Judgment & Decision · Quick Test
Gambler's Fallacy
After several heads, are tails due?
The experiment
Read the situation and answer with the number you think is right.
What happened?
The sequence ends with three heads in a row, and many people feel that tails are now “due”. But each toss of a fair coin is independent of the previous ones: the chance of tails is 50%, always. In a representative sample of adults in Germany, 21% gave a probability above 50%.
Why does your brain do this?
We know that, over many tosses, heads and tails come out roughly half and half. So the mind expects even a short stretch to look balanced, full of alternations, and a run of heads looks out of pattern. Hence the feeling that chance will correct itself. But it does not: a deviation is not made up for later, it is just diluted among many other tosses.
When does this shortcut work well?
When outcomes come from a stock that runs out, expecting the other outcome’s turn makes sense. If you have drawn several red cards from a deck, fewer red ones are left, and the next card is more likely to be black. The shortcut fails when it is applied to draws that use nothing up, such as coins, dice and roulette wheels.
Where does it show up in daily life?
Betting on the lottery number that “hasn’t come up in ages”, betting on black after a long run of red at the roulette table, or feeling that after three daughters the next baby “has to” be a boy. Before betting on a correction, it is worth asking: what took the previous outcome out of play?
Deep Dive
Open Nerd Mode
The error is old. Pierre-Simon Laplace already described, in his Philosophical Essay on Probabilities, the crowds in France rushing to bet on lottery numbers that had not been drawn for a long time.
Amos Tversky and Daniel Kahneman (1974) linked the fallacy to the representativeness heuristic: we expect a random sequence to look random not only as a whole, but in every part. After a long run of red on the roulette wheel, most people think black is “due”, because it would make the sequence look more like chance. In their words, deviations are not “corrected” as a chance process unfolds, they are merely diluted.
Thomas Dohmen and colleagues (2009) asked this same question to 1,012 people aged 16 or older, a representative sample of Germany interviewed in 2005. The sequence was chosen with four tails in eight tosses, so that no one would doubt the coin. 60.4% answered 50%, 21.1% gave more than 50% (the gambler’s fallacy), 8.8% gave less (the opposite error, called the hot hand) and 9.6% said they did not know. Those who fell for the fallacy estimated, on average, a 79.2% chance of tails.
Peter Ayton and Ilan Fischer (2004) proposed another origin: in nature, many sequences resemble drawing balls from an urn without putting them back, where a repeated outcome does become less likely. In a roulette game with 32 students, the longer a run of one color, the less often they picked that color next. At the same time, the longer their own run of correct guesses, the more confident they felt about the next one: two faces of the same misunderstanding about chance.
Technical terms
- Probability
- Representativeness
- Intuition
Discovery
Described by Pierre-Simon Laplace in 1820.
Level of evidence
Robust evidence
Replicated in multi-lab studies or consistent meta-analyses.
Basis: Dohmen, T.; Falk, A.; Huffman, D.; Marklein, F.; Sunde, U. (2009). Biased probability judgment: Evidence of incidence and relationship to economic outcomes from a representative sample.
References
- Original studyTversky, A.; Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131.
- ReplicationDohmen, T.; Falk, A.; Huffman, D.; Marklein, F.; Sunde, U. (2009). Biased probability judgment: Evidence of incidence and relationship to economic outcomes from a representative sample. Journal of Economic Behavior & Organization, 72(3), 903–915.
- Original studyAyton, P.; Fischer, I. (2004). The hot hand fallacy and the gambler’s fallacy: Two faces of subjective randomness? Memory & Cognition, 32(8), 1369–1378.
Review
This text has not been reviewed by a volunteer yet. Want to help review it?
Found a mistake? Write to [email protected]