Manifold was fairly incorrect in its predictions for the 2024 US presidential election. Why was that?
Manifold was considerably worse than real-money markets such as Polymarket, and very similar to forecasts like FiveThirtyEight and Nate Silver. It seemed to notably underestimate Trump's performance, especially in swing states (Manifold had Harris winning MI at about 65% on election day). I have noticed a very large sentiment in the Manifold community that is eager to disregard the accuracy of those who did make better predictions, oftentimes dismissing them as "lucky guesses". Is that really the case? Can we assume that Manifold will do better next time around, or has real-money betting taken over as the most reliable predictor for future elections?
How this will resolve:
Until 1/20, traders can bet on the options provided and will be able to add their own options (I will prevent the adding of options when it reaches 25). Once the market is closed, I will make one comment for each option on the market. The top three comments in like margin (likes - dislikes) will resolve YES and the remaining options will resolve NO.
Inspired by:
Thank you for participating! DM or comment with any questions.
Update 2025-21-01 (PST) (AI summary of creator comment): - Like/Dislike Period: Until January 31, users can like or dislike any of the creator's comments based on their contribution to Manifold's incorrect prediction.
Resolution Criteria: At the end of the month, the three options whose comments have the largest net likes (likes - dislikes) will resolve YES. The remaining options will resolve NO.
1,000🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ135 | |
| 2 | Ṁ127 | |
| 3 | Ṁ84 | |
| 4 | Ṁ74 | |
| 5 | Ṁ63 |
@traders The market has closed! Until 1/31, you can like or dislike any of my comments based on whether or not you felt they contributed to Manifold's incorrect prediction. At the end of the month, the three options whose comments have the largest net likes (likes-dislikes) will resolve YES. The remaining options will resolve NO. Thank you again for participating!
@vibhav What is meant by this? Is it something more like A) Disinformation caused Trump’s election, B) Disinformation gave a false impression of Harris’s chances, or something else?
@MikeElias Definitely B. This suggests disinformation created false impressions of who was in the lead, causing Manifold to predict incorrectly.
@barbarous I am not fully educated on what exactly happened here. Is such accurate information actually available at a cost?
@vibhav by "at cost" I mean he conducted his own survey using a method that is thought to be more accurate under preference falsification. I.e. instead of asking "who will you vote for?" he asked "who will your neighbour vote for?"
@barbarous very interesting. can this consistently be reliable or did it just work in this case? if the former, why are we not relying on methods like this to forecast?
@vibhav I don't really know, but he's ended up pushing the polymarket odds by a fair amount in the correct direction and won a lot of money. So it worked for him and it worked in this case.
He also had a strong hunch that polls and other models were still underestimating trump's advantage. That's not exactly a formal model that you can apply to every case, but the beauty of prediction markets is that you can incorporate information like that and have it pay off if you increased accuracy