Game Theory · GT-06
When there's no single best move, the best strategy is to be unpredictable — on purpose, in exact proportions.
An equilibrium where at least one player randomizes over multiple strategies according to specific probabilities, because no single pure strategy is a best response to the opponent's play. Being predictable is itself the losing move.
Formalized by John von Neumann's 1928 minimax theorem and generalized by John Nash's 1950 existence proof, which showed every finite game has an equilibrium once randomized strategies are allowed.
The Mechanism
Penalty kicks: the equilibrium mix, not a coin flip
Neither pure strategy survives — if the kicker always goes right, keepers learn and always dive right. The equilibrium is a specific randomization (roughly matching each side's save/score rates) that makes the opponent indifferent between their own options.
01 · INDIFFERENCE IS THE EQUILIBRIUM CONDITION
Not 'confuse them' — make them indifferent
In a mixed equilibrium, each player's randomization probabilities are set so precisely that the opponent's expected payoff is identical across all their pure strategies — the opponent literally has nothing to gain by favoring one option, which is what makes the mix stable.
02 · IT ONLY APPEARS WHEN PURE EQUILIBRIA DON'T EXIST
A fallback that guarantees a solution
Matching Pennies and penalty-kick games have no pure-strategy Nash equilibrium — whatever one side commits to, the other exploits. Nash's 1950 theorem guarantees a mixed equilibrium always exists even then, which is precisely why it was a landmark: it made equilibrium existence universal, not just common.
03 · REAL-WORLD MIXING NEEDS ACTUAL RANDOMNESS
Humans are bad unpredictability generators
Empirical studies of professional soccer penalty kicks (Chiappori, Levitt, Groseclose 2002) find real players approximate the theoretical mixed equilibrium fairly well, but most people asked to 'act randomly' produce detectable patterns — true equilibrium mixing requires a genuine randomization device, not an intuition for unpredictability.
Where It Fails / Inversion
Where it fails / inversion
Mixing is only optimal against an opponent who is also playing well and will exploit predictability. Against a weaker, pattern-following opponent, deviating from the equilibrium mix toward their specific exploitable tendency is worth more than sticking to theoretical purity — equilibrium play is a floor against a smart opponent, not always a ceiling against a weak one.
How To Use It
Worked example · pricing promotions
A retailer alternates between full price and surprise discounts in a way competitors can't predict, specifically so rivals can't time their own promotions to undercut a known pattern. The randomization isn't random for its own sake — its frequency is calibrated so a competitor gains nothing on average by betting on either 'they'll discount' or 'they'll hold.'
How to use it
When you notice a competitor or counterpart exploiting a predictable pattern in your own behavior, the fix usually isn't picking a new fixed pattern — it's calculating the mix of options that leaves them with no exploitable edge, and then genuinely randomizing (a die roll, a schedule from a random number generator) rather than trusting your gut to be unpredictable.
See Also