Strategic Economics

Mixed Strategies and the Value of Being Unpredictable

Set an audit, inspection, or promotion policy as a probability rather than a schedule, and discover that the right rate is determined by the other side's numbers, not yours.

  • Advanced
  • 13 min total
  • 12 chapters

What decision this helps you make: What proportion of the time to do the costly thing (inspect, audit, discount, defend) when any pattern the other side can learn is a pattern they can exploit.

What this topic is

A mixed strategy is a deliberate randomisation across your options, with stated probabilities. It exists because some situations have no stable answer in which each side simply picks one thing: whatever you settle on, the other side has a profitable reply, and whatever they settle on, so do you. In those situations the equilibrium is not a choice but a probability, and the striking result is that your probability is determined by the other side's payoffs rather than your own. What makes them stop exploiting you is that you have made their two options equally attractive.

Why it matters

Every inspection regime, audit programme, quality-sampling plan, promotional calendar, and security rota is a mixed strategy whether or not anyone designed it as one. Where they are run on a schedule, the schedule is the vulnerability, and the low violation rate the programme reports is often evidence of successful timing rather than compliance. The framework also produces one uncomfortable and very useful comparative static: raising the penalty for a violation reduces how much enforcement you need, and does not reduce violations at all. Only cheaper detection does that.

Who should learn it

Anyone who sets an audit, inspection, or sampling rate; anyone designing a promotional calendar customers could learn to wait for; and anyone negotiating repeatedly against a counterparty who is reading their patterns.

What you will understand

  • Why some situations have no answer in pure choices, and how to recognise one before you go looking for the answer that is not there
  • The indifference principle: how to compute each side's equilibrium probability from the other side's payoffs
  • Why doubling a penalty cuts enforcement cost and not violations, and which lever does reduce them
  • Where randomisation is the wrong tool, and what to use instead when the other side is not actually optimising

Prerequisites

Common misconception

"Randomising means the decision does not matter, since you are indifferent between the options." Backwards on both counts. You are indifferent, but that indifference is an output of the other side having chosen their probability correctly, and it is exactly what stops them exploiting you. The probabilities are not arbitrary: yours is pinned by their payoffs and theirs by yours, and moving even slightly off the equilibrium mix hands the other side a strategy that beats you systematically. Being unpredictable is a precise quantity, not a shrug.