I recently read the book "Superforecasters" by Philip Tetlock which discusses the science of making accurate forecasts of future events, particularly in the socio-economic/political spheres.
One of the studies done by the author (in the 80s?) showed that supposed political and economic experts were no better than chance at predicting things within their supposed areas of expertise, yet these political pundits continue to make confident predictions. The problem is that their predictions tend to be couched in weasel words or are so vague as to generally be untestable.
But the perhaps more important result is that some people are able to make accurate predictions about future events and the book looks at the methodologies these so called "Superforecasters" use. Things like Bayesian reasoning, quantitative analysis, and a non-ideological outlook seem to be important.
Anyway, after reading the book I started taking part in the forecasting game at the Good Judgement Project which asks very specific questions about world events and asks you to make a probabilistic forecast about their outcome. You then get a "brier score" based on the accuracy of your predictions. You can choose which questions to answer and which to ignore, but your score is always compared to those answering the same questions on the same days. You can also change your forecasts based on new information that arises, which is treated as a new forecast for the same question.
I thought others on here might find this interesting as a way of testing their own understanding of political and economic events in the world.
So far I've had 6 questions come up and have a brier score of 0.19 (compared to a median of 0.277 for all other users answering the same questions on the same days). My best prediction so far (in so much as I did better than other forecasters) was in predicting apple's reaching $1 trillion before January 2019, and before any other company, which I think I put at 70%. I've answered a total of 25 questions so we'll see if my brier score starts to rise up soon. A score of 1 is, if I recall correctly, equal to random guessing, and a score of 2 I think is being perfectly wrong, whereas a 0 is a perfect score. The brier score is based on a system developed for weather forecasting.
One of the studies done by the author (in the 80s?) showed that supposed political and economic experts were no better than chance at predicting things within their supposed areas of expertise, yet these political pundits continue to make confident predictions. The problem is that their predictions tend to be couched in weasel words or are so vague as to generally be untestable.
But the perhaps more important result is that some people are able to make accurate predictions about future events and the book looks at the methodologies these so called "Superforecasters" use. Things like Bayesian reasoning, quantitative analysis, and a non-ideological outlook seem to be important.
Anyway, after reading the book I started taking part in the forecasting game at the Good Judgement Project which asks very specific questions about world events and asks you to make a probabilistic forecast about their outcome. You then get a "brier score" based on the accuracy of your predictions. You can choose which questions to answer and which to ignore, but your score is always compared to those answering the same questions on the same days. You can also change your forecasts based on new information that arises, which is treated as a new forecast for the same question.
I thought others on here might find this interesting as a way of testing their own understanding of political and economic events in the world.
So far I've had 6 questions come up and have a brier score of 0.19 (compared to a median of 0.277 for all other users answering the same questions on the same days). My best prediction so far (in so much as I did better than other forecasters) was in predicting apple's reaching $1 trillion before January 2019, and before any other company, which I think I put at 70%. I've answered a total of 25 questions so we'll see if my brier score starts to rise up soon. A score of 1 is, if I recall correctly, equal to random guessing, and a score of 2 I think is being perfectly wrong, whereas a 0 is a perfect score. The brier score is based on a system developed for weather forecasting.
via International Skeptics Forum https://ift.tt/2MD7ruj
Aucun commentaire:
Enregistrer un commentaire