Most players fall into the same trap: play game after game, hoping volume alone produces improvement. Without review, volume mostly entrenches the mistakes you already make. A thousand games teach you very little; ten reviewed games teach you a lot.

Analysing your own games is the single highest-return activity in chess study, and losses are worth more than wins — a win often hides the same flaw that a loss exposes, because your mistakes simply went unpunished.

Here's a method that works, in two deliberate stages.

Stage 1: Analyse it yourself, engine off

The most common error in modern chess study is switching the engine on the instant the game ends. It feels efficient and teaches you almost nothing. Reading an evaluation bar is not analysis — it hands you the answer without ever making you do the work that builds calculation and board vision.

So before any engine touches the game, replay it yourself.

Find the turning point. Go through move by move and identify the single moment the position went from winning to equal, or equal to lost. One move. Being forced to commit to one is what makes this useful.

Reconstruct what you were thinking. At that move, what did you actually calculate? What candidate moves did you consider and reject? This is the part that transfers to your next game — you're auditing your decision process, not just the position.

Check your clock. Note the time you had left at each critical decision. A lot of "tactical oversights" are really time-management failures wearing a disguise.

Write it down. Something as short as:

Move 18...Nd5? — played it to centralise. Missed 19.Bxh7+ Kxh7 20.Ng5+.
I never checked his checks and captures before committing.

That last sentence is the actual lesson. "I hung a piece" is not a lesson; "I don't check forcing moves before committing" is something you can fix.

Stage 2: Verify with the engine

Now turn the engine on. Its job is to be an unbiased referee for the conclusions you already reached — confirming some, and exposing the blind spots you couldn't see precisely because they're blind spots.

The comparison is the valuable part:

Your conclusion:   "I lost because my endgame technique is weak."
Engine's verdict:  "You were winning (+3.5) until move 32, where you missed a
                    back-rank check and traded into a lost pawn ending."

Those are different problems with different fixes, and you'd have spent a month on the wrong one.

Three rules for using it well:

  1. Look for the big swings first. A move taking the evaluation from +1.5 to -2.0

has a concrete tactical reason. Find it.

  1. Play out the engine's line. If it suggests something surprising, walk the moves

until you understand why it's better. An engine move you don't understand teaches you nothing.

  1. Ignore small inaccuracies. +0.2 instead of +0.4 did not lose you the game.

Chasing these is the most common way to waste analysis time.

On LocalChess, Analyze game does this with the same local engine you played against, replaying the game and reporting where the evaluation moved. It runs on your device, so it works with no connection.

Categorise every mistake

Improvement comes from patterns, not incidents. Tag each error you find into one of four buckets:

  1. Opening slip — misremembered a line or came out of the opening with a bad structure.
  2. Tactical oversight — missed a check, capture or threat one to three moves deep.
  3. Strategic misjudgement — wrong plan, misread the imbalances, king left uncastled.
  4. Time trouble — a rushed decision under clock pressure.

The bucket matters more than the move. A player losing to bucket 2 needs tactics drills; a player losing to bucket 4 needs a clock rule, not more chess knowledge. They often feel identical from the inside.

Keep a mistakes logbook

One game's analysis is an anecdote. Twenty games' analysis is a diagnosis. Keep a running log:

| Date | Opening | Primary cause | Fix | | --- | --- | --- | --- | | 2026-07-10 | Italian Game | Missed a pin on f6 | 15 pin puzzles daily | | 2026-07-12 | French Defense | No space, no plan | Study the c5/f6 breaks | | 2026-07-14 | Slav Defense | Rushed the ending in time trouble | 30-second decision cap |

Review it monthly and count the categories. If 60% of your losses are tactical oversights, then reading another opening book is the wrong use of your time no matter how much you enjoy it. The category that repeats is your study plan — it's the whole reason to keep the log.

The hard part is emotional, not technical

There is a strong pull, after a painful loss, to close the board and immediately click "new game" to win it back. That's tilt, and it converts one bad game into a bad evening and sometimes a bad month.

Two rules that help:

  • **Never start a new game until you've found the exact move where the last one slipped

away.** It takes three minutes and it breaks the revenge loop.

  • Treat each loss as a lesson written specifically for you. It found a real gap in your

play, which is more than any book will do this week.

Put it into practice

Play one game on LocalChess and commit to reviewing it before you play another — win or lose. Do stage 1 with the engine off, then check yourself with Analyze game, and write one line in your logbook.

Do that for ten games and the pattern will be obvious. Then go and fix that one thing.