What do you know
about tomorrow? How much will it look like today?
Superforecasting:
the art and science of prediction by Philip E. Tetlock, Dan Gardner. I skimmed
through parts of the book. Yet I discovered a brand new of looking at the
world. Do read, I borrowed it from the New York Public Library — please
consider supporting them.
The most interesting
part for my purpose is toward the first half. There are some interesting stuff
later for team leaders and managers in a corporate setting.
Tetlock and Gardner
look at the superforecasters, people who spend a good chunk of their time
making forecasts, and getting them right more often than not. Using a couple of
simple rules, they are able to make very accurate predictions about very
uncertain, far-off events.
Prerequisite: you
need to leave behind the notion of right or wrong and think probabilistically.
If you can’t let go of that, move on to the next post.
How you look at an
event. That’s what matters.
You think you know
what tomorrow will be made of? You have that arrogance? Well, sorry to
disappoint. You don’t. Look away from the specific case you are presented with
and try to identify a generic case to which you can assign a probability.
You meet a fine
woman, she wears glasses, she carries a bag with a bunch of books inside of it
and likes to go and listen to literary conferences at night. Is she a librarian
or a nurse? You say a librarian of course. Yet in our society, there are many more
nurses than librarians and so the probability of a random stranger to be a
nurse is a lot higher than her being a librarian…
Once you have the
base probability, you can now work through your specific case and add
additional factors to move your base probability up or down.
Are equity markets
going to go down by September?
S&P500 is highly
valued right now, and in the past, markets at such valuation levels have a
tendency to come back to the mean and go down. Some argue we are in the 95th
percentile of valuation. You get a base probability of markets going down by
September.
But the economy is
strong, there are no signs of any worries on the horizon. You bump down the
probability of markets going down. But the Fed is raising rates, who knows
where inflation is going and how the new generation of traders is going to
behave once they enter a world where interests and inflation are materially
above zero? Bump the probability down.
And so on. Until you
get comfortable with the probability you assign to the events you’re trying to
forecast.
Then day after day,
update the forecast and probabilities based on the new information available
each day in the news, in the markets etc.
One key point of
this method is to be sincere about the probabilities and the updates. And to
back-test your predictions once the facts are known and to understand how one
could improve the probabilities.
What facts were not
properly incorporated?
The method departs
from anything you see in the financial news every day: it is balanced and
nuanced (financial TV is never balanced, always one side or the other). It
performs a reality check (which financial pundits never do, have we ever seen
the success rates of some very famous financial TV expert that I would not
name?). And eventually it has the potential to lead to better decisions, in any
context.
I unfortunately lack
some of the basics in probability theory and I wish I could appreciate even
more their discussion of base rate, Bayes theorem etc.
One caveat: Tetlock
makes a very good by calling Nassim Taleb’s work on black swan. What if the
probability assigned to a given event is subject to black swans? Then you can
still make some forecasts, but you run the possibility that everything will
come crashing down because of some extremely rare event you never thought could
happen. Something that had a chance in a billion to happen, and yet it
happened.
So my (very modest)
contribution to this would be to add a level of variations, volatility,
variance into the prediction one is making. Try to know how far what you don’t
know extends.
A lot of food for
thought. My first predictions start today.
What about you? What
will you forecast? What do you look at to form your forecast?
