El Farol Bar Game, explained.
The El Farol bar problem explores a coordination dilemma: an activity is attractive when not too many people choose it, but everyone is trying to anticipate everyone else.
Why it happens
A prediction can change the outcome it predicts. If many people expect a quiet evening and attend, it becomes crowded. If many expect crowds and stay home, it becomes quiet. Mixed attendance can avoid perfect synchronization, without guaranteeing a good evening.
The payoff to attending depends on how many other people independently make the same forecast.
Read the result
Change the attendance probability and compare crowding with missed opportunities. The bots in this simplified experiment use probability rules; they do not reproduce the full adaptive forecasting process of Arthur's original problem.
A worked example
A small venue
Ten friends each enjoy a venue only when total attendance stays at six or fewer.
If all expect space and go, everyone encounters a crowd. If all expect crowding and stay home, the venue is empty.
Their choices help create the condition they are trying to predict, making a single shared forecast difficult to sustain.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
Ninety-nine other agents independently go with probability 1−expected-attendance/100, a transparent toy response to the public expectation. A seeded round samples turnout; your attendance adds one. Going pays +1 if total attendance is at or below capacity, −1 otherwise; staying yields zero.
Where this idea is useful
A practical use
Decide whether to visit a popular place when everyone has the same crowd forecast.
A common misconception
“A correct crowd forecast stays correct after everyone follows it.”
Following the forecast changes attendance. That feedback can invalidate the original prediction.
What this explanation leaves out
- Other agents use a fixed response rule, not Arthur's adaptive heterogeneous forecasting strategies.
Why use randomized attendance?
Independent randomization can spread participation across evenings and reduce synchronized decisions. It still produces variation, and it is not the only possible coordination mechanism.
Could everyone acting on the same prediction make that prediction fail?
Associated thinkers
Further reading
Explore the original research or the teaching reference behind this experiment.