The Art and Science of Prediction

The book helped to put structure to everyday analysis. Often I would do one or two of the methods in the book but not with a strong focus on my process. I think in the end, I will now approach problems with a consistent level of rigor.

This book has some great ideas and content. The issue is a lot of it is filler that's a complete waste of time. If they only focused on what differentiates superforecasters, the book would be 1/2 the size and much better.

As an ex-supervisor of professional forecasters (economics and state tax revenues) I found the author's descriptions of the methods of his group of successful forecasters very interesting and made me think our crews might have held there own with his folks. But we used complex regression models as much as possible and used other techniques to supplement. Constant revisions though are clearly one of the best techniques regardless of what your base method is. The one part of the book I found unsatisfying was the discussion of the theories Nassim Taleb on the supposed central place that so-called black swan events play in turning the course of long term history, thus limiting the range of forecasting to just a few years. Tetlock has no data to offer on this. I think he needs to try longer range forecasts and see if they can be useful while not exact (embrace the probalistic future). Yes a big asteroid can all but destroy life and we may not currently be able to predict it but I dispute that World War 2 would not have happened had Hitler not been born. Yes he was a extraordinary villain but he had plenty of fellow travelers who would have similarly lead the Nazis to war.

Thought provoking and scientific yet easy to read. He opened my eyes to the global importance of this topic while pointing out forecasters lack of accountability. He also showed the potential for accurate and scientifically accountable forecasters could have.

Too many chapters filled with nothing useful or informative and over emphasizing too often. The core topics were you get lucky or you don't and some people are luckier at guessing than most there isn't much depth into the applications on how to forecast accurately. This book has been a big letdown.

This book contains plenty of information about the art of predicting the future, and why even some (but not all) experts are bad at it. You won't be given a formula for how to predict the future, not explicitly. Instead you will be asked to examine your own method of thinking to find the human tendencies that hamper our ability to make evidence based decisions.

Tetlock starts by examining the human tendencies that get in the way of objectively evaluating evidence, and why we should be skeptical of some predictions. When people make decisions we tend to make hasty generalizations, and never bother keeping score if the decisions we made were correct. This quick thinking is prone to error and a lack of reflection means we never have to own up to those errors in our judgement. Tetlock then discusses the idea of 'superforcasting' and applying evidence based methods to prediction. A fair amount of the book is dedicated to the criticisms of the ideas presented in the book, which underscores the skeptical nature of this work.

If you have read other books on human judgement and decision making such as Daniel Kahneman's "Thinking Fast and Slow" then much of this information will be old to you, but as a first introduction this book does its job well. I recommend that every person read this, not to become a superforcaster themselves but so they can become better critical thinkers.

I simply love these combination of research that teaches us new things and a very, very good read. I read with a smile on my face from the first to the last page. Than you Philip Tetlock for the science and the book!

The author is long on anecdotes but short on facts, which is what he deplores of others. We are being told why he thinks good forecasters are what they are, with names but not much else. What do those people predict, how and why they reached that conclusion? We are left with more questions than answers.
Examples of good forecasts and how they came about, would have been better. For a few pages in the 2nd part of the book, those insights appeared, but were short lived.
It's a "how to" book without much how.

I excitedly preordered this book, after reading about it in The Economist. So it was a big disappointment when it did not live up to my expectations.

If you're looking to learn all about Tetlock's study, how he came up with the idea, how his team performed in the forecasting competition, as well as the personal rebuttals to other people who criticized his findings, then this is your book. However, if you're actually looking to improve your own forecasting, you will be disappointed. There's only a tiny sliver of applicable information, lost in a sea of superfluous and blatantly obvious content. And little applicable information there is, is smothered in caveats. An example of blatantly obvious filler content: three entire sections of one chapter are actually devoted to talking about how sometimes adjusting to new information improves forecasts, sometimes it made the worse, and the best forecasters find a way to do both. Was there any person on the planet who didn't already know this?

In its defense, the book did introduce two new concepts to me which I did appreciate. It talked about the idea of Fermizing problems, which he describes as breaking up an unsolvable problem into smaller parts and then forecasting those. Yes, that is pretty obvious and pretty much the only way to handle a difficult problem (if you're serious about finding an answer). But it did make me feel better about the way that I do my forecasts. It also said that he found its better to think in smaller local probabilities (12% instead of 10%), which I had not been doing. So that was actually something new I learned from the book that was useful.

But again, these are tiny nuggets of information that would be better suited for a single blog posting to summarize them all. Stretching this out into an entire book was I'm sure very profitable for the author, but a disservice to anyone trying to improve their own forecasting.

Building on the writings of Kahneman, this book takes some of his research and applies it to forecasting. It's so heavy on the work in Thinking Fast and Slow, that it almost felt like Kahneman should have been listed as a co-author.

More generally it explains Tetlock's research on forecasting, and what it takes to be a superforecaster. I wanted something a bit richer, but it's a good elementary introduction.

Should be self-required reading for any rational person that cares about the future and how it will be impacted by the present. Forecasting is part of life, so to be better at it is beneficial. Also it is an unpredicted screed AGAINST bureaucracy and the Democrats' god, the "Elite" expert, especially the Harvard elite. They and the "Intelligence Community" the President of the Democrats praises constantly are far less able than a group of smart, dedicated people who predict for fun and challenge not money.

Think about that while you read this great book.

Before I read book I liked Jim Collins’ interpretation of the essay, “The Hedgehog and the Fox and hedge;” Great companies are hedgehogs. Yet, Tetlock argues Super Forecasters constantly update their forecast based on new information; they are foxes and they need to know many things. In my opinion great forecasters are foxes, but use many pieces of information to become hedgehogs.

Before I read book I liked Jim Collins’ interpretation of the essay, “The Hedgehog and the Fox and hedge;” Great companies are hedgehogs. Yet, Tetlock argues Super Forecasters constantly update their forecast based on new information; they are foxes and they need to know many things. In my opinion great forecasters are foxes, but use many pieces of information to become hedgehogs.

Tetlock builds upon Kahneman's expertise into forecasting, examining how people thought and the clear winners are those that think in possibilities and probabilities, gather all information, pragmatic. This distinguishes from pundits that "squeeze complex problems into the preferred cause-effect templates and treated what did not fit as irrelevant distractions." Embrace nuance, seek, measure, test, experiment in terms of probabilistic outcomes against Brier scores.

Excellent research on how to improve forecasting accuracy make it a useful read for those in a wide range of businesses. The rational for achieving precise forecasting is less convincing For example, did it really matter whether intelligence estimates of the location of Osama bin Laden was 84% or 87% accurate?

This book is absolutely worth reading if you are afraid of blind spots that we have as forecasters. It's simple, deep, and very insightful at times. The work on leadership and teamwork is less compelling, but the definition of what makes a super forecaster good is outstanding.