Glossary
Probability Matching
What is Probability Matching?
Probability matching is a decision-making strategy often seen in situations where individuals are tasked with predicting outcomes over a series of trials. Rather than favoring the outcome with the highest probability, an individual employing probability matching will choose outcomes in proportion to the likelihood of those outcomes occurring.
Background and Examples
-
Origins of the Concept
Probability matching has been studied in repeated binary choice experiments. It is suboptimal for maximizing expected correct predictions when outcomes are independent, probabilities are stable and known, rewards for correct answers are equal, and there is no additional predictive information.
-
Numerical Example
Suppose independent trials produce red with probability 0.70 and blue with probability 0.30. Always predicting red yields 70% expected accuracy. Independently guessing red 70% of the time and blue 30% yields 0.70 × 0.70 + 0.30 × 0.30 = 0.58, or 58% expected accuracy. These are calculated expectations, not guaranteed results for a short run.
Relevance and Impact
-
Impact on Decision Making
Matching frequencies can reduce expected accuracy in that setting, but observed matching does not uniquely identify a mental strategy. Learning, searching for patterns, and task interpretation can produce similar choice frequencies.
-
Implications for Behavioral Science
Vulkan’s review surveys repeated binary choice evidence and different theoretical interpretations. Whether maximizing is appropriate depends on the decision problem; unequal rewards, changing probabilities or informative cues can change the preferred strategy.