Superforecasting: Interview with Phil Tetlock

[dcollon]

Great Interview from Shane/Farnam Street with Phil Tetlock (author of Superforecasting). 

 

 

https://www.farnamstreetblog.com/2015/12/philip-tetlock-prediction/

 

 

[netnet]

The skills he promulgates can lead to better investing prowess; see his book or master class on Edge.org

[eclecticvalue]

I wanted to get involved with his contests.

[netnet]

You can join the forecasting contest here: https://www.gjopen.com/

[Jurgis]

You can join the forecasting contest here: https://www.gjopen.com/

 

Joined.  8) I wish they had more business/economics questions. I'm definitely not an expert in world affairs and I'm not sure I'd become a better investor if I try to predict political events.  8)

 

We'll see how this works out.  8)

[Jurgis]

So. :)

 

Anyone want to do a startup/fund (or service ala oddballstock's https://www.completebankdata.com/ ) based on superforecasting? Get Philip Tetlock as a partner, potentially get his superforecasting team(s) (or build your own - from CoBF participants haha?), make predictions, ..., PROFIT!

 

There are some organizational and methodological questions to be resolved, so this is not commodity-trivial (i.e. you might get competition, but not immediately and possibly not very competitive if you solve some of the issues well). Of course, if you go fund route, money raising would also be interesting challenge. If you go service route, then https://www.completebankdata.com/ might be a good blueprint on how to do it...

 

Fund would probably do much better if this really works.

Service is more inline with what Tetlock does now and perhaps easier to implement / get running.

 

In best case, you could do https://en.wikipedia.org/wiki/Renaissance_Technologies  8)

[netnet]

So. :)

 

Anyone want to do a startup/fund (or service ala oddballstock's https://www.completebankdata.com/ ) based on superforecasting? Get Philip Tetlock as a partner, potentially get his superforecasting team(s) (or build your own - from CoBF participants haha?), make predictions, ..., PROFIT!

 

There are some organizational and methodological questions to be resolved, so this is not commodity-trivial (i.e. you might get competition, but not immediately and possibly not very competitive if you solve some of the issues well). Of course, if you go fund route, money raising would also be interesting challenge. If you go service route, then https://www.completebankdata.com/ might be a good blueprint on how to do it...

 

Fund would probably do much better if this really works.

Service is more inline with what Tetlock does now and perhaps easier to implement / get running.

 

In best case, you could do https://en.wikipedia.org/wiki/Renaissance_Technologies  8)

 

I was think along those lines myself. hmmmmm.

[Jurgis]

I decided that I will make a Brier-scorable shorter than 1 year forecast related to the stock every time I buy or sell a stock.

 

Will figure out how to score adjusted forecasts.

 

Will evaluate Brier score results in 2-3 years.

 

This is not very scientific, since Brier scores are interesting mostly across forecasters. I.e. you may think that you have a great Brier score if you score softball questions. But if you compared it against other people you might discover that it's actually crappy. This is not something that I can solve if I'm the only one making forecast.

 

But perhaps some info is better than no info.

 

Will keep posted on the thread once in a while.

[Jurgis]

BTW, it seems that Tetlock thinks that even superforecasters can't outperform indexes in highly liquid (efficient?) markets:

 

Footnote 10 in chapter 8:

"... even superforecasters won't be able to beat deep liquid markets in which many very smart and well-capitalized traders are relentlessly second-guessing each other".

 

He does say though "That proposition has never been tested, but (sic) superforecasters can outperform shallower and less liquid markets".