Learning to Forecast

My fellow countryman Niels Bohr (Nobel laureate in physics) once said: “Prediction is very difficult, especially if it’s about the future”. To me that sounds rather obvious and Albert Einstein said “I never think of the future, it comes soon enough”. This leads us to believe that we should not forecast and just wait and see.

But Henri Poincarè who laid the foundations for modern chaos theory wrote “It is far better to foresee even without certainty than not to foresee at all.” – An astonishing opposite to Albert Einstein.

As analysts, it is clear that we must take the side of Poincarè as opposed to accountants, who probably prefer to side with Winston Churchill, who said “I always avoid prophesying beforehand because it is much better to prophesy after the event has already taken place.”

For 25 weeks we have been forecasting the weekly temp development year over year development and we have learned some lessons as well as having some fun discussing the results week by week. I would like to tell you what I have observed over this time:

• Persistence is important, meaning you need to do it over and over again.

• Total experience does not count, task specific experience does count, but not as much as current task involvement. This is opposite to our review work, where experience count a lot more and therefore

• It is an advantage to follow the Weekly reporting closely and to apply our knowledge every week including what we learn on the way. This is also in line with the American economist Edgar R. Fiedler, who said: “If you have to forecast, forecast often.”

• The average of the group outperforms the best individual. Actually, the result is approx. 10% better than the best and 25% better than the worst.

• A rule that just takes the last week’ development applied ruthlessly is only 2% worse than the best, which tells me that if you have a system to validate your predictions against, you will be ok. This has the advantage that you do not need to spend so much time forecasting. I do not know if we can generalize to other types of forecasts.

• No one is right all the time, in fact only 2 out of 104 observations were spot on. An anonymous said -” It is often said there are two types of forecasts… lucky or wrong!!!!”

• Over the long run you can still do well, even if you make large mistakes. The best forecaster had one or two observations very far from the result.

I want to close with an anonymous quote: “A good forecaster is not smarter than everyone else, but merely has his ignorance better organized.”