Long Live Polymarket

Polymarket got fined $1.4 million by the Commodity Futures Trading Commission and was ordered to cease noncompliant trading in the US.

Polymarket is probably the biggest prediction market currently available. US law considers unlicensed prediction markets to be somewhere between illegal gambling and illegal futures trading, ie definitely illegal. Polymarket and a few peers had survived anyway, through the “crypto is the Wild West and nobody has time to deal with all the illegal things happening there” exemption. Apparently they found time.

The rumor on the prediction market grapevine (which I absolutely cannot substantiate; please don’t sue me for libel) is that this might have something to do with competing prediction market Kalshi. Kalshi spent two years and probably a lot of money getting the CFTC to agree they were legal, and has a former CFTC Commissioner as a Director. Their legal status forces them to do an annoying and expensive regulatory dance all the time; illegal prediction markets were able to move more nimbly, provide better user experience, and eat their lunch. This was a big problem for them - but they’d just finished making lots of friends in the agency that decides which illegal things to crack down on, so, as Tyler Cowen likes to say, “solve for the equilibrium”.

For its part, Polymarket was an easy target. Despite “decentralized” being the fourth word on their website, they weren’t exactly at Satoshi levels of opsec. They had a normal identifiable guy as CEO, a normal headquarters building in New York City, and they did normal business things like raise millions of dollars in seed funding.

Some might call a headquarters building with a CEO sitting in it and millions in the bank account a “center”, so in what sense was Polymarket decentralized? See here for more discussion, and here for the full text of the CFTC decision, but my understanding is - all of the markets themselves were smart contracts on the blockchain run by automated market makers, but you could only access them through the Polymarket website, and the Polymarket people decided how they resolved. Polymarket did not charge fees, and made money by providing liquidity. The CFTC seemed angriest about the “you can only access contracts through the Polymarket website” part of this. Crypto attorney Collins Belton writes:

It’s hard to assess which factors were most aggravating and most mitigating from CFTC’s perspective. For instance, it’s hard to assess whether Polymarket may have been okay if its agents didn’t engage in any [liquidity provider] activity and operated no [front end].

So: Polymarket got fined $1.4 million, and was ordered to make its real-money markets inaccessible to US-based traders (the rest of the world is still fine). It’s v ery p oor n ews to hear that a v illanous p olitical n onentity blocked this v ital p rediction n exus, and I guess we Americans have no other options besides accepting that we’re v astly p oorer n ow.

Meanwhile, Polymarket put out a rainbows-and-butterflies press release saying that:

We are excited to continue championing our mission and building out our global footprint, information and educational initiatives, and U.S. product

I assume this means they’re excited to continue building their prediction market somewhere else, and will include a US version with play money, just like lots of other companies have done.

I Will Limit My Outrage To Four Paragraphs, Then Move On

My favorite commentary on this decision is Nuno Sempere’s The American Empire Has Alzheimers. He lists various bad decisions the US has made, from Vietnam to the bungled withdrawal from Afghanistan last year. In this last case, President Biden said there was “no circumstance where you see people being lifted off the roof of an embassy” barely a month before we saw exactly that.

Throughout these bad decisions, intelligence analysts and national security advisors were begging the government to come up with some kind of good forecasting infrastructure. By the early 2000s, many of them had settled on prediction markets as the most promising opportunity. In 2008, twenty-two prominent economists including five Nobel Prize winners wrote an editorial begging the CFTC to legalize prediction markets; the CFTC refused. In 2010, Philip Tetlock (one of the signatories on the pro-prediction market letter) did some pretty basic forecasting work, not even prediction market level, and proved that he could significantly outperform top analysts at the CIA with access to classified information. The government refused to hire him or use any of his methods, and continued shutting down new prediction markets as they arose.

Starting a few years ago, cryptocurrency provided a brief “thaw” when people thought they might be allowed to try innovative forecasting mechanisms. They tried, they created really impressive work, they made (and deserved) millions of dollars, and then the government kicked them out of the country anyway.

The US is becoming the North Korea of forecasting. Every other civilized country allows prediction markets. In a perfect world, they could ignore our constant own goals and move on without us. But because America has a disproportionate share of money, users, coders, and entrepreneurs, a US-less prediction market ecosystem won’t be living up to its potential. That means decreased ability to gather and process information and worse decision-making worldwide.

Where Do Prediction Markets Go From Here?

…aside from “to other countries”.

I think there’s a general sense among people interested in the field that prediction markets are vastly underperforming their potential. There ought to be a billion dollar prediction market, maybe a ten billion dollar one. Smart VCs clearly believe something like this, or Kalshi wouldn’t have gotten $30 million+ in investment. Sometimes people who incorrectly believe I know things about prediction markets ask me if know the missing secret sauce.

I don’t think there’s any secret. A prediction market will strike it big when it gets three things right at the same time:

  • Real money

  • Easy to use

  • Easy to create your own subsidized markets

“Real money” should be self-explanatory. Metaculus and Manifold are both very nice, but so far they’re limited to a small group of enthusiasts playing in their spare time. I value them both, but neither is the killer app that makes prediction markets as central to everyday life as stock markets or polls or whatever.

“Easy to use” is kind of self-explanatory, but with some caveats. A big part of ease-of-use is liquidity; you can get that from a big user base or from clever deployment of automated market makers. A market that requires crypto knowledge is harder to use than one that doesn’t; one that’s inaccessible from the US is harder to use than one that isn’t. Also all the normal things like UI and search.

“Easy to create your own markets” is where we’ve gotten stuck so far. Prediction markets are absolutely on top of questions about whether Donald Trump will win various elections. This is a solved problem. What I really wanted last year (and would have subsidized!) was a market about whether Alameda County, California, would permit indoor gatherings of 50 people on January 8th 2022 (ie would I be forced to cancel my wedding). But I also would have appreciated the ability to put a few questions to prediction markets before starting my psychiatry practice, or my grants program, or any of a dozen other things I did. A friend has gone further, and half-jokingly said they want to create conditional prediction markets about whether they’re compatible with various women in our friend group, to be paid out six months after the first date.

Some of these applications are attempts to route around the principal-agent problem. Maybe I have some question about whether a certain grant would succeed, I’m not sure who to ask, and even if someone gives me a “Bob Smith, Grant Evaluator” business card, I don’t know if he’s any good. A prediction market takes all the pain out of searching for information - if I subsidize it enough, it’ll attract people with the relevant skill set who will solve my problem for me.

Probably some of these ideas wouldn’t work, but probably other ideas I can’t even think of now would. I don’t know what the killer app for prediction markets will be. But we’re not going to find out unless people can create their own subsidized markets and play around.

Polymarket took some baby steps towards this before the settlement: they had a Discord server where anyone could propose questions, and a lot of those questions became markets. But they still had to be general interest, not “let Alice’s five friends predict her dating life”. And there’s a big difference between “talk it over with company representatives on a Discord server” and “press a button”. Imagine if you could only tweet by emailing Jack Dorsey and convincing him that your comment was a good thing to have on Twitter. Even if Jack had good judgment and approved most requests, this would be a long way from the limbic system < — > Send Tweet loop that real Twitter users know and love.

I asked some people in the business why they won’t do this. They said most people are bad at writing good resolution criteria. They don’t want their employees to get stuck resolving incredibly dumb questions about people’s dating lives, hunting down inaccessible or conflicting information, and making a bunch of people mad whichever way they decide. As far as I can tell, Manifold Markets solved that problem with their “proposer decides the resolution, caveat emptor ” strategy. But Manifold is US-based and can’t use real money, so there’s still no way to subsidize a market effectively.

(This is why I’m pessimistic about Kalshi. They could potentially do a lot of good in the “will Afghanistan collapse?” types of markets the Nobel laureates want, though even there I think some of their betting limits will give them trouble - $25,000 is good money, but not quite good enough to incentivize founding the prediction market equivalent of a Wall Street trading firm. But even if they solve this, I can’t imagine the regulators giving them permission to host “will this grant work out?” or “how will my dating life go?” markets; it’s just too weird, and the CFTC is too conservative. I don’t know, maybe their connections will come through and pull it off, but I don’t even know if they’re ambitious enough to want this, and I hate having to rely on one organization.)

Right now my hopes are, in ascending order of likelihood:

  • Manifold figures out some kind of weird crypto thing that isn’t real money from a legal perspective, but is real money from a “people really want it and will put a lot of effort into getting it” perspective.

  • Polymarket does this outside the US, and it succeeds so wildly that everyone agrees we need a US version.

  • Someone creates a genuinely decentralized prediction market which the CFTC either believes is legal, or can’t figure out how to shut down.

I’m most optimistic about this last one, but it would be tough. You could try a version of Polymarket without the centralized organization gating the front end and providing liquidity. But then how would it make money? It probably wouldn’t - which might be fine, Metaculus is a non-profit and is still exceptionally well-run and stable. If someone did a good job of this I would try really hard to get it funded, and would expect to succeed.

But what if the CFTC says no, they’re still angry?

In Search Of Lost Crypto Dreams

Sometimes it seems like everyone who’s ever used crypto has a different model of why it might be good. Here’s mine.

There are lots of financial products which people want, but which regulation prevents them from having. Some of these are totally without social value, like Ponzi schemes and Bored Apes. Others have a lot of social value, like prediction markets, initial coin offering style funding schemes, and cutting middlemen out of immigrant remittances. More than a few might even have negative social value, like easy ways to buy drugs, or super-high-interest loans marketed to very impulsive people. Without passing judgment on whether these things are good or bad, they are legion. Collectively, they’re a zillion-dollar market.

In theory, crypto is hard to stop and hard to trace (yes, I know the blockchain is public, but I also know that TornadoCash and Monero exist). Anonymous users could create these services so easily and in such numbers that governments would never be able to stop all of them. Or they could be run as smart contracts, where even if regulators arrest the original programmers, they can’t stop the program from existing and continuing to offer its service.

This vision is a lot like the original vision of the Internet: a magical place that nobody could censor, where information would flow freely across national and ideological borders.

That vision was . . . maybe 25% achieved? It’s pretty great that I can write a blog like this instead of begging for my supper at a major media organization. But after a brief period of discombobulation, dictatorships found it easy to create their own walled-garden Internets through light-touch censorship; although there are ways around most of their tricks, ordinary people don’t bother with them (v ery p oor n ews indeed!) And in practice most people ended up basing their Internet explorations at a few big businesses like Google, Facebook, and Twitter, which became easy prey for censors and in some cases rush to self-censor even more zealously than governments demand. It’s not that the Internet can’t create a magical censorship-resistant infrastructure, it’s that it’s 5% easier to sell your soul to FAANG, and so many people take that option that the few people who don’t aren’t really a critical mass for escaping governments or building new communities.

Is crypto following the same path? I am a lazy crypto user, and I notice that nowadays I have to grovel harder to the government to access my crypto than to acces my fiat - every crypto site is a gauntlet of “please tell us your name/birthdate/SSN/address”, “please upload your ID”, “please wait for us to complete our KYC process before trading”, etc. I assume there’s some really flashy way that cool people use their crypto without doing any of this, but there’s also Tor and Brave and Mastodon, and that doesn’t mean the Internet is a free speech privacy paradise.

But then, what’s the point of crypto? I mean, right now the point is to be a Ponzi scheme, and it works great; 1000% annual returns are a perfectly adequate substitute for there being a point. But most people think crypto only has another 0 - 2 doublings left in it; after everyone who’s going to invest invests, what then?

In the early days, you could make a small fortune by buying Bitcoin and holding, or a large fortune by minting a new token and telling other people to hold it, or a gigantic fortune by creating the infrastructure people needed to do A or B. All these activities are more or less legal and there’s no point in angering the government while you do them. And brilliant coders generally wouldn’t work on the illegal projects when it was so easy to make money doing the legal ones. So everybody rolled over and let the government regulate them, because why not?

At some point, the Ponzi-ing will run out. Optimistically, the brilliant coders will need something else to do, and someone will try creating things like genuinely decentralized, impossible-to-shut-down prediction markets. I feel like this should be possible, but I also don’t understand why it hasn’t happened already. Maybe the technical challenges are too hard.

Pessimistically, by then the crypto infrastructure, crypto social norms, and crypto user base will be so comprehensively locked into the current regulated model that this won’t be able to get off the ground. You’ll try to use Coinbase to send your crypto the prediction market, and it will warn you that this is a Non-Preferred Site that isn’t a Coinbase Partner and they’ll be informing the IRS of this transaction so don’t try anything funny. A few smart people will know ways around this, and everyone else will just suffer.

Will crypto ever live up to its potential? Only time will tell - since they’ve banned every other method of forecasting things.