Rank your chess mistakes by cost, not embarrassment

The ranking

Ordered by what it costs you in rating over a hundred games, not by how much you wince when you see it in the analysis tab. This is my ranking based on what shows up in game data and in coaching; treat the order of the middle entries as arguable.

The gap between the top of this list and the bottom is enormous. Most players spend their study time on entries 5 to 7.

• Dropping material to a one-move tactic in the middlegame. Nothing else is close.
• Missing what your opponent's last move threatened because you were busy with your own plan. Usually the same mistake as (1), one move earlier.
• Spending clock on moves that were never critical, then moving in four seconds when the position actually branched.
• Starting the next game within a minute of losing the last one.
• Volunteering complications in a position you only needed to hold.
• Getting slowly outplayed positionally — real, but it costs you half a point at a time, not a piece.
• Not knowing the sixth move of your opponent's sideline.

One move in 13 is a real error

Across 225,521 engine-classified moves from 7,134 analysed games on Mated, 3.7% of moves played are outright blunders and another 4.2% are mistakes. That is roughly one move in 13 being a genuine error, before you count inaccuracies, which run at 18.4%.

Read that the other way round. In a 40-move game you are making around three real errors. You probably remember one of them. The other two got away with it because your opponent was making three of their own.

This is why "I only blundered once" is usually false. You blundered once *that was punished*.

Where the errors actually are

Same dataset: 9.0% of middlegame moves are a mistake or worse, against 2.3% in the endgame. Errors are not spread evenly across a game. They cluster in the phase where the position is complicated and your plan is only half-formed.

I want to be careful about what this does and does not mean. It does not mean endgames are easy. It means that at this level, by the time you reach an endgame there are fewer pieces to hang, fewer forcing lines to miscalculate, and many games never get there at all. The error rate is low because the moves are simpler, not because you are good at them.

But the study implication is real. If you are choosing between an hour on rook endgames and an hour on middlegame tactics from your own games, the second hour is working on four times the error rate. Rook endgames will matter more at 2000 than they do now.

A few moves do most of the damage

The average move in that dataset gives away 93 centipawns against the engine's best. The median is far lower. A small number of very bad moves carry most of the total loss.

That asymmetry should change what you work on. You cannot meaningfully improve your average move — the median is already close to fine. You can remove your worst three moves per game, and removing them is worth more than sharpening the other thirty-seven.

Concretely: you are up a piece, you have a plan, you play the natural developing move, and it loses the exchange to a knight fork you would have found in a puzzle in two seconds. That single move erased forty good ones. Post-game, it registers as one blunder among many. In centipawns it was the game.

This is also why chasing your inaccuracy count is a poor use of an evening. 18.4% of moves are inaccuracies. Fixing them individually is slow work with a small payoff each. The 3.7% is where the rating lives.

The ones that feel worse than they cost

Losing to a cheap trick you had seen before. It costs exactly the same as any other lost game. It just has your name on it in a way that a slow positional squeeze does not.

Getting out-prepared in the opening. At 800 to 1800 the position after move 12 is almost never the reason you lost. Look at where your losing move actually came — usually twenty moves later, in a position neither player had prepared.

Losing to someone rated 300 points below you. Rating points are symmetric; your feelings about them are not. The value of that game is the same as any other, and treating it as a catastrophe is what produces mistake (4) on the list above.

Opinion, but strongly held: the reason players study openings is that opening mistakes are legible. You can name them, look them up, and fix them. Middlegame blunders are just "I didn't see it", which feels unfixable, so people avoid working on it.

The mistake in how you train

Across 2,099 rated puzzle attempts on Mated, 94.2% were solved first time. That is a problem with the puzzle set, not a compliment to the solvers. A set you pass 94 times out of 100 is testing patterns you already own.

Puzzles that feel comfortable are recall. Puzzles that make you sit for ninety seconds and still get it wrong half the time are the ones adding anything. If your puzzle rating has been flat for six months and you are still solving almost everything on the first try, the difficulty is set too low.

There is a caveat. Fast, easy puzzles do have a use: keeping basic motif recognition quick enough to fire during a blitz game. But that is maintenance. Do not confuse it with training.

Find your own ranking in twenty minutes

Do not take my ordering. Take your last twenty losses and build one from your own games. This takes about twenty minutes and needs nothing you do not already have.

Open each loss. Find the move where the evaluation swung by more than about 200 centipawns in your opponent's favour — there is usually exactly one, sometimes two. Write down three things: the move number, whether you were better, equal or worse before it, and one word for the cause. My four buckets are *hung* (one-move loss of material), *missed* (their threat), *calculated* (I looked and got the line wrong), and *clock* (under ten seconds).

Twenty rows takes twenty minutes. Then count. If fourteen of them say *missed*, you have your answer and it is not an opening problem. If eight say *clock*, your problem is time allocation and no amount of tactics will touch it.

The count matters more than any individual game, because individual games lie. The one you remember is the one that hurt, and the one that hurt is rarely the one that repeats.