A Contrarian Sort
By Adam Siegel on May 03, 2022
Questioning the assumptions and probabilities of the consensus is a simple
best practice of forecasting. Do I currently agree with the prevailing winds,
or do I predict something different will occur?
To find people who are currently disagreeing with the consensus, you'd
typically have to scroll through the activity feed of a question which, if
there are a lot of forecasts, can make your eyes hurt.
Instead, we've recently introduced the "contrarian sort." Here's how it works:
- We start by computing the difference between each person's forecast in that question and the current crowd forecast
- If there are multiple answers, we sum the difference of each
- We then sort by the total difference, with the biggest being at the top
As an example, in this question on the INFER platform: Will clashes between Chinese and Indian security forces result in 25 or more fatalities between 28 April 2022 and 1 April 2023? the current consensus as of 3 May, 2022 is a 9% likelihood. That's pretty extreme, so let's see if there are any contrarians, by using our contrarian sort. And indeed...
We can clearly see there are some who recently forecasted with a much
higher likelihood this would occur, which is very useful as you're
considering your own forecast. In the future we'll likely introduce a
time element as well, so you can see "recent" contrarians vs.
contrarian perspectives that may simply be out of date. For now though,
we're already getting positive feedback from people saying: "huh, I
wasn't considering that" which is exactly what we were going for!
To stay updated on what we’re doing, follow us on Twitter @cultivatelabs.
You may also be interested to read How We're Assessing the Credibility of a Crowd Forecast.