London Grip offers another fable for our times –
showing there is more than one way to make successful predictions
… and also exposing potential flaws in the compilation of performance statistics
THE WEATHER FORECASTERS
The scene is a convention of meteorologists; more specifically the bar outside the main meeting hall where, at the end of a long day, the American, British, Bulgarian and Romanian delegates are having a quiet drink. The American issues a light-hearted challenge: off-the-record and notwithstanding what we might say from the conference platform – do we dare to tell one another just how accurate we think our forecasts really are? And, with disarming frankness the American goes on to admit that even with the aid of satellites and weather balloons and networks of super computers the USA manages only about 65%-70% reliability in forecasting.
The British delegate, well used to smiling self-deprecation, has to acknowledge that UK resources do not compare to those of the USA. But he mumbles something (which is more or less unintelligible to the others) about the predictive properties of a piece of seaweed that has been hanging up in the Ministry of Defence since Winston Churchill’s time; and this, he claims, when combined with a bit of old-fashioned British know-how, produces a success rate of about 55-60%.
The Bulgarian representative does not know how to fall back on English-style understatement and a mythical reputation for muddling through. Hence he can express nothing other than embarrassment when confessing that his own country’s lack of expensive technology means that he would claim no more than 30-35% accuracy for his forecasts.
To everyone’s surprise the Romanian delegate looks very pleased with himself as the group’s eyes turn towards him. We get 65-70% success he says. But how can Romania’s meteorologists possibly be as good as those from the USA when their technical expertise is no better than Bulgaria’s? It’s simple, he says. We listen to the Bulgarian weather forecasts and then predict the opposite.
Moral: An appearance of success often rests on exploitation of other people’s mistakes.