Getting a MultiChoice Forecast Just Right

This is a follow-up to Multi-choice “Publication” Markets.

This extended example is just to give you a better feel for linked multi-choice questions. It’s probably a bad idea to forecast like this because:

  • I’m making big, expensive trades for dramatic effect. Unless you are either impatient or skilled, make smaller moves first, and see how the market reacts.
  • The best strategy is always to fix the most-wrong. After the first trade below, all three probabilities move towards your goal, so it’s likely some other preprint question is now most-wrong.
  • Getting it just right is a waste of time until the end: someone will just come by and change it on you. 

That said, let’s go.  Here’s a preprint with the three Publication claims – High / Low / No – starting at ~33% each.


Without even looking at the claim, I decide they should be 10% / 40% / 50%.

Using Quick Forecast for speed:

  • 1) I’ll raise No to 58% – overshooting a bit because I know Step 2 will reduce it.

That will drop the others to 21% each. (Wait a few seconds for the refresh.)

  • 2) Next, I’ll raise Low to 40%. That will reduce both High and No, proportional to their values, so High will reduce by more — that’s why I overshot in Step 1:
  • 3) Nuts, not close enough. I now think High is the most wrong -> it should be near 10%.
  • 4) Good enough!
  • 5) No, wait, I can’t put it down. My next move will be No 47% -> 50%. The extra three points should go mostly to Low which is 4x as large as High. Watch:

              Not perfect, but this time it’s good enough.

The end result is I have the following shares (use Forecast or Quick Forecast to see):

High: 77 “No” shares.


Low: 129 “Yes” shares


No:  163 “Yes” shares.


Charles Twardy

There is NO COST to participate in Replication Markets.

This research is supported (in part) by the Fetzer Franklin Fund of the John E. Fetzer Memorial Trust. It uses a platform developed for DARPA SCORE, and some staff are supported by SCORE while working on this. We are grateful for their support.

We’re sorry to see you go! Please visit our social media sites.

This site uses cookies to provide you with a better browsing experience.

Visit our Privacy Policy for more information.