses and lucky wins. Then once a week, re-read your log of named patterns and ask, honestly, am I still doing these? That weekly check is the whole trick. One review that closes the loop beats a marathon you never repeat.
Does the engine tell me why a move was bad?
No, and that's the core limitation. An engine gives you an evaluation and the better move, a number and a line. It never explains, in words, why your move lost: what it hung, what plan it dropped, what threat you missed. Reconstructing that reason is the actual skill. If you can look at "Blunder, better was g6" and rebuild the why yourself, the engine is all you need. If you can't, a coach that explains, checked against a real engine first, is what closes that gap.
A fancier engine won't make you better. A repeatable habit will. Replay it yourself, find the swing, understand the why, name the pattern, log it. Do that on your losses and your lucky wins, once a week, and the same old mistakes quietly stop coming back. Import a game you actually played, hunt down the move that lost it, and get the why in plain words.
Play a Game Against a Coach That Talks You Through Your MovesFree 7-day trial, no credit card required. Voice coaching is in early access.