New forecasters guide 🎉

Do you want to forecast stuff? You’re in the right place!

I don’t know how!

First, don’t worry! This is actually very simple.

Overall, the process is forecast, talk, update forecast.

First, get a forecast in. This is easy:

  • All you’re being asked is to type some numbers.
  • You won’t have to show or discuss your work.
  • You’re not committing to anything long term.
  • Forecasts are default anonymous.
  • You will finalize it later, so just get something down.

Next, we talk about our forecasts. Talking is optional! You can lurk where the discussion is happening instead. What is important is that you think critically about the problem. Review other perspectives. Contributing to the discussion helps the whole process and is very much welcomed. You can share your actual forecast, or just your reasoning, or discuss points others have brought up.

After you’ve had time to work through the problem with others, you have a chance to update your forecast.

Here’s some discussion on the numbers you might be asked for.

If you’re being asked for an interval: That’s two numbers, a maximum and a minimum, assuming some percentage of confidence that the future value will land within it. That confidence level (like, say, 90%) is usually assigned and shared between everyone.

For example: Jelly beans in a jar: 100 (a minimum) to 200 (a maximum).

If you’re being asked about a scenario: That’s few probabilities that you believe some outcomes will happen over others.

For example: A 15% chance it will rain tomorrow.

You’ll probably be given a link to a forecasting tool (like e6e) or a spreadsheet to collaborate in.

There’s likely going to be some discussion about people’s forecasts that you can participate in as well. You might want to update your forecast after seeing discussion, and that’s very much encouraged.

What happens next?

We wait for the actual outcome to occur. Then, we take a look at our individual and average forecasts. Hindsight analysis allows us to better understand what our beliefs were, why they may have been wrong, and how to improve them in the future.

Some analysis will probably be written about the discussion, the measurements, and the actual outcome. These are usually shared more widely, while hiding individual participants (unless they want to step forward and discuss their forecasts).

I want to do this more!

If you’re not added to the Substack, enter your information here.

← Measurement
Risk →