Chess strategy

Chess strategy in five numbers from 89.4 million games

We read every rated Lichess game from September 2026 (89,396,008 games, players grouped by rating) and counted what castling, queen trades and material leads do to the result. Each one is charted by rating further down the page.

• 51.8% vs 49.8%White scored 51.8% when it castled by move 10 and 49.8% when it never castled, across all ratings. Stronger players do both, so this is an association.
• 61% and 77%A player two points of material ahead after move 30 won 61.1% of games; five or more points ahead won 77.2%.
• 3.8% to 7.1%Draws were 3.8% of games with queens on at move 20 and 7.1% of games with queens off.
• 2.17 to 1.36Blunders per player per game, from 2.17 at 1000-1199 down to 1.36 at 2200+, on games Lichess analysed.
• +2.2 pointsSide with the bishop pair at move 20, material otherwise equal, scored 52.2% at 1000-1199 and 53.2% at 2200+.

Try the strategy ideas on a live board

Each position asks for one move. Play it, then read why it works and what Stockfish thinks. Positions 1 to 5 are the five plans, 6 is the check for a tactic that comes before any plan, and 7 is the six-move win that people search for and what it costs.

1. 1. Improve your worst piece

   Black to move. The knight on a5 is on the edge. Bring it back, from a5 to c6.

   Nc6: legal.

   From c6 the knight presses d4 and guards e5 instead of looking at the edge. Stockfish picks Nc6 first, and the score stays level for White: +0.39 before and +0.44 after.

   Try it on the board

2. 2. Rooks on open files

   The d-file has no pawns on it and Black has no rook there yet. Put a rook on it, from f1 to d1.

   Rfd1: legal.

   The rook now eyes d7 and d8. The position stays level, White +0.01 before and -0.01 after, because Black can answer with Rcd8. Stockfish agrees with Rfd1; the file is a way in, not a win.

   Try it on the board

3. 3. Pawn structure: stop the isolated pawn

   Black's pawn on d5 has no pawn on the c- or e-file next to it. Put a knight on the square in front of it, from f3 to d4.

   Nd4: legal.

   No Black pawn can attack d4, so a piece there cannot be pushed away by a pawn, and the d5 pawn cannot advance. Stockfish picks Nd4 first: White +0.39 before and +0.41 after.

   Try it on the board

4. 4. King safety: castle early

   White to move. Both centre pawns are in place and the e-file could open soon. Castle kingside, from e1 to g1.

   O-O: legal.

   The king goes behind the f2, g2 and h2 pawns and the rook leaves the corner. Stockfish picks castling first: White +0.37 before and +0.35 after, so it costs nothing.

   Try it on the board

5. 5. Trade when ahead

   White has an extra knight. Black's rook on d8 is guarded by the king. Swap rooks, from d1 to d8.

   Rxd8+: legal.

   After Rxd8+ Kxd8 White is still a knight up with fewer pieces left to defend against. The score barely moves, White +4.67 before and +4.72 after. Stockfish's first pick is the quiet Kf1, so the trade is not the only good move.

   Try it on the board

6. 6. Check for a tactic before the plan

   White to move. Black's king and rook stand a knight's jump apart. Play the knight, from d5 to c7.

   Nc7+: legal.

   Nc7+ attacks the king on e8 and the rook on a8, so White wins the rook next move. Stockfish picks it first: White +7.78 before and +7.92 after. A slow plan would have missed it.

   Try it on the board

7. 7. The six-move win, if Black cooperates

   Black has just played Ngf6 in the line shown. The queen on e2 pins the e7 pawn to its king, and the knight on d7 shuts in Black's queen. Play the knight, from e4 to d6.

   Nd6#: legal.

   Nd6 is checkmate: the king on e8 has no square and nothing can capture the knight. Stockfish 17.1 at depth 16 finds the mate in one. It works only because Black played Ngf6 on move five.

   Try it on the board

Should you castle early? What 89 million games show by rating

White scored 51.78% in games where it castled by move 10 and 49.79% in games where it never castled (89,396,008 Lichess games, September 2026).

At 1000-1199 only 40.3% of White players castled by move 10 and 29.1% never castled at all. By 2200+ the figures were 58.1% and 12.2%. The gap in score between the two groups was 1.2 points at 1000-1199 and 2.3 points at 2200+.

These tables show an association. Stronger players castle earlier and also win more, so the gap cannot be read as what castling alone is worth. The part that is not in doubt: players who never castled scored lowest at every rating, 50.59% at 1000-1199 against 51.8% for those who castled by move 10. The rule for how it works is on the castling page, and position 4 above lets you try it.

How often does a material lead win? By size of lead and rating

A player two points of material ahead after move 30 won 61.1% of games across all ratings; a lead of five or more points won 77.2% (89,396,008 Lichess games, September 2026).

One pawn ahead is a small edge at any rating: 52.0% of games won at 1000-1199 and 54.6% at 1800-1999. Two points lifted that to 58.08% and 61.55%, and three to four points to 65.25% and 67.41%.

The odd result is at the top of the table. With five or more points ahead, 1000-1199 players won 78.5% and 2200+ players won 74.4%. Draws are 6.55% of games at 2200+ against 3.13% at 1000-1199, which is one reason a big lead converts less often at the top. When you are ahead at club level, the way to cash it in is to trade pieces and keep your king safe, which is the plan in position 5.

Do trading queens lead to draws? Draw rate with queens on and off

Games with queens off the board at move 20 ended in a draw 7.05% of the time; games with queens still on ended in a draw 3.81% of the time (89,396,008 Lichess games, September 2026).

Queens were off by move 20 in 27.8% of games at 1000-1199 and in 18.4% to 19.1% of games at 2000 and above. At every rating the draw rate with queens off was 1.7 to 2.2 times the draw rate with queens on: 6.65% against 3.05% at 1000-1199, 10.64% against 6.24% at 2200+.

If you need a win, a queen trade has a price: it roughly doubles the chance of a draw. Keep the queens on and look for attacks on the king. If you are ahead, count that price before you offer the trade.

Is the bishop pair worth it? Score with two bishops against none

With material equal at move 20, the side that had two bishops against bishop and knight, or two knights, scored 52.85% across all ratings (89,396,008 Lichess games, September 2026).

The edge is real and small: 2.2 points above an even score at 1000-1199, rising to 3.2 points at 2200+. The data do not say why the edge grows with rating. The bishop pair is one of the positional advantages Steinitz's school put first, described below.

How many blunders per game? Lichess's own count at every rating

Lichess marked 1.83 blunders per player per game on average, from 2.17 at 1000-1199 to 1.36 at 2200+ (8,896,940 computer-analysed games, September 2026).

Only 4.5% of 1000-1199 games had no blunder at all, against 14.8% at 2200+. A player at 1200 makes about two blunders a game, and 95.5% of 1000-1199 games had at least one. That is why the three questions in the section below ask about forcing moves and your opponent's threats before they ask about plans.

Chess strategy for beginners: choose a plan

In the opening, develop your knights and bishops, fight for d4, e4, d5 and e5, and get your king safe. In the middlegame, use the ideas below to improve your position. In a safe endgame, bring your king toward the action and support a passed pawn.

The numbers above back four of the rows: king safety is in the castling charts, trading when ahead is in the material lead chart, the tactics check is in the blunder chart. Pawn weaknesses and rooks are in the sections below.

Chess strategy by rating: what to work on at 1000, 1500 and 2000

The numbers above point to a different first job at each level.

1. Beginner, 1000 to 1399. Castle early and check what your opponent threatens before every move. At 1000-1199, 29.1% of White players never castled and 12.1 games in 10,000 ended in Scholar's Mate (Lichess, September 2026), so king safety and one blunder check do more than any plan. Learn the piece values so you know what a trade costs.
2. Intermediate, 1400 to 1799. Work on trades and structure. A lead of two points won 60.9% to 61.4% of games here, so learn to convert it: swap pieces, keep your king covered and aim for a passed pawn. Add the worst-piece question and the open file from the sections below.
3. Advanced, 1800 and above. Small edges decide games. The bishop pair scored 52.7% to 53.2%, queens-off games drew 7.4% to 10.6% of the time, and a five-point lead won only 74% to 76% of games. Study endgames and pawn structure, and decide early whether you want to keep or trade queens.

Chess strategy vs tactics: which one is this position asking for?

Strategy is the plan, tactics is the forcing line. Run the same three questions on every move.

1. Look for forcing moves first. Checks, captures and moves that attack something valuable. If one wins material by force, play it. That is position 6.
2. Check what your opponent threatens. A slow plan fails when the opponent has a threat you did not see. Answer it before you improve anything.
3. Only then pick the plan. No forcing move, no threat against you: improve the worst piece, take the open file or fix a pawn weakness.

Wikipedia's article on chess strategy defines strategy as "setting and achieving long-term goals during the game" and tactics as "move-by-move threats and defenses". Emanuel Lasker described the approach Wilhelm Steinitz founded as aiming for small advantages, accumulating them and only then looking for a combination (Wikipedia, Wilhelm Steinitz, section Assessment). Richard Teichmann is credited with the line that chess is 99% tactics, which Erwin Voellmy recorded in 1930 as a slight exaggeration (Wikipedia, Richard Teichmann).

Improve your worst piece: the question to ask when nothing is forced

When no tactic is on the board, ask which of your pieces does the least, and give it a better square.

1. Find it. Look for the piece with the fewest useful moves: a knight on the edge, a bishop behind its own pawns, a rook with no open file.
2. Pick the square. Decide where it would attack something or guard something. Central squares and outposts are the usual answers.
3. Plan the route. A knight may need two moves to get there. Check that each step does not hang anything.
4. Trade it if it cannot improve. A piece stuck behind its own pawns can be swapped for an opponent's active piece.

The principle of the worst piece is simple: improve the piece doing the least before starting another attack. Positional chess means improving lasting features such as piece activity, pawn structure and weak squares; position 1 shows this with ...Nc6.

Rooks on open files: why the file matters

An open file is a file with no pawns of either colour on it, and a rook there reaches the other side of the board.

Wikipedia's article on the open file says having rooks or queens on open or half-open files is considered advantageous, because they can attack more easily down the file into the opponent's position. It credits Aron Nimzowitsch with first recognising the power of a major piece on an open file in My System, where he names the goal as "the eventual occupation of the 7th or 8th rank".

Position 2 is level on Stockfish's scale. A rook on an open file gives you a way in, and your opponent can answer by putting a rook on the same file, so take it first and then look for the seventh rank.

Pawn structure and king safety: the two plans that last

Pawns cannot move backward, so a pawn move can leave a lasting weakness.

1. Pawn weaknesses are permanent. Wikipedia's chess strategy article lists isolated, doubled and backward pawns as weaknesses that, once created, are usually permanent. A piece in front of an isolated pawn, as in position 3, stops it advancing.
2. Keep the king sheltered while queens and attacking pieces remain; activate it in an endgame when it is safe.. The same article says the king is often best protected in a corner behind its pawns, and that castling often connects the two rooks. Castle before the centre opens.
3. Do not castle into an attack. If the opponent's pieces already aim at your castled position, a different plan may be better. Check first, as in the tactics steps above.

Trade pieces when you are ahead: how to cash in a lead

With more material, every swap leaves your extra piece worth more of what remains.

Chess.com's tactics glossary describes the aim in its entry on simplification: "A player tries to simplify a winning position as an act of good technique, which is the skill of converting an advantage into a victory." In position 5, White is a knight up and swaps rooks, so fewer pieces remain to attack White's king and the extra knight counts for more.

Trade pieces of equal value, not a rook for a bishop. If you are behind, do the opposite and keep the pieces on. Stockfish's own first choice in position 5 is the quiet Kf1, so a trade is one good way to cash in a lead, not the only one.

What strong players say about chess strategy

Five ideas from the players who wrote the rules of positional chess, each with the page we read it on.

• Lasker summed up Steinitz's method as aiming for small advantages and accumulating them before looking for a combination. Steinitz's later play gave priority to pawn structure, space, outposts for knights and the bishop pair.

  Emanuel Lasker on Wilhelm Steinitz, Wikipedia, Wilhelm Steinitz (paraphrased)

• Wikipedia's chess strategy article says to defend your pieces even when they are not threatened, and links the habit to Nimzowitsch, who called it overprotection.

  Aron Nimzowitsch, Wikipedia, Chess strategy (paraphrased)

• “He who fears an isolated Queen's Pawn should give up Chess.”

  Siegbert Tarrasch, Wikiquote, Siegbert Tarrasch

• “The rooks belong behind passed pawns”

  Siegbert Tarrasch, St Petersburg 1914 tournament book, p. 83, Wikiquote, Siegbert Tarrasch

• “you must study the endgame before anything else”

  José Raúl Capablanca, Last Chess Lectures (1966), p. 23, Wikiquote, José Raúl Capablanca

Read against the numbers: the accumulate-small-advantages idea is the bishop pair (a 2 to 3 point edge) and the lead table; Tarrasch's rook rule is the open-file and passed-pawn plan; Capablanca's advice is why the by-rating section sends 1800 players to endgames. Those readings are ours; the quotes are theirs.

A chess strategy checklist to run before every move

Run the checks in this order; stop at the first one that gives you a move.

1. What did my opponent just threaten. Check every capture, check and attack on your pieces. Only 4.5% to 6.5% of games below 1600 had no blunder.
2. Do I have a forcing move. A check, capture or attack that wins material by force goes first. See chess tactics.
3. Is my king safe. Castle if you have not, and do not open the centre while your king is still there.
4. Which piece is doing the least. Move your worst piece to a square where it matters, as in position 1.
5. Is there an open file or a pawn weakness. Put a rook on the file, or aim pieces at an isolated or doubled pawn. See pawn structure.
6. Should I trade. Trade when ahead, avoid trades when behind, and remember queens-off games draw about twice as often.

A model game: Sämisch against Nimzowitsch, Copenhagen 1923

Playing Black, Nimzowitsch took away White's squares one by one until White ran out of good moves. The game is known as the Immortal Zugzwang game.

How to win a chess game in six moves, and what it costs

A win in six moves exists, and it needs your opponent to play badly.

Position 7 is a Caro-Kann line that ends in checkmate on move six: 1.e4 c6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nd7 5.Qe2 Ngf6 6.Nd6#. It works only if Black answers 5.Qe2 with Ngf6: then the knight on d7 blocks the queen, the bishop is blocked by the e7 pawn and that pawn is pinned, so nothing can take on d6. Stockfish 17.1 at depth 16 finds the mate in one from that position. Black has other fifth moves. Fool's Mate can end a game on Black's second move; Scholar's Mate is a familiar four-move trap. Both depend on mistakes.