Epistemic status: I probably have no idea what I'm talking about.

Prediction markets have several constraints due to laws against online gambling (at least in the US). PredictIt, for example, must operate as a nonprofit, limit each contract to 5,000 total traders, and limit investment to $850 per contract. [1]

Given the usefulness prediction markets, it would be great if we could remove these contraints on financial incentives for the users and operators of these platforms. [2] Currently, we rely on fake internet points for incentive on websites like Metaculus; although, people do seem to like those.

So, I tried to think of an alternative model. In a betting market, the winners are paid with the losers' money. Instead, imagine the prediction platform is a bank. Forecasters deposit money in the bank, and are given ballots to vote on questions. The bank, as banks do, lends that money out to earn interest. Then, interest earned from all forecasters' deposits is paid to only to the winners. Winners win by earning interest from others' money, and losers lose by opportunity cost and inflation.

Legally, I don't think this counts as online gambling. This would be like if your regular bank only let you earn interest on the money in your account if you can guess how many jelly beans are in the jar in the lobby.

Here's a simple version of the model. Forecasters buy 1-year CDs from the prediction bank, with the stipulation that they only earn this interest if they get questions right. The bank buys a bunch of US 1-year Treasury Bills with the money deposited. At the end of the year, the earned interest is divided among winners for each question, minus what bank keeps as profit.

There are a lot of variations for this model, but hopefully you get the general idea. Note that risk for forecasters is limited relative to the betting market approach, thus constraining both profits and losses. Also, there's the risk that the bank makes a bad investment and cannot pay out interest.

I think it could work, but I'm pretty confident I'm missing something here, so I wanted to share it for feedback. Let me know if (a) I got the economics of this right, and (b) it seems feasible. 

  1. ^
  2. ^


Sorted by Click to highlight new comments since: Today at 5:41 PM

I think another possible route around gambling restrictions to prediction markets is to ensure all proceeds go to charity, but the winners get to choose which charity to donate to. I wrote about this more here:


Great idea! Makes me think, it would be interesting to see a political prediction market where the winnings go to your preferred candidate in the race. Not sure about if that would have a positive impact, but would be cool to study.

Edit: Just read your post and see that you discuss this haha

This pays far too little to the winners to make it worthwhile to have any money in this. It wouldn't have much more liquidity than a moneyless prediction book.

That entirely depends on the return on the bank’s investment, right? I have no idea what that could be in practice. If it were similar to the stock market, say 8% annually, then I think that would be very attractive to forecasters. Being right on a poll where 50% of respondents were also right would be like doubling what you expect to earn from stocks. But obviously that’s risky investing, so probably not feasible. Or if you were already a big bank, you could afford such risk, and make it worthwhile for winners. Was that what you were thinking?


In the crypto world, Hedgehog Markets is built around this concept - you stake your money in a tournament, and then bet with their play money; the staked rewards go to the winners, and everyone gets their initial money back.

At https://manifold.markets/ we elected to start with fake internet points, but hope that careful rationing of the fake points can make them valuable the way in-game currencies can become valuable.

One more spinoff of your idea - if the information provided by the prediction market is valuable enough, perhaps the platform could pay out without ever having to take in money, and not qualify as gambling. Eg if the platform sells early access to market data to a hedge fund and distributes the proceeds to its users?

Oh wow, that's cool! Do you know how Hedgehog invests the play money?

Your last idea is a lot like a company giving bonuses to internal forecasters, seems promising. If prediction markets prove themselves worthy, maybe an EA organization will eventually decide it's worth it to regularly sponsor forecasting tournaments for global priorities work.  I'm excited to see where the space goes.

Came here to say this.


This is a good effort, but I'm not sure if the return would be worth it. You said stock market in a different comment -- yes, it returns 8% annually, on average, but what happens when it returns -10%? You're subjecting the forecaster's to investment risk, whereas what you want to do is subject them only to the forecasting risk (and compensate them appropriately).

CFTC regulations have been at least as much of an obstacle as gambling laws. It's not obvious whether the CFTC would allow this strategy.


You can build a blockchain-based prediction market like Polymarket from outside the US - if the restriction is only developers not users. For users there's Tor + Tornado cash and that make your funds and activity much harder to trace.

In theory the developers can even be anonymous but you want someone public in a country where they can be sued, to ensure they don't feel false results into the market.

One significant challenge is getting liquidity on new markets. Paying liquidity providers is the obvious solution, but this requires the platform to eat losses and I'm not sure they can recover them from fees once that market matures, for most markets.

It seems like there is a lot of value in creating something that feels legitimate. Ideally more than just a niche group of EA/Rationalist folk would be interested in participating, and I don't think a "the feds would never know" argument would be convincing for a big audience. Thanks for sharing those examples though!


Fair enough.

P.S. Your suggestion could also be done by non-US developers on the blockchain, since it's legally ambiguous. That way the developers can't be sued, it's mainly the users which is harder.