Forecasting newsletter: May 2024

Polymarket raised $$, CFTC prohibiting political prediction markets, and an anti-prediction markets lobbying group.

Nuño Sempere

Jun 05, 2024

Polymarket has raised $70M: $25M at some unclear point in the past, and $45M recently.
The CFTC is in the process of banning politics prediction markets in the US.
Manifold might be adding real money payouts.

Index

Prediction Markets and Forecasting Platforms
- The Manifold pivot
- The Polymarket round
- The Kalshi commitment
- The Metaculus leadership transition
Sympathy for the CFTC
- The Better Markets lobbying group
Jobs
Odds and ends

Prediction Markets and Forecasting Platforms

The Manifold pivot

Manifold is pivoting into having a few real-money markets (a) (2 (a)). They are also offering rewards to referrers (a)

As a result, and because the code for the platform is open source, some people are pursuing a fork/rewrite. You can see their work in progress code here (a). There is also a Discord link somewhere in the bowels of the Manifold Discord.

Here (a) are some reflections by Austin Chen on why he is leaving Manifold. He mentions less excitement about 1 to n growth as opposed to 0 to 1 growth, not enthusiastic about the pivot to real money, prediction markets feeling insufficiently powerful, expecting AGI to arrive, focusing on Manifund & public goods funding, and having a baby.

Amana (a) was a way for making anonymous bets on Manifold. The tool is dead now because of API changes, but the website looks cool and I wonder what the person behind it is now up to.

Manifest (a), a ”festival for forecasting and prediction markets”, will be happening relatively soon.

The Polymarket round

Polymarket has raised $70M from VC funders (a) (2 (a)), including Vitalik Buterin and Funders Fund. This is divided in two rounds, a previously undisclosed “series A” of $25M at an unknown point in the past, and a new $45M “series B” now.

News media reports this as coming from “Peter Thiel’s Fund”, but more specifically, this is instead coming from Joseph Krug, the erstwhile co-founder of Augur (a), who basically moved on to other things once Augur’s token got high enough.

On the positive end, the large $70M does guarantee Polymarket’s operations & expansion for the next while. It is also plausibly more money than everything Open Philanthropy (a) has spent on the forecasting ecosystem (!!). I am very glad to see it. Major win for Polymarket.

I am also just following the jurisdictional saga with fascination. The CFTC, perhaps because of Kalshi’s lobbying, banned PredictIt and Polymarket in the US. Polymarket left US jurisdiction. And now that the CFTC is further wanting to ban all prediction markets on politics, not being subject to US law seems like the better move. But at the same time, there is no Great Chinese Firewall for the US, and “the internet interprets censorship as damage and routes around it”, so people who want to bet on Polymarket will be able to. One could say that this was the CFTC’s mistake, because it issued an order that it couldn’t enforce. If this is the end of the story, I admire Polymarket’s chutzpah.

Although this funding takes the form of a VC investment, I’m wondering whether the internal rationale and the success criteria are that of a profit-seeking investment, or that of funding given to a public good. Like, if Polymarket does something new, like ending up becoming sustainable through a DAO to which people donate because they want to keep the public good alive, and because they value the public probabilities, is this fine? Or is this investment meant to capture something like a share of future trading or access fees and be capitalistically profitable? In other words, is Peter Thiel a softy?

On the other hand, though, a friend mentions that the New York Stock Exchange’s parent company has a market cap of $77B. In the optimistic case, it doesn’t seem crazy for Polymarket to reach some fraction of this.

Otherwise, Polymarket had a rules fight on whether one of two necessary bureaucratic steps counts as “approving” Ethereum ETFs—a particular type of investment vehicle. Various sources on this: Domahhhh (a) (2 (a)), The Block (a), and Polymarket itself (a):

This market will resolve to “Yes” if any spot Ethereum ETF receives approval from the SEC by May 31, 2024, 11:59:59 PM ET. Otherwise, this market will resolve to “No”.
The primary resolution source for this market will be information from the SEC, however a consensus of credible reporting may also be used.
[After resolution, a clarification:] The SEC approved eight proposals to list and trade Ethereum ETFs per the May 23 filing here (https://www.sec.gov/files/tm/lk87adfs99.pdf), concluding that the proposals “hereby are approved on an accelerated basis.” Thus this market will resolve to “Yes.”

My take on this is that resolution in this instance seems basically fine. The fear, however, is that Polymarket’s market resolution criteria tend to be on the terse side—as opposed to e.g. Metaculus—and that a contentious enough rules dispute will kill the site.

Polymarket also had some growing pains (a), with a trade confirmed in the UI and not on the blockchain back-end. Donald Trump posted Polymarket odds (a). And Kristi Noem’s chances of VP visibly went down (a) from 10% to 2% (a) after a brouhaha about she shooting her over-aggressive dog.

The Kalshi commitment

Kalshi has gotten Susquehanna (a), a big market maker, to commit to adding liquidity. On the positive side, this is a big deal, as it addresses a practical problem with prediction markets. And making a profit from providing liquidity has been rough for non-pros. On the other hand, from the announcement I can’t quite figure out how much money they are committing. Say 100k shares per market, average price of $0.5 per share, for 270 markets (a), that’s $13.5M of liquidity (?!). You can be cleverer by having a smaller pot and only spending if someone wants to bet. Still, it seems like a pretty big deal.

The Metaculus leadership transition

Metaculus has a new CEO (a). He seems to be doing some cool yet undetermined stuff in Taiwan.Not sure how much it would be good form to say publicly, but expect changes.

This page (a) displays median Metaculus forecasts for AI related questions.

Sympathy for the devil CFTC

The Commodity Futures Trading Commission is in the process of banning election betting contracts, through the process of defining a political contest as “gaming”. You can follow their process here (a), or read a few articles about it (Maxim Lott (a), Reuters, Wall Street (a), Vox (a)).

The statement of the chairman (a) and the two (a) dissenting statements (a) are also informative.

Some quotes from the chairman:

The proposal defines “gaming” and provides illustrative examples of gaming, including the outcome of a political contest, the outcome of an awards contest, the outcome of a game in which one or more athletes compete, or an occurrence or non-occurrence in connection with such a contest or game.
To be clear, that means that event contracts on the outcome of a political contest such as an election could not be listed for trading or accepted for clearing under the proposed rule
To be blunt, such contracts would put the CFTC in the role of an election cop.
Political control contracts on CFTC-regulated exchanges would push the CFTC far beyond this historical expertise and jurisdiction, and potentially place the CFTC in the position of monitoring such markets for fraud and manipulation in elections themselves.

Some quotes from the first dissent:

The Proposed Definition of “Gaming” is Significantly Overbroad
But rather than remain true to the legislative history that equated “gaming” with only sporting events, the Proposal broadly sweeps all “contests” into its definition of “gaming.” And it then concludes that elections and awards are “contests” and, therefore, “gaming” – even though neither Senator Lincoln nor Senator Feinstein ever mentioned elections or awards (or “contests,” for that matter).
But the Commission lacks legal authority under the CEA to make public interest determinations by category.
The CFTC is a creature of statute, and has only the authorities granted to it by the CEA. There is no provision in CEA Section 5c(c)(5)(C) for public interest determinations regarding event contracts involving enumerated activities to be made by category. Accordingly, the Commission cannot claim that authority through the ipse dixit of “Congress didn’t say we couldn’t.”
to say that there are limits to the hedging utility of an event contract is simply a statement that the contract may not be a particularly good hedging vehicle. Market participants should be permitted to make their own choices about what financial products meet their hedging needs. It is not the CFTC’s role to deny them that choice altogether because we feel a given product’s hedging value is “limited.”
The fact that certain portions of the Proposal are inaccurate, extremely weak, or simply make no sense suggests that it either was hastily prepared, or is motivated primarily by the sheer hatred that the Commission seems to bear towards event contracts. Here are a few examples:
The Proposal says (at page 44) that “the public good” is a relevant factor for consideration in an evaluation of whether an event contract is contrary to the public interest. It makes no sense that the Commission should consider “the public good” in evaluating whether a contract is contrary to “the public interest.” This is tautological – “the public good” and “the public interest” mean the same thing.

Some quotes from the second dissent:

The Proposal flatly contravenes Congress’ direction in the CEA that the CFTC “promote responsible innovation.”
the Event Contracts Proposal bigfoots into State regulation of gaming by drawing unintelligible lines in the sand that will either at best result in confusion for State gaming authorities, or at worst push event contracts into illegal, unregulated offshore markets.
Missing Comment Letters
The Event Contracts Proposal completely omits any discussion of the comment letters the Commission recently received on the definition of gaming, as well as Rule 40.11 and event contracts more broadly. All told, the Commission has received around 200 comments in response to requests for public comment on an exchange’s political control contracts.[16] These comments came from exchanges, academics, former CFTC officials, and other industry participants, and were directly on point on the issues raised in today’s Proposal.
The Commission cannot selectively decide to tell one side of the story. It strains credulity that the Commission has selective amnesia and makes no mention of these letters in the Event Contracts Proposal.

Articles on this topic generally report that the CFTC is “in the process” of amending its rules for political prediction markets. But my weak impression is instead that they’ve pretty much reached the conclusion already, and that they’re going through the motions.

The Better Markets lobbying group

Better Markets (a) is an advocacy group seeking moderate markets and financial institutions, and hold them accountable for misdeeds. Ideologically, I disagree with them on prediction markets. However, I find them interesting. Here are three of their outputs I enjoyed:

An opinion piece (a) on the CFTC’s proposed rule-making on US elections
An article on Kalshi’s Dangerous Efforts to Allow Gambling on U.S. Elections (a)
A previous ethics complaint (a) they raised on account of one of the CFTC Commissioners sharing confidential information with Kalshi.

As a background fact, reviewing Better Markets’ Form 550 (a), I found out from their part IV.4 that 503c1 organizations are permitted engage with administrative agencies (a) like the CFTC, and that this is not considered “lobbying”. This is important because my sense is that a few people in the Effective Altruism cluster of nonprofits have been hesitant to advocate more actively in favor of prediction markets because of these concerns.

Jobs

The Forecasting Research Institute (a) is hiring for researcher and assistant roles.
Sage (a) is hiring a software engineer to build forecasting tools, like Fatebook (a) or Quantified Intuitions (a).
Kalshi is hiring (a) for a variety of roles (a).
Polymarket (a) hiring for multiple jobs.
Citi (a) hiring for a “Sr. Quantitative Model Developer – Economic Forecasting”, for $177k to $266k. 6+ years post Master experience required. Outside our normal judgmental forecasting Spiel and more into economics, though.

Odds and ends

Works in Progress (a) makes the case against prediction markets. They are zero-sum rather than wealth-building mechanisms, and they don’t attract gamblers. Subsidies might work, but would be too expensive, and overall “the current size of the prediction market universe reflects market demand”. My sense, however, is that this article reasons about efficient markets too seriously. “The market” hasn’t provided these instruments at scale yet, but “the market” isn’t instantaneous, and entrepreneurs—like Polymarket—are in the process of exploring the space.

Language Models Can Reduce Asymmetry in Information Markets (a). Traditionally, it’s difficult to buy information, because, how can you assess the quality of that information before learning about it? Language models solve this, because they can assess and then forget.

Tom Adamczewski has been working on Carlo (a)—work with distributions on Google sheets, and MakeDistribution (a)—fit a probability distribution to your subjective human judgment.

I have been setting up and running Alert (a)—name change pending—a foresight effort and an emergency response team meant to see large catastrophes coming—recently, solar flares, nuclear escalation in Iran and Ukraine, and the H5N1 flu strain—and to perhaps be able to react to them. One of the artifacts which the project produces are weekly minutes, which you can read here (a).

My formed boss has a few articles out about higher order forecasts (a) and scorable functions (a).

This paper (a) looks at how communicating uncertainty numerically and verbally affects perceived uncertainty of the figures, trust in the numbers, and trust in the source. Perceived uncertainty increases, trust in the numbers is reduced, and trust in the source stays about equal.

My friend Marwan Riach (a) has been modeling UK elections (a) and attained some fame and renown by being sharper than the competition. UK readers might be interested in following him.

Weather forecasting in Africa (a) is underdeveloped, making them more vulnerable to severe weather events.

I found an INFER (a) report on Iran nuclear weapons frustrating. I like that it exists, and it might be better than other options. But it’s also underwhelming. I would be hugely surprised if INFER couldn’t do better.

A news-wire reports an inventory forecasting acquisition (a). It doesn’t mention what the company was acquired for, but the area has been smelling profitable to me for some time.

I liked an article on Bloomberg (a) arguing for the US Federal Reserve to shift from point-based forecasts to more complex scenario analysis.

Note to the future: All links are added automatically to the Internet Archive, using this tool (a). "(a)" for archived links was inspired by Milan Griffes (a), Andrew Zuckerman (a), and Alexey Guzey (a).

Y’all ever thought about how many wild missions we’ve been on, and we somehow always survive?
Heh, yeah, so far so good, I guess

—Dialogue in Fast and Furious 9, around the 35’ 50’’ mark.

Forecasting