The things that actually move your rating, ranked

The ranking, before anything else

Between 800 and 1800, the single biggest lever on your rating is not losing pieces to one-move tactics, and the second is giving yourself enough time on the clock to notice them. Everything else — openings, endgame technique, positional understanding, calculation depth — sits below those two and stays there until they are handled.

Here is the full order, as I would rank it. Items 1 and 2 are supported by where errors actually appear in games. Items 3 to 6 are my ordering and I will say so as I go.

What the move data says

Across 117,841 engine-classified moves from 3,781 analysed games on Mated, 3.7% of moves played are outright blunders and another 4.2% are mistakes. That is roughly one real error every 13 moves, before you count inaccuracies, which run at 18.2%.

Those errors are not spread evenly. In the same set of 117,841 moves, 8.9% of middlegame moves are a mistake or worse, against 2.2% in the endgame. The middlegame is where your rating is decided, by a wide margin.

One more number worth sitting with: the average move gives away 94 centipawns against the engine's best, but the median is far lower. A small number of very bad moves carry most of the damage. Your play is not uniformly leaky. It is mostly fine, punctuated.

That last point is what makes the ranking obvious. If your losses came from a steady 30-centipawn drip across every move, you would need to get broadly better at chess. They do not. They come from a handful of moves per game, and those moves are findable.

1. Stop the one-move losses

Not tactics training. Not pattern recognition. Specifically: a habit that runs before your hand moves the piece.

The habit is two questions, and it takes about four seconds:

Most players at this level know these questions. They do not ask them, because the position looks calm and the move looks obvious. The errors cluster in the middlegame — 8.9% of moves there are a mistake or worse in that same 117,841-move sample — precisely because the middlegame is where positions look calm and are not.

• What did my opponent's last move change? What is now attacked, what square just opened, what piece stopped defending something.
• After my intended move, what is undefended? Check every piece of yours that is not currently protected by another piece.

2. Play the time control that lets you use question one

If you play three-minute blitz almost exclusively and your rating has not moved in six months, this is very likely why. You cannot run a four-second check on move 22 with 40 seconds left.

I do not have a measurement that says rapid raises your rating faster than blitz, and I have not seen one I trust. Nobody has measured this well, partly because the players who choose long controls differ from the players who choose short ones in a dozen other ways. So treat this as opinion: play 10+0 or 15+10 as your main control, and use blitz when you want to play chess rather than when you want to improve. The mechanism is at least clear — the habit in item 1 needs time to exist.

The related change is playing fewer games. Three rapid games you look at afterwards beat fifteen you do not. Again, opinion, but the alternative claim — that volume alone does it — is the one that gets repeated without support.

3 to 6, my ordering

Below the top two, the ordering is a judgement call. This is mine.

3. Converting won positions. You reach +4 and draw or lose. This is partly technique and mostly item 1 again, under the specific pressure of a position you feel you should not be able to lose.

4. Middlegame plans in your own structures. Not general strategy — the three or four pawn structures your openings actually produce, and what the pieces do in each. This has real weight and gets less attention than opening move orders that lead to it.

5. Basic endgames. King and pawn, rook and pawn, the Lucena and Philidor. Worth knowing cold. But endgame moves are the cleanest part of most games in the data — 2.2% mistake-or-worse — which suggests it is not where the points are leaking for most players at this level.

6. Opening theory past move eight. This has more content written about it than everything above combined, and it is last. You need a first move, a reply to the four things you actually face, and to know why the moves are played. Twenty moves of a Najdorf line you meet twice a year is not improvement work.

A single move, worked through

Say you play a 15+10 game. You get a decent Italian, you are up a pawn by move 20, and the engine later says your first nineteen moves averaged about 30 centipawns of loss — perfectly reasonable club play. On move 24 your opponent plays a rook to the b-file. It looks like a slow improving move. You continue with your plan on the kingside. The rook was attacking your b7 knight through a pawn that had moved two turns earlier, and you lose it.

That one move costs around 300 centipawns. It undoes the other nineteen combined and then some. Nothing you learn about the Italian changes it. The only thing that changes it is asking what their rook move touched.

This is the shape of most losses in this rating band, in my reading of the games I see. Not a slow strategic drift. One or two moves where a check did not happen.

How to turn the ranking into a week

Play four rapid games. After each, spend five minutes finding the two or three moves where the evaluation actually swung — not every inaccuracy, just the swings. Write down what you failed to look at. Not "I need to calculate better" but "I did not check the diagonal after their bishop moved." After ten games you will have a list with two or three items on it repeating. That list is your training plan.

You do not need software for this. A board, the game, and your own honesty will do it, and the effort of finding the swings yourself is part of the value. What software helps with is the second week, when the review starts feeling like a chore — Mated does the engine pass in the browser, sorts your errors by category, and hands back a fifteen-minute session built from your own positions, which mainly buys you the fact that it still happens on a Thursday when you cannot be bothered.

Either way, the object is the same: a specific, written list of the checks you skip.

The ranking changes, and you will notice when

This ordering is for 800 to 1800. It stops being true somewhere above that. When your blunder rate drops and your games start being decided by whether you understood the position rather than whether you saw the fork, items 4 and 6 climb and item 1 becomes maintenance.

You will know the switch has happened when your losses stop having an obvious moment. When you can look at a loss and not find the move — when the game just got gradually worse and you cannot point at where — that is the signal that the top of this list has been paid off and you need a different one. Until then, if you can find the move, you are not yet in the market for a better opening.