The Post-Loss Review Checklist: What To Do Before You Open The Engine

You lost. The game is sitting in your Lichess or Chess.com archive with a red dot next to it, and the fastest thing in the world would be to click “Analysis” and let Stockfish tell you where you went wrong. Thirty seconds later you have a squiggly evaluation graph, four red “Blunder” tags, and a vague sense of shame. You close the tab. Nothing changes.

That workflow feels like study. It isn’t. The engine is answering a question you never asked, which means you can’t remember the answer a week later. This page is about the twenty to thirty minutes that come before the engine, and why skipping them is the single biggest reason club players analyse hundreds of games and improve at none of them.

If you want the full picture of how engine work fits into a training routine, the parent guide on analysing your own games with an engine covers the whole arc. What follows here is just the pre-flight checklist.

Why Engine-First Analysis Fails

Here’s the mechanism. When you look at a position cold and Stockfish tells you 22…Rfd8 was the move, your brain files it as a fact about that position. Facts about specific positions are almost worthless: you will never see that exact position again. What transfers is the reasoning — and reasoning only gets corrected if you produce some first.

Psychologists call this the generation effect, and it’s brutal in chess. A 2019 study of adult learners found recall roughly doubled when subjects attempted an answer before seeing it, even when their attempt was wrong. Especially when their attempt was wrong. The engine can only correct a belief you’ve actually stated.

There’s a second problem, more specific to chess. Stockfish 17 running at depth 30 on your laptop plays at something north of 3500 Elo. Its top line frequently rests on a tactical resource eight moves deep that neither you nor your 1400-rated opponent would find in a hundred games. When it says 18.Nd5 is +1.4 and your 18.Be3 is +0.2, that 1.2-pawn gap may be entirely invisible from the board. You cannot learn from it. Meanwhile the move that actually lost you the game, the one where you spent 4 seconds and dropped from +0.3 to -0.1, gets no tag at all because it isn’t a blunder by the engine’s threshold.

Step One: Sit With The Result For Ten Minutes

Not metaphorically. Do not analyse a loss within ten minutes of playing it, and ideally not within an hour. Tilt makes you review defensively: you go hunting for the moment you were “winning” so you can tell yourself the loss was a fluke. That’s confirmation bias with a chess clock attached.

The practical version: finish your session, close the app, and queue the game for review later that evening or the next morning. Lichess makes this easy. Open the game, hit the bookmark icon, and it lands in your Bookmarks list. On Chess.com, add the game to a Study or just paste the URL into a notes file. Build a small backlog and review in batches of three or four, which also lets you spot patterns across games that are invisible one at a time.

Step Two: Reconstruct The Game From Memory

Before you open the game record, write down as much of it as you can from memory. Not the whole move order necessarily, but:

• The opening you played and when you left theory
• Your plan in the middlegame, in one sentence
• The three or four moments where you felt uncomfortable
• The move where you think it went wrong

This takes five minutes and does two things. It tells you what you actually understood at the board, and it exposes the gaps: if you can’t name your middlegame plan, that’s your finding, and no engine was ever going to give it to you.

A worked version, from a 1450-rated player’s London System game:

Opening: London, out of book around move 8 when he played ...Qb6
Plan: attack kingside with Ne5, Qf3, Rf3-h3. Never really got going.
Uncomfortable at: move 12 (his queenside pawns), move 19 (my knight
   on e5 got traded and I had nothing), move 25 (time trouble)
Think it went wrong: move 31, hung the exchange

Note what’s already visible. The player thinks the game was lost at move 31. But they’ve written “move 19: had nothing” without connecting it. Ten to one the engine graph is flat at move 31 because the position was already lost, and the real error is somewhere around 17-19.

Step Three: Play Through It Once, Slowly, With No Evaluation Visible

This is the core of the checklist and the part people skip.

Open the game on Lichess and press the “l” key to toggle local engine off, or on Chess.com use the board without starting the Game Review. Set it so you see no evaluation bar, no move classifications, nothing. Then walk through your own game as if it were someone else’s, one move at a time, and at every move where you had a genuine decision, stop and write:

1. What did I think my opponent was threatening?
2. What were my two or three candidate moves?
3. Why did I pick the one I picked?

Do this in a notation field, not in your head. Lichess lets you type comments into a Study chapter; Chess.com has a notes pane in Analysis. Either works. The written record is the point, because in two weeks you’ll want to compare what you thought against what was true.

Expect this to take fifteen to twenty-five minutes for a 40-move game. That’s correct. If it takes four minutes, you’re clicking through, not thinking.

Step Four: Build A Time-Spend Map

This one’s mechanical and it’s the highest-yield thing on the list for players under about 1700.

Lichess shows you move times by default on the game page; Chess.com shows them in the move list if you toggle “Show timestamps” in board settings. Go through and note every move where you spent under 5 seconds in a position that wasn’t forced, and every move where you spent over 90 seconds.

Here’s a real pattern from a 15+10 game:

Read that table again. The player spent almost two minutes on move 15 and two seconds on move 14. But move 14 is where the game turned: they had a real choice and treated it as automatic. This is the classic club-player failure, and you can diagnose it with zero engine assistance. The engine will not flag 14…Nc6 because it’s fine or near enough; the problem isn’t the move, it’s that no thought went into it.

Mark those fast moves. They’re your candidate list for engine checking later, and they’re far more instructive than the moves the engine tags red.

Step Five: State Three Hypotheses In Writing

Now, and only now, commit to what you think went wrong. Three claims, specific enough to be falsified:

1. I played 14…Nc6 without considering b4-b5, and that’s when I lost the initiative.
2. My kingside attack plan was wrong for the structure after he played …Qb6 and …c5 — I should have been playing on the queenside.
3. I was down to 2 minutes by move 28, so the move-31 blunder was a time-management symptom, not a calculation failure.

These are excellent hypotheses because the engine can confirm or refute each one within about four minutes. Hypothesis 1 is a concrete position check. Hypothesis 2 is a plan question you’ll answer by looking at what Stockfish’s top lines do across moves 12-20, not at any single evaluation. Hypothesis 3 isn’t an engine question at all: you’ve already answered it with your time map.

Compare that to the standard approach, which is to open the engine with no hypotheses and let it show you six moves you don’t understand.

Step Six: Check Your Opening Against A Database, Not An Engine

Separate task, separate tool. Before Stockfish, find where you left book.

On Lichess, open the game in Analysis and click the “Opening explorer” tab. Set the database to Lichess games, filter to ratings 1400-1800 and your time control, and step forward until your move drops to a handful of games or vanishes. That move is your out-of-book point. Note it and the score for the position: if the explorer shows White scoring 54% after 10 moves and you’re Black, you’re fine; if it shows 61%, you have an opening problem worth ten minutes of work.

Chess.com’s Explorer works similarly under the Openings tab, though the Masters database is less useful for club players than the rating-filtered amateur pool, because masters don’t play the moves your opponents play.

Write down the position where you left theory. You’ll want it later when you decide whether this loss belongs to your opening file or your middlegame file.

Step Seven: Decide What This Game Is For

Every reviewed game should end up in exactly one bucket, and you should pick the bucket before the engine speaks:

Opening repertoire gap. You didn’t know what to do by move 10. Fix: add a line to your Lichess Study repertoire chapter. Budget 20 minutes.

Tactical pattern miss. You missed or allowed a motif you should recognise. Fix: identify the motif by name (deflection, back-rank, knight fork on a loose piece) and set a themed puzzle filter. Lichess’s puzzle themes at lichess.org/training/themes let you drill exactly that one.

Strategic misunderstanding. Wrong plan for the structure. This is the expensive one and the one that actually raises your rating. Fix: find the structure by name (isolated queen’s pawn, Carlsbad, hanging pawns) and read the relevant chapter of something like Chess Structures by Mauricio Flores Rios.

Time or nerve failure. You knew better and didn’t do it. No engine input required. Fix: change your clock discipline, not your chess knowledge.

Most players discover, after twenty games processed this way, that 60-70% of their losses land in buckets three and four, and almost none in bucket one. That’s a training plan, and it came out of a notebook rather than a 3500-Elo oracle.

What You Take Into The Engine

At the end of the checklist you should have a single page containing: your out-of-book move number, your time-spend map with two or three fast moves circled, three written hypotheses, and a bucket. That page is your agenda. When you finally open Stockfish, you are checking specific claims at specific move numbers, not browsing.

The whole thing runs twenty-five minutes, which is roughly the length of the blitz session you’d otherwise play instead. One game processed this way beats ten games clicked through, and the difference shows up in your results inside two months.