Where your blitz clock actually goes

Most of your clock goes to about six moves

In a 3+0 game, the bulk of your time disappears into five or six decisions, nearly all of them in the first half of the middlegame. The rest of your moves take one to three seconds each, and they are fine. Time management in blitz is not about moving faster overall. It is about which six moves get the money.

The failure pattern is almost always the same. You spend 18 seconds on move 14 choosing between two moves that are worth the same, arrive at move 28 with 22 seconds, and hand back the game in a position you would have solved in eight seconds an hour earlier. The seconds were not wasted because you thought. They were wasted because you thought in a position where thinking did not change the move.

Do the arithmetic once and it stops being abstract

Take a 3+0 game that lasts 40 of your moves. That is 180 seconds for 40 decisions: 4.5 seconds a move if you spread it flat. Nobody spreads it flat, and you shouldn't.

Here is a realistic shape. Ten opening moves at 1.5 seconds each: 15 seconds gone, 165 left. Two long thinks of 20 seconds in the early middlegame: 125 left for 28 moves, which is still 4.4 seconds a move, so you are fine. Now add the third and fourth long think, the ones where you were choosing between Rfd1 and Rad1, or checking for the fourth time that the knight fork does not work. That is another 35 seconds. You now have 90 seconds for 26 moves, which is 3.4 a move, and the position is getting sharper rather than simpler.

The point of the arithmetic is that the average is never the problem. Two extra 18-second thinks are the problem. You cannot feel a 4.4 second average turning into 3.4, but you can feel the last ten moves, and by then you are playing on instinct in exactly the positions that punish it.

The test for whether a long think bought anything

After a long think, one of three things is true. You changed your move, and the move you changed to was better. You changed your move, and it wasn't better. Or you played the move you were going to play anyway.

Only the first case bought you something. The third case is the common one, and it has a tell: you spent the time confirming rather than comparing. Re-checking a tactic you already checked is confirmation. Counting the attackers on d5 a second time is confirmation. If you notice yourself verifying, move.

The positions where thinking genuinely changes your move share features you can spot in under a second:

• Your opponent played something you had not considered at all.
• Material is about to change hands, and there is more than one way to let it.
• There is a check or a capture available to either side that you have not accounted for.
• A piece of yours is undefended and a piece of theirs just got closer to it.
• The pawn structure is about to become permanent.

What the error data says about where to spend

Across 180,442 engine-classified moves from 5,773 analysed games on Mated, 3.8% of moves played are outright blunders and another 4.2% are mistakes. That is roughly one move in 12 being a real error, before you count the 18.5% that are inaccuracies. So you have about three genuine errors per game to prevent, not thirty.

Those errors are not spread evenly. In the same set, 9.1% of middlegame moves are a mistake or worse, against 2.3% in the endgame. The middlegame is roughly four times as dangerous per move. If your clock graph shows a long think on move 34 of a rook endgame while moves 15 to 22 went at two seconds each, you spent your money in the safe part of the game.

The third number is the one that should change how you think about a long think. The average move in that set gives away 95 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. Your clock is not there to squeeze five centipawns out of an already-fine move. It is there to catch the one move that gives away a piece. Those two jobs feel identical from the inside and have completely different payoffs.

Audit your last three losses in ten minutes

Both sites will show you your move times. On Lichess, the analysis board of any game shows the clock at each move and a time graph. On Chess.com, the review screen shows time spent per move. You do not need software for this part.

Open three recent blitz losses. For each one, write down your five longest thinks with the move number. Then label each one: forced (there was only one sensible move), near-equal (you were choosing between moves of similar value), or critical (a check, capture, or threat you had genuinely missed). Add up the seconds in the first two buckets.

In my own games and in games I have looked at with students, the forced and near-equal buckets are usually the bigger half — that is an impression, not a measurement, so check it on yourself rather than taking my word. The useful output is not the total. It is noticing the type of position that eats your clock, because it repeats. Some people burn time on recaptures. Some burn it on rook placement. Some burn it when their opponent's rating is higher.

If you want the other half of the audit — which of those positions were actually the ones where you went wrong — that is what Mated does with your imported games: Stockfish over every position, then a fifteen-minute session built from the positions where you lost material. But the clock half you can do yourself tonight with a notepad.

Seconds that reliably buy nothing

Some spends are almost never worth it at blitz speed. Thinking in an opening you half-remember: you either know move 9 or you don't, and 15 seconds will not retrieve it. Thinking about which of two winning endgames to enter: pick one and play it. Thinking in a lost position about whether it is lost: it is, play fast and make them prove it.

One spend is worth defending. When your opponent plays a move you did not consider, stop, even if the move looks harmless. That is the single cheapest moment to catch an error, because you are already surprised, and surprise is information.

What nobody has measured well

I do not know what fraction of blitz losses in the 800–1800 band are decided by the flag versus decided on the board. It would be measurable from public game data, and I have not seen it done properly by rating band and time control. Be suspicious of anyone who quotes you a number for it.

I also cannot tell you whether 3+0, 3+2, or 5+0 improves you fastest. The increment obviously changes the shape of the endgame and lets you think when you are low, and plenty of strong players recommend it for that reason. Whether it transfers to your slow games is an open question rather than a settled one.

Pay before the fork, not after

The cheapest second in blitz is the one you spend on your opponent's clock. Not calculating — deciding. While they are thinking, pick the plan: I am playing on the queenside, my rook is going to d1, I will trade the bishop if offered. Then when it is your move you are executing a decision instead of making one.

This is also the argument for repertoire depth over repertoire width. Not because theory wins games at 1400, but because a middlegame you have seen forty times costs you three seconds a move instead of eight, and those five seconds are the ones you will want on move 30. Narrow your openings and you are buying clock, not knowledge.