How to make better predictions? Lessons from the book Super-forecasting at Data Science Speakers

On the 2nd November, I presented “How to make better predictions? Lessons from the book Super-forecasting” at Data Science Speakers. Below you will find a summarised version of this presentation

Lets start by making a forecast, what would be your forecast for the next question?

Who is going to win the American presidential election?

In answering this question, what elements are you considering in this prediction? Are you considering your political views? Are you taking into account the latest results from the polls? Or perhaps are you considering the weather in the swing states?

Many of these considerations are the basis of a book that I recently read and that enjoyed quite a lot that is “Super-forescasting. The Art & Science of Prediction” by Philip Tetlock and & Dan Gardner.

From this book I would like to share the following three lessons that I learned

Lesson 1: Break the problem into sub-problems

How many piano tuner are there in Chicago?

One lesson that the author shares is that it is important to break the problem into sub-problems. This will help to use any information you may know already about the problem and to start to understanding which parts you feel more certain about and which parts you are entirely guessing. For example you can break this problem in the following questions and provide different estimates which you may feel more confident about it and might need research

  1. How many people are there in Chicago?
  2. What percentage of people own a piano?
  3. How many institutions like schools, concert halls or bars own pianos?

Lesson 2: Balance the outside and inside view

Was Yasser Arafat poised by Israel authorities?

Another lesson that the author shared is to take different views in answering the questions. From the previous question the outside inside view could be as follows

  1. The outside view should not be focused on the politics at all!
    At this view you should be thinking in points that could be focus in the poisonous substance that was apparently found in the belongings of Yasser Arafat after found death. Some points to consider can include:
  • How fast does Polonium decays?
  • What is the science behind Polonium testing

2. The inside view analysis should focus on what would it take for the question to be true?. At this point it can include the political view of the question. Some points to consider can include:

  • How can Israel obtain polonium?
  • How badly enough did Israel want dead Arafat?
  • How did Israel have the ability to poison him?

Lesson 3: Forecasts are not set in stone

And now back to the first question:

Who is going to win the American presidential election?

The third lesson that I learned from this book is that every forecast can change constantly, like the weather forecast! Two main elements to consider on how frequent you should update a forecast are:

  1. Forecast are judgement based on available information that should be updated in light of changing information. So if there is any relevant information considering how it impacts your previous assumptions and what sub-problems or outside/inside views are affected
  2. Strike balance in under- and overreaction on new information. Some information might need to update your forecast more radical while others will require just minor adjustments.

So considering these three lessons, what will be your forecast for the candidate to win the next American presidential election?

The following are the slides presented at this event

To know more about Data Science Speakers and their coming events visit:

Data Science Speakers Club

London, GB
1,676 Members

The Data Science Speakers Club is a Toastmasters International club enabling the current and next generation of Data Scientists, Entrepreneurs, Innovators, Mathematical and Lo…

Next Meetup

Data Science Speakers Club Meeting – ONLINE MEETING

Monday, Nov 16, 2020, 6:30 PM
12 Attending

Check out this Meetup Group →



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s