The problem of using ENSO forecasts is not at all straightforward. The basic difficulty arises from the fact that forecast information is probabilistic – our knowledge of the future is given imperfectly and we must learn to use this imperfect knowledge in an intelligent manner, especially when the skill is not high.

To illustrate the problem, we begin with an (admittedly fanciful) analogy. Suppose a stranger whispers in your ear that he is offering you a rare and unusual gift: a coin that looks and feels like every other coin of its type but will fall heads 55% of the time. The coin is yours to keep but it is up to you to find out how to make use of this gift.

The first problem is to find out if the stranger is telling the truth. So you flip the coin and it shows tails. This, of course, does not indicate that the stranger's words are fraudulent: one must flip the coin a very large number of times. So you flip the coin 100 times and 53 times it shows heads and 47 times it shows tails. This is promising, but it still does not prove that the coin is what the stranger said it is. So you flip the coin 1000 times and it falls heads 552 times and tails 448 times. Now it seems to be true that the stranger has told the truth – the coin is indeed a 55% heads coin.