Chess

You hang a knight on move twelve. Then the evening dissolves into chasing opening theory and distractions, not fixing the simple glance that would have saved the game.

About 20% of the positions create 80% of the losses. The other 80% is quiet moves you already play fine. That 20% and 80% is a rule of thumb. The counts further down are rougher.

Name the few positions that actually lose the point. Drill those. Leave the opening encyclopedia until the tags shrink.

Five moments that decide most losses

1. A piece left hanging.
2. A one-move tactic: check, capture, or threat.
3. A loud position you played like a quiet one.
4. A first move you meet every week and still do not know.
5. A rook ending you postponed for a queen study.

Look once before you let go

Club games do not fall apart because the plan was subtle. They fall apart because a piece was loose and nobody looked. The work that fills the night is opening videos and long study sessions that skip the simplest glance at hanging pieces.

A hanging piece is a simple object. You moved it to a square the opponent attacks, and nothing of yours defends that square. Or you left it there while they added an attacker. Masters miss this less often because the glance is automatic. Club players still lose pieces they would have seen in a puzzle with a caption.

Drill: before your hand leaves the piece, say which enemy units attack that square. If you cannot name a defender, and an attacker exists, do not play it. Use the same glance in puzzles and in games. It takes one second. If you skip it, the rest of the study does not matter that night.

Checks, then captures, then threats

The next pile is short tactics. Forks, pins, skewers, discoveries. Not a twenty-move combination you will not see on a Tuesday. The work that feels like training is a puzzle streak you never review, then a skip past the simple one because it looked too easy.

Most of those simple ones are checks, captures, or threats. A quiet move can be right. It is rarely the first thing to search. If you search the quiet moves first, you will find a plan and miss the fork.

Drill: on each puzzle, say the checks out loud, then the captures, then the threats. Only then look for a quiet move. Give it one minute. If the tactic was a check you did not say, that is the miss. Stop the session when you can do ten in a row without skipping the list.

Spend the clock where moves lose

Ashton Anderson, Jon Kleinberg, and Sendhil Mullainathan went through millions of recorded games. The paper is in ACM Transactions on Knowledge Discovery from Data, 2017 (arXiv:1606.04956). They studied positions a computer has fully solved. Mostly spare endgames, not a messy middlegame with queens still on.

In those spots, the danger in the position predicted a blunder better than the player's rating, and better than the clock. They built a balanced set of 600,000 decisions, half blunders and half not. Features that describe how dangerous the position is were right about 75% of the time. Features that describe the player's skill were right about 55%. Time left on the clock was about 53%. Skill and time added nothing once the danger features were already in the model.

Their danger measure is plain. If you picked a legal move at random, how often would it be a blunder? A position full of losing moves is dangerous even when you are the higher-rated player. On the Free Internet Chess Server, for ratings from 1200 to 1800, a step of 0.2 on that scale changed the blunder rate more than 600 rating points. For players rated 1500 to 1599, eight to twelve seconds left in a safer position beat more than a minute left in a position 0.2 more dangerous.

So the clock is not the first villain. The position is. A quiet move in the opening does not deserve the minute you just spent. The position with three checks and two captures does. This is a decision-making result, not a claim that rating is fake. Rating still matters. It matters less than whether the board is full of losing moves. Same idea as concentration: the minute goes where the board is loud.

Drill: in your next rapid game, find one moment where a check or a capture was legal and you played a quiet move instead. Next game, that moment gets an extra minute. Take the minute from move six of theory, not from the ending.

Two first moves, then stop

White has 20 legal first moves. In the New In Chess Yearbook for 2000, two of them carried most of a large database. Of 731,740 games, 349,855 began with 1.e4 and 296,200 began with 1.d4 (Sosonko and van der Sterren, Yearbook 55, p. 227). That is about 88% of the games, from 2 of the 20 legal first moves. The yearbook is master-leaning, and it is a quarter-century old. Your Tuesday rapid pool will not match it move for move. The shape is the point. A short list of first moves is the game people actually play.

80/20 example: two first moves, about 88% of that database. A tidy split would have given the other eighteen moves more of the games. The yearbook did not.

The work that feels serious is fifteen moves of a line you meet twice a year. It is the studying trap in a smaller room. Every chapter, instead of the chapter that shows up.

Drill: write one first move as White, and two answers as Black to 1.e4 and to 1.d4. Stop at move eight. Play them until move four is automatic. Add a line only after the tag log says openings, not hanging pieces, are what you are losing.

Rook endings come up

Karsten Müller and Frank Lamprecht, in Fundamental Chess Endings (2001), put rook versus rook at 8.45% of games. That was the most common material ending on their list. Queen versus queen was 1.87% on the same list (the table, cited to pp. 11-12 and 304). A wider description, endings with rooks and any number of pawns, is put at about 8 to 10% of all games, including games that never reach an ending (rook and pawn versus rook). Neither number is an 80/20. Both say the same thing. If you study one ending family, it is rooks.

The work that looks like mastery is a beautiful queen-versus-rook study you will not reach. Or an equal tour of forty ending types, one page each, so the common one never sticks.

Drill: Philidor and Lucena, ten minutes. Philidor holds the draw a pawn down. Lucena wins with the extra pawn. Then one rook-and-pawn idea: keep your king in front of a passed pawn, or cut the enemy king off with the rook. Stop there for the month unless your tags say otherwise.

The week you can stop filling

Tag ten games

After each game, write one line. If you lost, tag the move that decided it. If you drew or won and still got lucky, tag the move you should not have survived. Five tags only.

• Hanging piece
• Missed short tactic
• Rushed a loud position
• Opening you did not know
• Rook ending

Illustrative: a club player tagged four hanging pieces, three rushed positions, two opening holes, and one rook ending across ten games. The next week was only the one-second glance. The opening book stayed shut.

8020 move: when the ten lines are in, drill only the tag with the highest count. Leave the other four until that count drops.

Misreads that waste a month

"I need more time on the clock." In the solved positions Anderson and colleagues studied, a slightly more dangerous spot outweighed a large gift of time. Add time where the board is loud. Do not add it on every move.

"I need a bigger opening book." Two first moves were about 88% of the 2000 yearbook. A new chapter does not fix a knight you left loose.

"Every ending deserves a week." Rook versus rook was several times more common than queen versus queen on the Müller and Lamprecht list. Start with the rooks.

The next game

You do not need a new personality, or a repertoire that runs to move twenty. You need the glance, the short tactic list, and the minute spent where many moves lose. If the night is a club board game and not a title shot, that is the whole practice plan.

Tag ten games. Then practice the tag that actually showed up. The encyclopedia can wait until hanging pieces are no longer the top line.