Decision Making · 7 min read
When to Randomize a Decision (and When Not To)
A practical framework for distinguishing decisions that benefit from random choice, decisions that demand reasoning, and the surprising overlap between them.
Random decision-making sounds, on its face, like a confession of failure. If reasoning could produce the right answer, you would use reasoning; using a coin flip or a spinner wheel feels like an admission that you cannot tell the options apart, or worse, that you do not want the responsibility. Both intuitions are wrong. Randomization is, in well-defined circumstances, the correct strategy — and sometimes the only correct strategy. Other times it is genuinely an abdication. The skill is knowing which case you are in.
When Randomization Is Strictly Optimal
Game theory provides the clearest case for randomization. When you are facing an adversary who is observing your behavior and can exploit any predictable pattern, the unique optimal strategy is often a randomized one. A penalty kicker who always shoots to the same side will be saved every time. A poker player who only bluffs with weak hands will be called every time. A military commander whose movements follow a discernible schedule will be ambushed.
The formal result is John Nash's theorem on mixed strategies: in many strategic games, the equilibrium consists of each player choosing among their options according to specific probabilities. The probabilities are not arbitrary — they are calculated to make the opponent indifferent between their own choices. But the choices themselves, drawn from those probabilities, must be genuinely random for the strategy to work.
When Randomization Saves Time
Outside of adversarial settings, randomization is useful whenever the cost of deciding exceeds the cost of being wrong. If you face ten roughly equivalent options for what to have for lunch, the time spent comparing them carefully is probably worth more than the marginal benefit of choosing the best one. Randomly picking among the equivalent options frees up the mental resource for decisions where careful reasoning actually pays off.
This is the principle behind everyday delegation to a spinner wheel. You are not abdicating thought; you have already done the thinking by identifying that the options are roughly equivalent. The randomization is what closes the loop on a decision your reasoning has already finished but cannot resolve.
When Randomization Reveals Information
A subtler case: randomization can produce information that deliberate choice cannot. Randomized controlled trials in medicine, A/B tests in software, and randomized policy experiments in economics all rely on this. By assigning treatment randomly, the researcher ensures that any difference in outcome can be attributed to the treatment, not to the characteristics that drove the assignment. Deliberate choice — even well-intentioned deliberate choice — would smuggle in selection effects that contaminate the result.
In organizational settings, this extends to random audits. Auditing the same departments repeatedly produces a culture where those departments behave well during audits and poorly otherwise. Auditing departments randomly, with no predictable pattern, produces a culture where every department behaves well all the time — because at any moment, any of them might be selected.
When Randomization Distributes Fairly
When something good or bad must be allocated and the recipients are roughly equivalent, randomization is often the only mechanism that all parties can accept as fair. Random draft orders in sports leagues, random selection of jurors, lottery-based allocation of school placements, and random distribution of unpopular tasks all rely on this. No party has grounds to object that they were unfairly chosen, because no party was chosen — the randomness was.
This makes randomization a powerful tool for managing situations where any deliberate choice would create grievance. A manager who randomly assigns weekend shifts faces no accusation of favoritism; a manager who assigns them based on judgment, no matter how fairly, faces some accusation no matter what.
When Randomization Is Wrong
Randomization is the wrong tool when the options are not equivalent and the cost of choosing wrongly is high. A medical treatment selection should not be randomized; the best option for the patient's specific case should be reasoned to. An investment decision involving non-trivial money should not be randomized; the option with the highest expected return given the investor's constraints should be reasoned to. A safety-critical choice should not be randomized.
More subtly, randomization is wrong when the act of randomizing would itself signal carelessness. A hiring manager who randomly picks among finalists conveys, accurately, that they could not tell the candidates apart — which may or may not be the message they want to send. A leader who randomly chooses strategic direction conveys, accurately, that they have no view — which is rarely the message they want to send.
A Practical Test
Three questions disambiguate most cases. First: are the options genuinely equivalent in expected outcome, or is one better than the others? If one is clearly better, do not randomize. Second: is the cost of careful deliberation greater than the difference between options? If yes, randomize and save the effort. Third: would the appearance of randomness improve trust or fairness in the outcome? If yes, randomize and publish the method.
Closing the Loop
The cultural caricature of randomization — the indecisive person who flips a coin because they cannot make up their mind — captures only the laziest use case. The full picture is richer. Randomization is a tool for adversarial settings where unpredictability is the point, for trivial settings where deliberation costs more than it gains, for experimental settings where unbiased information is the point, and for fairness settings where any deliberate choice would create grievance. Used well, it is not an admission of failure but a recognition that some decisions are best made by methods that humans, with all our biases and limitations, cannot replicate.
Recommended Reading
If you found this article useful, these books go deeper into the same topics. Each title is hand-picked for the material covered above.
- Algorithms to Live By: The Computer Science of Human Decisions by Brian Christian, Tom Griffiths — When and why mixed strategies (randomization) beat deterministic decision rules in real life. View on Amazon
- Theory of Games and Economic Behavior by John von Neumann, Oskar Morgenstern — The original mathematical foundation for randomized strategies in game theory. View on Amazon
As an Amazon Associate, Roulety earns from qualifying purchases. This does not change the price you pay and helps support the writing on this site.
© 2026 Roulety. Free online spinner wheel for decisions, games, and fun.