Does a 'refuted' opening cease to exist?

Not really. A local refutation proves that one path in the variation leads to a clear advantage for one side. The rest of the graph (move orders, transpositions, sidelines) can preserve playable local equilibria. That's why historical 'refutations' are often partial: the variation survives with a different move order or an in-between move. The Sicilian Sveshnikov, 'refuted' in the late 1990s, came back strong in the 2010s on new equilibria.

Why don't engines 'solve' all openings?

Because solving = reaching the Zermelo value (win, loss, draw under perfect play to mate). Engines produce a heuristic evaluation at finite depth (typically 40-60 half-moves); this lets them rank openings, find local equilibria , and break weak equilibria, but not prove absolute truth. The distinction is fundamental: Stockfish says 'this variation seems +0.2'; it never says 'this variation is drawn in the Zermelo sense'.

Does a Nash equilibrium guarantee the best collective outcome?

No, and that's central. The prisoner's dilemma example shows it: the equilibrium can be Pareto-dominated (both could do better by coordinating, but neither has incentive to deviate alone). In chess, 'collective Nash' would be the mutual draw agreement; 'competitive Nash' pushes toward unbalanced positions where each seeks to exploit an error. Tournament format (must-win, ranking, prize money) distorts the utility function and therefore the chosen equilibrium.

What is a 'mixed equilibrium' in practice for an amateur?

For an amateur, it means playing two or three different openings with stable probabilities (e.g., 50% Italian, 30% Spanish, 20% Scotch). The defensive advantage: prevents a frequent opponent from preparing specifically for you. The offensive disadvantage: you know each line less deeply. The right repertoire size depends on your weekly study time; under 5h/week, a pure repertoire is often more effective than a poorly maintained mixed one.

Why talk of 'graphs' and not 'trees' of openings?

Because openings transpose : different move sequences lead to the same position. A pawn to c4 at move 1 or move 3 can give the same Réti position in certain lines. A tree supposes a unique path to each leaf; a directed acyclic graph allows multiple paths. Professional opening databases (ChessBase, Lichess Masters) are actually compressed graphs, with millions of pre-calculated transpositions.

Science 17 min May 11, 2026

Nash Equilibria: Balance and Openings in Chess

Why do some openings last decades without being refuted? The Nash equilibrium explains the stability of theoretical variations and the deep logic of chess openings.

Updated on May 21, 2026

ContentsJump to a section

Why is the Sicilian Defense played at every level for a hundred years? Why hasn't the Ruy Lopez been "refuted" despite centuries of analysis? Why do some theoretical variations collapse in a few years while others seem indestructible? The answer lies in a mathematical concept developed by John Nash in 1950: the Nash equilibrium.

John Nash and the Equilibrium That Bears His Name

John Forbes Nash Jr. received the 1994 Nobel Prize in Economics for his contributions to non-cooperative game theory. His equilibrium notion, published in a two-page article in 1951, is one of the 20th century's most influential ideas.

A Nash equilibrium is a strategy profile (one strategy per player) such that no player can improve their outcome by unilaterally modifying their own strategy, assuming others maintain theirs. It's a state of mutual stability.

Openings as Dynamic Equilibria

In chess, theoretical openings can be analyzed as local Nash equilibria. When a variation is called "theoretically equal," it means neither White nor Black has found a deviation that improves their result provably.

Consider the Sicilian Defense, the most popular response to 1.e4. Why has this opening lasted for centuries?

Because it corresponds to a deep equilibrium. Black accepts a slightly asymmetric pawn structure for counter-chances. If White tries aggressive exploitation, Black has solid defensive resources. Neither can improve their outcome by unilaterally remaining in the theoretical framework: it's a Nash equilibrium.

When the Equilibrium Breaks

An opening's Nash equilibrium breaks when a player finds a deviation that actually improves their result. The history of chess is marked by these equilibrium ruptures.

Case Study: The Berlin Defense After Kramnik 2000

The most famous illustration of a rediscovered Nash equilibrium is the Berlin Defense (1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6) of the Lopez. For nearly a century, the Berlin was considered slightly inferior to the classical 3...a6. Vladimir Kramnik brought it out against Garry Kasparov in the 2000 World Championship match in London: he held every Black game without concession, directly contributing to winning the title.

From the Nash viewpoint: Kramnik demonstrated that an alternative equilibrium existed in a variation theory had classified as "Pareto-dominated" by 3...a6. Once demonstrated at high level, dozens of GMs adopted the Berlin in subsequent years.

The Graph Representation of Chess Positions

A directed graph is a set of nodes connected by directed edges. For chess, nodes are legal positions and edges are legal moves.

This graph is immense: about $10^{44}$ nodes. But its structure is revealing. From the initial position, the tree of games branches exponentially. But many different variations converge to the same positions (transpositions). The structure isn't a pure tree but a directed acyclic graph.

Attractors in the Graph

In this giant graph, equilibrium positions correspond to attractors. These are nodes toward which many paths converge and from which both players prefer to maintain their strategies.

Nash Equilibria in Endgames

Endgames offer more precise analysis terrain for Nash equilibrium, as the number of positions is small enough for exhaustive analysis.

In a King and Pawn vs King endgame, under perfect play, the position is either winning for the side with the pawn or drawn. This "under perfect play" state is precisely the endgame's Nash equilibrium.

Tablebases are the complete documentation of these equilibria for endgames with few pieces.

Opening Preparation as a Repeated Nash Game

In high-level chess competition, opening preparation isn't just learning theory. It's a strategic game itself, a repeated Nash meta-game.

The Nash equilibrium of this meta-game is a distribution over openings: playing each variation with a certain frequency to make your global strategy unpredictable.

Magnus Carlsen is known for a particularly sophisticated approach to this meta-game. He plays a wide opening repertoire, including unusual variations, precisely to disrupt opponent preparation. It's a mixed strategy in the Nash sense.

Standard Nash equilibrium assumes perfectly rational players. But what if the opponent commits an error with probability ε? Reinhard Selten (Nobel 1994 with Nash) proposed the concept of trembling-hand perfect equilibrium: a move is "robust" if it remains optimal even when the opponent slightly deviates.

This has direct translation in chess. A move can be theoretically perfect under strict Nash while being fragile: it depends on opponent precision until the last move. Another move, slightly suboptimal in evaluation, can be more robust because it keeps the advantage even if the opponent plays three or four inaccurate moves.

Refused Variations: Suboptimal Equilibria

A counterintuitive result of game theory is that there can exist Nash equilibria that aren't the best possible outcomes for both players. These suboptimal equilibria also exist in chess.

In a World Championship match where a draw is insufficient, both players have interest in choosing more unbalanced variations, even if those variations are theoretically less solid.

What Nash Reveals About Chess Progress

Progressing at chess means progressing in the capacity to maintain and exploit strategic equilibria. A strong player isn't simply one who calculates faster or knows more theory. They are a player who perceives more finely the equilibrium of each position.

Sources

Nash, J. F. (1951). Non-cooperative games. Annals of Mathematics, 54(2), 286-295.
Selten, R. (1975). Reexamination of the perfectness concept. International Journal of Game Theory, 4(1), 25-55.
Osborne, M. J., & Rubinstein, A. (1994). A Course in Game Theory. MIT Press.
Kramnik, V., & Damsky, I. (2000). My Life and Games. Everyman Chess.

Key Takeaways

A Nash equilibrium is a situation where no player can improve their outcome by unilaterally changing strategy
Openings considered "theoretically equal" are local Nash equilibria
A variation that gives a real advantage breaks the equilibrium and forces theoretical correction (Berlin Defense after Kramnik 2000)
Selten's refinement (trembling hand) explains why a "robust" move can beat a "theoretically optimal" move in real play
Position graphs allow visualizing these equilibria as attractors in the space of possible games
A mixed repertoire is a mixed strategy in the Nash sense: it diversifies to avoid exploitation

Frequently asked questions

: Not really. A local refutation proves that one path in the variation leads to a clear advantage for one side. The rest of the graph (move orders, transpositions, sidelines) can preserve playable local equilibria. That's why historical 'refutations' are often partial: the variation survives with a different move order or an in-between move. The Sicilian Sveshnikov, 'refuted' in the late 1990s, came back strong in the 2010s on new equilibria.
: Because solving = reaching the Zermelo value (win, loss, draw under perfect play to mate). Engines produce a heuristic evaluation at finite depth (typically 40-60 half-moves); this lets them rank openings, find local equilibria, and break weak equilibria, but not prove absolute truth. The distinction is fundamental: Stockfish says 'this variation seems +0.2'; it never says 'this variation is drawn in the Zermelo sense'.
: No, and that's central. The prisoner's dilemma example shows it: the equilibrium can be Pareto-dominated (both could do better by coordinating, but neither has incentive to deviate alone). In chess, 'collective Nash' would be the mutual draw agreement; 'competitive Nash' pushes toward unbalanced positions where each seeks to exploit an error. Tournament format (must-win, ranking, prize money) distorts the utility function and therefore the chosen equilibrium.
: For an amateur, it means playing two or three different openings with stable probabilities (e.g., 50% Italian, 30% Spanish, 20% Scotch). The defensive advantage: prevents a frequent opponent from preparing specifically for you. The offensive disadvantage: you know each line less deeply. The right repertoire size depends on your weekly study time; under 5h/week, a pure repertoire is often more effective than a poorly maintained mixed one.
: Because openings transpose: different move sequences lead to the same position. A pawn to c4 at move 1 or move 3 can give the same Réti position in certain lines. A tree supposes a unique path to each leaf; a directed acyclic graph allows multiple paths. Professional opening databases (ChessBase, Lichess Masters) are actually compressed graphs, with millions of pre-calculated transpositions.

Have a reaction?

A thought, a nuance, a story: an email lands directly with me, I read every one and reply.

contact@blogdungaucher.com