What to do after losing a chess game: a review routine
The hour after a loss is the worst hour to analyse in and the best hour to collect in. Save the game, write down what you thought happened, then stop. The analysis is a separate job on a calmer day, and it is short: two positions, three questions, one sentence you can carry into the next game.
The first twenty minutes
What to do after losing a chess game, in the first twenty minutes, is almost nothing. Do not queue for the next game, and do not start analysing. A rematch played while annoyed is the single most reliable way to turn one loss into four, and analysis done while annoyed finds the move you are angry about rather than the move that decided the game.
There is one thing worth doing immediately, because it expires. Write two or three sentences about where you thought the game turned, what you were trying to do, and the move you already suspect was wrong. That account is only available now, it is the only record of what you believed at the board, and comparing it later with what actually happened is more informative than any single best move.
Save the game while it exists
Whatever you use, make sure the game is somewhere you can find it in a week. An online game is in your history; a club game exists only in your scoresheet until you type it up, and a scoresheet you did not type up is a game you did not play as far as your review is concerned.
In Mikhail either route works. Import your recent games by giving your Chess.com or Lichess username, or paste a single PGN from anywhere for one review. A pasted PGN is analysed and not kept; stored game history starts only once you turn game storage on in Settings, and turning it off stops it. Nothing about a review requires you to store the game, so the decision is yours and reversible.
The routine, on a calmer day
Read your own account first, before any engine. Then have the engine find the moves where the assessment actually jumped, and go to the first one. At that position: what did I think was happening, what was actually happening, and what would I have needed to notice in time? Then do the second one. Then stop - two positions properly examined beat sixty moves skimmed, and a review that takes an hour usually means the whole game got the attention the turning points needed.
Add one small step people skip: look up the opening you actually played and the exact move where the game left it. Mikhail's review names the line, its ECO code and the move that left the book, including whether you or your opponent left first. It is a fact about your preparation rather than about the loss, and it is usually earlier than you expect.
The loss you know about and the one you do not
Most players can name the mistake before opening anything, and are frequently right about the move and wrong about the game. The move you remember is the one that hurt. The move that decided it is often eight plies earlier and felt like nothing - a plan chosen, the queenside stopped being looked at, a piece left on a square that stopped being safe.
This is why the two accounts are worth writing down separately. Where they agree, you have a mistake you can already see coming, which is a good sign. Where they disagree, you have found something more valuable: a blind spot in your sense of your own games, which no amount of solving puzzles will show you.
When a loss is not worth reviewing, and when to stop
Some losses carry no information. You were flagged in a won position, or you played at one in the morning, or the game lasted eleven moves because you forgot a variation you never understood. Note the cause in one line and move on; there is nothing in the moves to learn from, and grinding through them is a way of feeling diligent rather than being it.
End every review that is worth doing with one sentence in the form "when X happens, I will do Y", and keep the sentences together. When the same sentence appears three times you have found a habit rather than an accident, and that is the thing to train. Mikhail does the same arithmetic from the other direction: the coach report ranks counted errors and stays silent below two engine-reviewed games, because one loss is not a diagnosis of anything.
Questions
Should I review a loss on the same day?
Collect on the same day, analyse on another. Write your account of the game while you still remember what you were thinking, and do the engine work when the result has stopped mattering to you - usually the next day.
Is it worth reviewing blitz losses?
In batches, not one at a time. A single blitz loss is mostly noise, but twenty of them reviewed together show you which mistakes survive at speed, and those are the ones that are really automatic. Treat the batch as the unit of analysis.
What about the games I won?
Review the ones you nearly lost. A win in which you were worse for fifteen moves carries more information than a clean win, and reviewing only losses gives you a picture of how you lose rather than of how you play.
How do I stop tilting after a loss?
Decide the rule before you sit down, because you will not decide it well afterwards: a fixed number of games, or a stop after two losses in a row. This is a scheduling problem rather than a chess problem, and nothing in a game review can solve it for you.
Try it on your games
The parts of Mikhail this guide describes are Game review and Coach report, which need a free account — your games and your counts have to live somewhere.