Stuck at 1000? It's your blunder rate, not your openings

The short answer

You are stuck at 1000 because you give away material or mate in one move, two or three times a game, and the player at 1200 does it slightly less often. That is nearly the whole difference. It is not opening theory, it is not endgame technique, and it is not that you do not know the Sicilian.

This is unsatisfying because it does not give you anything to study. Studying feels like progress. Cutting your worst move per game feels like nothing, right up until your rating moves.

What 128,917 moves actually look like

Across 128,917 engine-classified moves in 4,134 games analysed on Mated, 3.8% of moves played are outright blunders and another 4.2% are mistakes. Add those together and roughly one move in 13 is a real error, before you count the 18.3% that are inaccuracies.

Do the arithmetic on your own games. One in 13 across a forty-move game is about three genuine errors per side. Three chances per game to lose a piece, hang a mate, or drop from winning to equal. Your opponent gets three too. At 1000 the games are decided by who spends theirs first and whether the other player notices.

That number is not a rating-band figure. It is every game analysed on Mated pooled together, which spans a wide range of players. We have not split it by rating, and until we do, nobody should tell you that a 1000 blunders exactly twice as often as a 1200. What the pool does tell you is the shape of the problem: at amateur level, error is frequent and it is chunky.

The errors are not spread evenly

From the same 128,917 moves, 9.0% of middlegame moves are a mistake or worse, against 2.3% in the endgame. The middlegame is roughly four times more dangerous per move than the endgame.

Do not read that as "my endgame is fine". Two things inflate it. There are fewer pieces on the board, so there are simply fewer legal ways to hang something. And plenty of games never reach an endgame at all, because someone already lost a rook on move 22 — the games that survive to an endgame are the ones where nobody had a catastrophe, which selects for cleaner play.

The honest reading is narrower: if you have limited attention, the moves that decide your games are the ones with all the pieces still on, right after the opening ends and before there is a plan. That is where you are moving fastest and thinking least.

Why your average move quality is a bad target

In that same set, the average move gives away 94 centipawns against the engine's best — but the median is far lower, because a small number of very bad moves carry most of the damage. The mean is being dragged up by a tail.

This matters for how you train. If your average loss were spread evenly across every move, you would need to get better at chess in general: better plans, better piece placement, better everything. That is a years-long project. But it is not spread evenly. Most of your moves are already fine. A handful per game are catastrophic.

So the lever is not "play better moves". It is "play fewer disasters". Those are different skills. The first is understanding. The second is a checking habit, and it is much cheaper to install.

One move, whole game

Here is the shape of it. 1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Bg5 Nbd7 5.cxd5 exd5 6.Nxd5. White has just won a pawn. The knight on f6 is pinned to the queen, so it cannot recapture. Every part of that assessment is a normal, reasonable thought.

6...Nxd5 anyway. 7.Bxd8 Bb4+ 8.Qd2 Bxd2+ 9.Kxd2 Kxd8, and Black is a knight up for a pawn. This is the Elephant Trap and it is centuries old, but that is not the point. The point is that White's position after move 5 was fine, White's position after move 6 was lost, and the other 30 moves of the game did not matter.

White's error was not tactical weakness. White saw a pin and stopped calculating. Nobody checked what happened after the recapture — specifically, whether the pin still existed once the queen came off. That is one skipped question, and it cost the game.

If you looked at White's engine report for that game, the average centipawn loss might not even look bad. Thirty-five decent moves and one that dropped 600. That is the mean-versus-median problem in a single file.

What to actually do on the board

The habit is boring and it works: before you move, look at the position your move creates, not the one in front of you. Two questions, every time, in the middlegame.

In blitz you cannot do this on every move, and pretending otherwise is useless advice. Pick the moments: the first capture, any check, any move that changes the pawn structure, and any time your opponent does something you did not expect. Those are the moves where the tail lives.

The other half is knowing which of your errors repeat. Most players have three or four recurring ones — the same undefended bishop, the same back rank, the same habit of playing a natural developing move when there is a threat on the board. Reading your own games back is how you find them, and it is the thing Mated is built to do: it pulls your recent games from Chess.com or Lichess, runs Stockfish over every position, and builds a short daily session out of the positions you actually got wrong rather than generic puzzles.

You do not need software to do this. A notebook and the analysis board on the site you already play on will find the same patterns, if you will sit down and do it after a loss instead of hitting rematch.

• What does my opponent's last move attack, and what did it uncover behind it?
• After my move, what is every capture and check my opponent has?
• Before any capture that wins material: what recaptures, and what is still pinned or defended after the dust settles?

What these numbers cannot tell you

Our figures are pooled across every analysed game, not sliced by rating, so they describe amateur chess in general rather than the 1000-to-1200 step specifically. If someone gives you a precise blunder rate for exactly 1000 and exactly 1200, ask where it came from.

Engine classification also misses the things that lose you a lot of rating points. Flagging in a winning position does not show up as a blunder. Neither does the four-game losing streak after the first loss, or the fact that you play worse at 11pm. Those are real and they are not in any centipawn figure.

And there is a genuine gap in the public evidence. Nobody has published a clean, large study of what changes in a player's moves as they go from 1000 to 1200 — how much is fewer blunders, how much is better time use, how much is just having played more games and seen more patterns. The distribution above is consistent with "mostly fewer blunders". It does not prove it.