Suppose that you've got a free response market about which thing will score highest on some numerical metric, but when it's time to resolve the market two or more options have scored exactly the same. What do you do?
Prompted by discussion on https://manifold.markets/Multicore/what-is-the-best-strategy-in-ddsci
For this market I will resolve choosing all serious answers with their default payouts.
Resolve to all tied answers with equal payout weighting.
Chosen 77%
61%
Resolve to both the tied answers and the separate answer for a tie, with a predetermined weighting
Chosen 11%
9%
Resolve to all tied answers with payout weighting proportional to the market probability at resolution time.
Chosen 7%
5%
Have a separate answer for a tie, and resolve only to that.
Chosen 6%
4%
Multicore answered
Resolve to all tied answers with equal payout weighting.
61%
Adrian Kelly answered
Resolve to both the tied answers and the separate answer for a tie, with a predetermined weighting
9%
Multicore answered
Resolve to all tied answers with payout weighting proportional to the market probability at resolution time.
5%
Multicore answered
Have a separate answer for a tie, and resolve only to that.
4%
💬 Proven correct
Matt P bought M$30
@Multicore only right answer IMO. Picking "tie" punishes the people who correctly picked the winners in their individual answers, and picking tied answers to PROB is resolving to the one people spent the most money on, instead of resolving based on the original criteria (aka the winner). Keynesian beauty contests are a bad way to resolve free response markets.
0
Matt P made M$24!
Multicore answered
Resolve to all tied answers with equal payout weighting.
61%
Matt P bought M$30
@Multicore only right answer IMO. Picking "tie" punishes the people who correctly picked the winners in their individual answers, and picking tied answers to PROB is resolving to the one people spent the most money on, instead of resolving based on the original criteria (aka the winner). Keynesian beauty contests are a bad way to resolve free response markets.
0
Adrian Kelly answered
Resolve to both the tied answers and the separate answer for a tie, with a predetermined weighting
9%
Adrian bought M$10
I agree with Matt, and this isn't useful in all markets, but in markets like the NATO one, this method allows people to express whether they think X and Y are independent or linked.
0
The market creator should specify "all tied answers with equal payout", "all tied answers plus the combined answer with equal payout", or eg "50% to the combined answer, 50% split between tied answers", depending on how how linked the market creator thinks the answers are. But the most important thing is market clarity.
0
Multicore answered
Resolve to all tied answers with payout weighting proportional to the market probability at resolution time.
5%
Multicore answered
Have a separate answer for a tie, and resolve only to that.
4%
General Comments
Here's another example of this being relevant:
https://manifold.markets/Duncan/i-will-select-all-options-with-grea
0
https://manifold.markets/BoltonBailey/what-will-be-the-next-country-to-jo is a currently open market that currently has an explicit tie resolution policy. Are there any others?
0