Why do some openings remain in use after decades of analysis? Nash equilibrium provides a language for thinking about strategies, but does not turn a popular variation into a theorem.
“Nash graphs” here is a teaching connection, not the name of a standard theory proving the Sicilian or Berlin correct.
What Nash defines
In Non-Cooperative Games (1951), John Nash describes a strategy profile in which no player can improve their payoff by changing strategy alone while the others keep theirs. The article spans pages 286–295, ten pages; it should not be confused with his short 1950 note.
A strategy specifies choices in situations within the game, not merely the first move’s name. The existence theorem for finite games allows mixed strategies; it does not automatically provide the optimal moves of a chess game.
What that means at the board
In the usual zero-sum model, each side prefers winning to drawing and drawing to losing. Russell and Norvig’s minimax analysis searches for a strategy against the opponent’s best replies.
A line called “equal” by an engine expresses something different: an estimate with a method and search budget. A draw between champions is another kind of information again. Neither alone proves that every possible deviation has been covered.
You can speak metaphorically of a practical balance in preparation if you label it as such. If a novelty changes a line’s evaluation, our knowledge changed; a mathematically proven equilibrium is not refuted by a theoretical fashion.
Game trees and position graphs
A tree keeps move sequences separate. A graph can merge sequences reaching the same position: transpositions. It must also specify what it calls a position.
Piece placement is not always enough. Side to move, castling rights and en passant can change legal moves. Applying all draw rules also requires historical information.
The configuration graph can contain cycles. Knights may leave and return to their starting squares. A tree retaining the entire game history does not confuse those occurrences. Calling every opening graph “acyclic” without qualification would therefore be misleading.
Endgames and calculated values
Tablebases can provide exact values within their domain and conventions. Distinguish win/draw/loss, distance to mate and distance to resetting the move counter. Syzygy documentation emphasises draw rules: a table distance is not always a number of winning moves in every tournament situation.
That precision makes endgames useful for discussing models: which information describes the state, which rule ends the game and which quantity is calculated?
Varying a repertoire: a practical decision
If you often face the same person, varying openings may make preparation less targeted. It does not prove that a 50%, 30%, 20% mix is a Nash equilibrium, or establish a universal five-hour study threshold for choosing a repertoire.
Examine the cost of each change: do you know the ensuing plans? Can you explain resulting positions? Are you gaining surprise at the cost of playing something you do not understand? These questions suffice to organise a practical choice.
Sporting results and the model
A draw may or may not suit the tournament standings. That can change the preferences you want a model to represent. Define them before announcing a new equilibrium.
In a strictly zero-sum model, both players cannot simultaneously increase their payoff. Importing the prisoner’s dilemma or a collective Pareto improvement without qualification therefore confuses the analysis.
Further reading: game theory, minimax and Zermelo.
Sources
- Nash, J. (1951). Non-Cooperative Games. Annals of Mathematics, 54(2), 286–295. Article.
- Russell, S., & Norvig, P. (undated online chapter). Adversarial Search. Artificial Intelligence: A Modern Approach. Chapter.
- Syzygy project (undated). probetool: endgame-metric documentation. GitHub. Documentation.
Comments