Why Better Forecasts Can Make the Forecast Wrong
The El Farol problem is a model of a decision that depends on how many other people make the same choice. Each person would enjoy going to a bar only if attendance stays below a comfortable limit. The difficulty is that the limit is not the only thing to predict: each person must also predict everyone else’s prediction.
If a forecast suggests the bar will be quiet, going seems sensible. But if many people act on that forecast, the bar fills up. A prediction of a crowded night can push people to stay away, leaving it nearly empty. The forecast changes the outcome it was meant to describe.
Why There Is No Single Winning Strategy
A conventional forecast works best when people’s choices do not strongly affect the thing being forecast. Here they do. If everyone uses the same rule—such as repeating last week’s choice or trusting the recent average—their decisions can move together and overshoot. Each person may be reasoning sensibly alone while the group produces a poor result.
In the model, people can try different simple ways of predicting attendance and favor the rules that have worked recently. Because conditions keep shifting in response to their choices, a strategy that succeeded before may stop succeeding. The point is not that people cannot learn; it is that learning from the same public history can make their behavior more alike.
The Value of Different Guesses
The model highlights a practical tension in markets and other crowded decisions: accurate information is useful, but widely shared reactions to it can amplify swings. When everyone rushes toward the same opportunity—or retreats from the same risk—the crowd itself becomes part of the problem.
A less crowded bar does not require every visitor to predict perfectly. It requires their choices not to synchronize too tightly. That is why, in situations shaped by collective behavior, a variety of expectations can sometimes serve the group better than one apparently obvious forecast.
