Context

Most of the value in prediction markets is not in the gambling side of them, but on the information embedded in the markets.

A prediction-market platform that can produce forecasts-as-a-service would have enormous value. There’s people willing to pay for accurate forecasts to their questions. However, neither Manifold nor other markets currently solve this problem at scale.

Problem Statement

• In Manifold Markets, the questions which are most relevant to the userbase get the highest betting volumes. It incentivizes markets on:
  • Topical issues (current examples: Elon’s twitter acquisition, Russia-Ukrainian conflict, stable-coin crashes, etc).
  • Topics relevant to the forecasting userbase (Effective-altruism related markets, AI-risk markets, meta-markets about Manifold/PredictionMarkets, etc)
  • Personal markets made by people known to the userbase (Scott Alexander, cofounders, Aella, Nuno, etc).
  • Mega-mainstream stuff (Sports, presidential races, etc).
• But markets that don’t fit these categories get barely any betting volume.
  • Markets narrowly relevant to the creator.
    • Ex: Will flight Will VIV 3109 be on time on March 19?
  • Markets that are broadly-relevant but not to the existing userbase.
    • Ex: Will Jay Shambaugh become the next Treasury Secretary for International Affairs?
  • Anything in between
    • Ex: Will DearMoon launch in 2024 (1 year behind scheduled date of 2023)
    • Ex: Will a new Build for Tomorrow podcast episode be released this Thursday, April 28th?
• The creator can add liquidity to a dead market, but it only weakly incentivizes bets, and it makes the market less accurate.
  • Ex: Will I quit my SWE Job at Meta/FB to work at Manifold Markets?
• There should be a way for any user to reliably get high-quality forecasts for any question they ask.

The Idea

MVP: Play-money prize pools

• Allow market creators to add a $M prize-pool to their questions.

  • Allow any other user to donate $M to the prize-pool as well.

• Distribute the prize-pool among the people that bet correctly (proportionally)

  • Naive Example (doesn’t necessarily have to be this mechanism)
    • In a particular market, Alice won 50$M, Bob won 20$M, Charlie won 30$M.
    • But the market had a 1000$M prize-pool, which also gets distributed proportionally at closing time. So Alice additionally gets 500$M, Bob 200$M, Charlie 300$M.

• Benefits:

  • Lets market creators incentivize betting as much as they want.
    • The larger the prize-pool, the more upside for bettors without any extra risk.
  • Bettors are still incentivized to be forecast accurately.
  • Potential growth in the number of markets created daily, as the platform can now reliably answer more questions.

• Open questions, limitations & Problems

  • Relevant: Zvi’s post on Subsidizing Prediction Markets
  • What mechanism we can use to distribute the $M is the big question, and the one I’d want to discuss with the founders
  • Relevant:

  

  • One mechanism would be to distribute the prize-pool at resolution time, proportional to earnings
    • Toggle for example
    • But users wouldn’t be able to know which percentage of the prize-pool they’d get until resolution time (as more bets from other people could come in later)
  • It might not increase the total number of bets on the platform, just redirect them to questions with the largest prize-pools.
  • Game theoretic issues depending on the mechanism.
    • Depending on the mechanism you might be incentivizing accurate predictions to a different degree
    • Example: distributing the prize-pool equally among everyone that bets incentivizes low-effort predictions

v2.0: $USD prize pools

• Allow Real-dollar ($USD) prize pools instead.
• Distribute the $USD prize-pool among the people that bet correctly (at resolution time, proportionally to their bets).
  • Example (with naive mechanism)
    • In a particular market, Alice won 50$M, Bob won 20$M, Charlie won 30$M.
    • But the market had a 1000$USD prize-pool, which also gets distributed proportionally at closing time. So Alice additionally gets 500$USD, Bob 200$USD, Charlie 300$USD.
• Benefits:
  • Lets market creators incentivize as much betting volume as they want.
    • Think 10USD prize-pool, vs 1000USD, vs 100K USD.
  • Grows DAU.
    • The larger the prize-pools in the website, the more “free-money” lying around. Users will flock to the platform to predict, and get money.
  • Compensates forecasters for their work.
• Limitations & Problems
  • The same problems as with the MVP would apply
  • Creates a strong incentive to create multiple accounts and use the free 1000$M repeatedly.
  • You probably (?) run into gambling regulators if you allow people to bet Mana they bought with USD, rather than earned