Definition Player-2 LP Formulation of a Matrix Game
definition admitted

Player-2 LP Formulation of a Matrix Game

For a matrix game $A : I \times J \to \mathbb{R}$ played row-by-column, the player-2 (minmax) linear program seeks a mixed strategy $y \in \Delta(J)$ ([[game_theory.strategic_game.zero_sum.core.mixed_strategy_simplex]]) and a scalar $w$ such that every pure row response of player 1 yields at most $w$:

$$ \boxed{\begin{aligned} \text{minimise} \quad & w \\ \text{subject to} \quad & \sum_{j \in J} A_{ij}\, y_j \;\le\; w &&\text{for every } i \in I, \\ & \sum_{j \in J} y_j = 1, \\ & y_j \ge 0 &&\text{for every } j \in J. \end{aligned}} $$

The constraints encode "$A y \le w \cdot \mathbf{1}_I$" plus probability normalisation on $y$.

Duality with player 1's LP

The player-2 LP is exactly the dual of player 1's LP ([[math.linear_programming.minimax_bridge.player_1_lp]]) in the standard LP duality sense (write player 1's LP in equality / slack form; take the LP dual; the result is player 2's LP after relabelling).

Strong LP duality ([[math.linear_programming.strong_duality]]) then yields:

  • Feasibility on both sides: both LPs are feasible — player 1's LP has $v = \min_j A_{ij}$ for $x = e_i$; player 2's LP has analogous feasibility.
  • Optimal value equality: $v^* = w^* = \operatorname{val}(A)$ ([[game_theory.strategic_game.zero_sum.core.value]]) — the minimax theorem ([[game_theory.strategic_game.zero_sum.von_neumann_minimax]]) re-derived from LP duality.
  • Complementary slackness: optimal solutions $(x^*, y^*)$ satisfy support / complementarity conditions ([[math.minimax.support_complementarity]], [[math.minimax.strong_complementarity]]).

See also

  • [[math.linear_programming.minimax_bridge.lp_optimal_iff_optimal_strategy]] formalises the bijection between LP optima and game-theoretic optimal strategies.
  • [[math.minimax.lp_duality_minimax_proof]] is the explicit proof-plan turning LP duality into the matrix-game minimax theorem.
  • [[math.linear_programming.minimax_bridge.zero_sum_lp_bridge]] discusses the AGT-style presentation of the same equivalence.

References

  • [AGT Section 1.4.2] Nisan, Roughgarden, Tardos, and Vazirani, Algorithmic Game Theory.
  • [MFoGT Section 2.5] Laraki, Renault, and Sorin, Mathematical Foundations of Game Theory.

Used by

Also in