CM 004: A New Home for PMs We DAObt You Saw Coming

Plus a useful tool for every prediction market trader, and the sketches of a Superforecaster aggregator.

❤️ Please click the little heart at the top of the post—it’s like a tip, but free!

✍️ Share your thoughts with us about the issue in the comment section.

📧 And a warm welcome our 13 latest Crowd members who joined since the last podcast! Please don’t forget to tell your friends to #JoinTheCrowd!


💰 Welcome back to Crowd Money, the monthly newsletter covering real-money prediction markets (PMs). Two weeks after the newsletter we publish the Crowd Money Cast, a podcast where we discuss the latest prediction market news and interview key players in the space. Last episode we spoke with George Yu, CEO of Hedgehog Markets—a crypto-based prediction market on the Solana blockchain featuring innovative no-loss markets.

In Crowd Money 004:

  • 🔨 Useful tool for all prediction market traders and forecasters;

  • 🏊 Deep dive into the rise of DAOs and how PMs can support them; 

  • 💊 Interview with the Founding Steward of VitaDAO about the future of decentralized science and the prospects of PM integration in DAOs;

  • 🔭 Initial thoughts on a superforecasting aggregator for PMs and potential applications.

🔨 The Good Kind of Meta

While prediction markets and prediction polls have been around for decades, their popularization increased dramatically in the past half-decade-or-so following notable catalysts such as Phillip Tetlock and Dan Gardner’s Superforecasting, the release of online platforms such as Metaculus and Good Judgement Open, and the expansion of prediction markets vis-a-vis new entrants such as Polymarket and Kalshi.

Due to this fragmented ecosystem, a lot of useful information about improving performance on these platforms is decentralized: Reddit, Twitter, and in-person conversation are all common ways that prediction market insights are communicated. As a result, things that we consider to be common knowledge remain unknown to many. This is why, when it makes sense, we want to highlight useful tools and practices for traders and forecasters that we’ve discovered in the newsletter.

In this issue we will focus on a high-impact, low-friction tool that can benefit prediction market bettors and forecasters alike: Metaforecast.

What is Metaforecast?

Simply put, Metaforecast is a search engine for predictions. There’s a search bar where you type in a topic and then Metaforecast looks across a variety of different prediction markets and polls, as well as forecasts made by certain individuals and on betting exchanges, and displays the (mostly) relevant results to you. The site also filters predictions based on their quality.

Let’s say you’re interested in trading on PredictIt’s market on who will be the next chairperson of the U.S. Federal Reserve. Given the mainstream nature of this question, it would be a good guess that other platforms will have similar questions. Instead of going to each platform (assuming you can always remember the growing list of them!), you can go to Metaforecast, search ‘Powell’ (finding the correct term to search can be tricky, but results load fast), and receive this output:

You instantly see how traders on Kalshi, Smarkets, and Polymarket are feeling about the question, as well as the superforecasters at Good Judgement (which is different from GJOpen). The site is updated daily. Clicking on any of the boxes will open the relevant market link in a new tab.

You might also use Metaforecast to discover tangentially-related predictions to the main question you are interested in. This can be helpful as inputs or guidance for your primary prediction; e.g., to forecast the 2024 Presidential Election, questions on potential candidates might be helpful.

Metaforecast is useful, and simple to use (incl. mobile) so give it a shot! The back-end and front-end code are also publicly available. The project was developed by Nuño Sempere, with help from Ozzie Gooen, and is part of the work produced by the non-profit, EA-aligned Quantified Uncertainty Research Institute.

Share

🏊 Incentives & operationalization

In the first three issues of Crowd Money we focused on the potential for prediction markets to produce actionable information; we explored methods of facilitating the generation of accurate predictions by way of incentives, both monetary and non-monetary; and we have also written about traditional market makers and automated market makers as tools to fuel liquidity and trading volume, arguably two of the most indicative metrics for prediction market platform success. But once a framework for accurate prediction markets is established, how can it be optimally leveraged?

Today we will be exploring a new prediction market incentive, one that has the potential to eclipse the incentives we dug into last month. This incentive framework also takes predictions one step further than question resolution, and does what hasn’t yet been done in the policy realm: operationalize PMs to inform decision-making. 

What do you GNO about Futarchy?

Our journey starts here: Almost exactly one year ago, CoinDesk published an article about GnosisDAO, a DAO launched to let Gnosis (GNO) users provide input on governance decisions for the currency. 

  • Gnosis is a PM built on the Ethereum blockchain. It supports two tokens: GNO tokens and OWL tokens. GNO is an ERC-20 token, and there is a fixed amount of ten million minted. GNO tokens can be staked on the network for OWL tokens (stable coins) to pay fees.

For those unaware, DAO stands for decentralized autonomous organization. Think of DAOs as web-first businesses that are owned and operated by their members. Members vote on proposals to make decisions for the businesses, and similarly business actions require approval from the collective.

While the rise of DAOs has been overshadowed by movement in NFTs (non-fungible tokens) and various alt-coins in the last year, DAOs offer immense value potential for members and inventors alike. For example, some DAOs are launching automated market maker (AMM) exchanges that work with centralized exchanges to attract investors and drive liquidity. We’ve covered AMMs in the past, including their pros and cons, but DAOs add new wrinkles to our analysis. 

DAO-based AMM exchanges offer lower transaction costs, faster transaction speeds, and both staking and yield-farming opportunities for members. They also offer the unique selling proposition of NFTs as well such as community and digital asset rewards. And although AMM exchanges are an interesting new use case for DAOs, their potential far exceeds this area of DeFi.

  • More color on DAOs can be found here.

Now back to our journey. In addition to hosting their own prediction market, Gnosis has an open-source infrastructure and services arm which allows developers to build their own decentralized prediction market applications.

Coverage of GnosisDAO was limited, and although the price of GNO has jumped considerably since the time of the article, the theory behind GnosisDAO’s prediction market integration still has untapped potential.

The ‘theory’ we’re referring to actually has a name and you might have heard of it: Futarchy. The theory was developed by Robin Hanson, an economics professor at George Mason who is known for his work on prediction markets. The theory has also been explored in web3, notably by Vitalik Buterin the creator of Etherum.  

In short, Futarchy offers a theoretical governance model where elected officials manage ‘national welfare’ (the standard of living of the average citizen) while market participants speculate on which policies will raise national welfare. And according to Hanson, the primary tenet of this governance model is “When a betting market clearly estimates that a proposed policy would increase expected national welfare, that proposal becomes law.” This is the core of Hanson’s vote on values but bet on beliefs slogan. Citizens can vote on their values through a representative democracy structure, but bet on their beliefs through policy prediction markets.

Now, this theory is just that. Hanson points out several points of contention with his theory. These include practical hurdles around adoption, encoding policies with a welfare ultimatum, and limiting wealth market manipulators. Finally, one of the largest barriers to affirming Futarchy is testing it.

Here is Vitalik’s conception of Futarchy ⤵️

A (decentralized) match made in Heaven?

This is why GnosisDAO’s implementation of Futarchy is fascinating - it is a testing ground for a potentially revolutionary governance framework. It also creates a new paradigm; aligning decentralized information (PMs) and decentralized governance (DAOs).

Pairing DAOs and prediction markets solves problems for both technologies.

DAOs are the ultimate incentive for prediction markets. A prediction market in a DAO with enough members should not have issues generating enough liquidity to produce good intelligence because the DAO has pre-qualified market participants through sub-incentives inherently built into DAO. DAOs are a proxy for several incentives: monetary incentives, mission incentives, community incentives.

And for DAOs where decision-making and governance is critical, prediction markets offer a clean way for members to provide input on the potential outcome of proposals.

Since the inception of Crowd Money we have pondered the question: why haven’t prediction markets seen greater adoption? Both in the corporate world, and within government institutions, prediction markets have seemingly either remained undiscovered to date, or actively cast aside as a reliable input for making decisions. 

The answer to that question may be centralization. Prediction markets as a means of decision-making might directly compete with the already established decision-making framework that exists in companies and government agencies. These frameworks are heavily centralized, and are supported by rigid hierarchy which determines who makes decisions, and how they are made. Prediction markets would meaningfully disrupt that structure, and as such, the two might be incompatible. 

The Crowd Money twist

As we’ve considered the integration of prediction markets and DAOs, we have theorized ways to leverage already-existing markets to provide value and information to DAOs. The result is the notion of ‘mirrored markets’. 

In this framework, a DAO prediction market on a certain topic would have a corresponding market(s) on a traditional prediction market platform. Individuals who aren’t members of the DAO would trade on the mirrored markets, however exceptional performance on mirrored markets would warrant membership in the DAO.

The idea is that, for DAOs, granting membership to great forecasters will result in strong prediction market data and in turn strong decision-making. These markets would also present an alpha-generating opportunity for DAO members, and increase aggregate trading volume for the prediction market platform. And if markets charge episodicially or introduce subscription programs for DAOs to create markets on their platforms, this could create a new business line altogether.

The GNOxt frontier (last one we promise)

If you have followed our forecasting content at Global Guessing, then you will know that we have done a lot of forecasting in the public health space. We have also spoken with a number of accomplished forecasters on our podcast The Right Side of Maybe including Superforecaster Tom Liptay, Professor David Manheim, and researcher Juan Cambeiro. Through those conversations and independent exploration, it seems that the healthcare space is primed for the integration of prediction markets and DAOs. 

To this end, let’s look at VitaDAO. VitaDAO has leveraged the DAO framework to create a vehicle for decentralized intellectual property (IP) and patents. The organization is funding early-stage longevity research that may not get the funding or attention it needs from larger institutions. In exchange for providing funding to researchers globally without the restrictions of many common crowdfunding platforms, the DAO maintains the IP.

Individuals can stake currency (the process of holding funds in a cryptocurrency wallet to support the operations of a blockchain network or organization) to join VitaDAO, and offer a range of services such as providing capital, IP assets, labor, or data. Through a combination of these services, members support and finance the development of new therapeutics. This is IP and patent creation. Those assets are licensed to third parties to generate funds which are used by the DAO. Research data is also licensed for revenue. With these funds, VitaDAO intends to fund more research, and also buy valuable healthcare IP to hold and license, bringing more value back to the organization. In this model, value is distributed to the creators and researchers of drugs in a more equitable way, and not just the final manufacturer, publisher, or seller.

Now let’s imagine the integration of prediction markets within VitaDAO.

VitaDAO is organized into working groups, one of which if the Governance group which is tasked with overseeing and improving “all infrastructure and processes related to governance and decision-making among the VitaDAO community” as well as voting on, and sharing proposals. 

What if there were prediction markets within VitaDAO on the likelihood of a drug receiving FDA approval, or on a vaccine’s efficacy? Metaculus has a number of relevant markets currency operating on topics like Implantable Bioartificial Kidney FDA Approval and especially topical markets like Will there be a culturally significant development in aging research by 2030? They also have markets on even more granular trial-related questions like What will be the mean expected lifespan gain from one year of the TRIIM-X trial, as measured by the epigenetic clock GrimAge? And while these time horizons are far off, the idea is that topics like medicinal trials, government approvals, and drug efficacy are forecastable. 

If these markets were run within VitaDAO to inform funding decisions, the impact of VItaDao in the long-run could be greatly improved. Similarly, as these markets already exist on other platforms, these could act as our ‘mirrored market’ concept, and those who prove their forecasting ability in these markets might gain entry into VitaDAO.

We think VitaDAO would welcome talented public health and pharma-focused forecasters, and we are certain that Metaculus forecasters would like to leverage their on-platform achievements in an off-platform, money-generating entity. We are certainly missing something in our analysis here, as most do, but this idea is worth further scrutiny - if we could offer decentralized information to decentralized science it could result in a powershift within the worlds of synthetic biology, bioengineering, and Big Pharma. 

💊 PM Interview 

Disclaimer: We are not currently members of VitaDAO

Luckily, these questions and hypotheses don’t just have to live in the world of theory. In this issue of Crowd Money we were fortunate enough to have a member of VitaDAO provide an interview to discuss the DAO, and the potential of prediction markets integration both within VitaDAO and DAOs writ large. 

Vincent Weisser—Founding Steward & Tech and Product—has been with VitaDAO since its inception, and has been integral the the growth of the platform, the community, and the way that the DAO is governed. He is also a member of the Effective Altruism community, and has explored issues of machine learning, cryptocurrencies, and existential risk through artificial intelligence on his personal website. His interview gives insights into the future of decentralized science and prediction markets as true decision-making tools.

Interview with Vincent Weisser

Question: In recent months, VitaDAO has made a name for itself as one of the pioneers in decentralized science (DeSci). As we know, the DeSci collective funds early-stage research focused on age-related diseases. Your background is in Ethereum exchanges, however. Can you share how your interests led you to decentralized science, how you got involved with VitaDAO, and what opportunities the DAO framework has created for drug development?

Answer: Ethereum was the first thing that got me really excited about crypto, especially the concept behind The DAO, of a collective venture fund of sorts. In the following years I went down the rabbit hole of DeFi, and helped build dex.blue in 2017, an Ethereum based decentralized exchange. Simultaneously I also explored health care and longevity and self-taught myself molecular bio basics. I went to a bunch of conferences covering longevity (Undoing Aging), crypto (ETH DevCon), and effective altruism (EA Global), and was really interested in the intersection.

When I met the molecule.to team I got excited by their idea for VitaDAO and immediately joined full-time to make it happen. Given the interest in longevity by the crypto community, I think it makes a lot of sense to explore how we can collectively fund research into it through crypto. I was always personally interested in funding longevity research, but it was fairly impossible without a group bundling its resources and expertise. 

VitaDAO is organized by working groups, with one of those groups being the Governance working group. On the VitaDAO website, the Governance working group is described as being responsible for overseeing and improving governance infrastructure in the DAO, including decision-making. The group also votes and comments on proposals. What is that voting process like at VitaDAO? What information is considered, how are members of the Governance working group chosen, and how do you plan to measure success or failure of voting decisions?

Like in most DAO, governance is ultimately up to the token holders, who can initiate or give feedback on proposals and vote on them. Basically in the first step people within the DAO or even outside of it can create proposals on our governance forum, after which they will be discussed and through a poll going to the next stage of being put on-chain for people to vote for or against.

In the context of funding research, we have a scientific research evaluation group that makes individual public recommendations on a specific research funding proposal. So it could end up that two researchers make the case why we should fund a specific proposal and another researcher makes the case why we shouldn’t, although at the end it's up to every token holder to vote for his preferred outcome. 

Are you familiar with GnosisDAO or Futarchy? The idea of integrating prediction markets into DAOs to inform voting decisions by letting the DAO place bets on the outcomes of events relevant to ongoing proposals. Are there adequate mechanisms in place for DAOs to facilitate this?

Yes, some of us are based in the same coworking space as Gnosis, so we share a close connection. Also a big fan of Hanson’s ideas! I’d say there aren’t yet, but DAOs and governance is also still in its infancy regarding tooling and experiments. One example is quadratic voting, which we are now exploring to fund non-profit longevity research together with Gitcoin.

Another idea we are exploring is creating more on-chain identity around specific voters, as well as introduce delegation (for example that one can delegate votes to a researcher), as well as mechanism around signaling for specific IP-NFTs through mechanisms like Dutch auctions for fair-price discovery or staking on or against a research project. 

Do you think this could work for VitaDAO given the breadth of forecasts on medicinal trials and bioengineering that exist on other platforms i.e. Metaculus, Good Judgement Open? At Crowd Money we have been thinking about how DAOs can leverage intelligence from outside prediction market platforms. For markets that are relevant to your DAO, for example, would you value input from individuals who perform well. Could forecasting be a means of gaining membership to the DAO?

I’m a big fan of Metaculus and Good Judgement Open! Ultimately I think the main issue right now with prediction markets in crypto is that they are fairly illiquid and niche. Although I would consider financial markets such as crypto in a way as prediction markets in themselves, for betting on specific outcomes. Creating a fair price discovery mechanism around our research projects and making them liquid would in a way create a strong incentive to become a good predictor of the outcomes of that research. 

Last question! We have heard that one of the biggest pain points for DAOs is onboarding talent - funneling talent into right DAOs and finding the right talent to fit a DAO’s needs. Do you think that someone’s forecasting ability or prediction market performance could become a way to discern and onboard DAO talent at scale?

Definitely! At VitaDAO for example I consider our deal flow group one of the most important ones to make our model work long-term, and predicting which research could end up working would help to fund the most research in general.

One goal of mine is to refine and iterate on the best outcome-based incentives to incentivize the best people to contribute. Our second proposal has been on incentives related to the outcome of a research project, basically rewarding the people that made a successful project happen. It’s similar to the incentives of a VC scout or analyst, they make great rewards if they help fund the next breakthrough company. 

In a similar way if you imagine that we have a range of 100 projects and someone beats the average and ends up having a better success rate in picking successful research projects, that could become a strong indicator to get them more involved on deal flow. But even without pristine forecasting abilities, we welcome everyone! Feel free to check out our website, join our discord and introduce yourself! We constantly onboard new people to actively contribute, on everything from media, research, deal flow to tech and governance.

🔭 Introducing Supertrading

We want to conclude by sharing some initial thoughts on one of many concepts we’ve been toying with privately these past few months: A superforecaster aggregator for prediction markets.

In the forecasting world there are people who have been deemed Superforecasters - a term originating from, and trademarked by Good Judgement Inc 😬. The title refers to the idea that there exists a subset of forecasters who, across a wide number of questions spanning a variety of topics, are able to consistently out forecast the rest of the group (by having lower brier scores).

Looking at the paper Bias, Information, and Noise (BIN) paper by Satoppa et al., we know superforecasters derive their superior foresight by lowering their bias, effectively discerning signals from noise (recently seen with Jerome Powell’s renomination to the Fed, per Tetlock), and inputting more information, than the average forecaster. And if we return to the other paper by Satoppa et al., we also know that prediction markets are the most effective form of prediction aggregation at improving BIN.

So in theory, you could combine these two ideas to identify supertraders—individuals who have the best relative risk-adjusted returns, rather than lower brier scores given the differences between PMs and their prediction polls counterparts—after they have traded on a large number of questions spanning a diverse set of questions to:

  1. ⭐ Produce highly accurate predictions to use as an information source, which could potentially be sold as a way to offset or eliminate fees;

  2. 👍 Generate alpha as a trading bot by selling data to firms, possibly paving the way for fee-less platforms;

  3. 🤷 Create a new type of market maker that offers no-fee trading to people outside of the current superforecasting trader range which would allow for the entrance of dumb money while also allowing supers to move the markets when new information becomes available.

There are some difficulties with this approach. The first set of comes from the concept itself. As we said, brier scores aren’t the best measure here because traders don’t exclusively buy shares based on their true beliefs (i.e., I forecast a 80% likelihood so buy shares at 80c), but also based on relative price difference (i.e., I would buy 60c shares for that 80% forecast). Moreover, as Peter Wildeford mentioned on The Right Side of Maybe podcast, PM traders play a two level game that forecasters don’t: the second being their expectation on how the market itself will move. So it remains to be seen the data aggregated from supertraders will be more useful than that from traditional superforecasters.

The second set of difficulties come from putting the idea into practice. For example, we run into the risk that top traders might create new accounts if they notice someone shadowing their trades. However, since the aggregator would be shadowing their relative cohort rather than an individuals this might be less of an issue. Plus, the rate of supertraders out of the collective might be offset by the entrance of new supertraders in.

What do you think?

Leave a comment


We want to connect with you, our readers, on a deeper level! So we are going to start sharing articles or books that we have read, or new technology we purchased in the last month that we think are worth sharing.

Andrew

  • Bacteria, specifically E. coli, are being programmed and used as a distributed computing system. This is an exciting breakthrough in synthetic biology, and this is a field I will definitely be following

  • Quantifying experience and sensation will become increasingly relevant as we approach the metaverse (whatever that looks like). This article takes some large strides in getting there! 

  • Hyundai unveiled their second heritage Series restomod, a 1986 Hyundai Grandeur, in what might be the coolest-looking EV designs that I’ve seen this year. Elon Musk, watch out!

Clay

  • I finally finished Brandon Sanderson’s Rhythm of War, the fourth book (and my favorite so far) in his epic fantasy series The Stormlight Archives. The series is the only work of Sanderson I’ve read so far, but I can highly recommend the series to anyone who enjoys epic fantasy or enjoys deep character development.

  • I’ve now had my Samsung Galaxy Z Fold 3 5G for three months, and could not be happier. The device is everything I dreamed of when I saw the original Galaxy Fold in 2019. Having a phone-tablet duo in my pocket feels like the future, and has rendered my previous tablet usage to a shadow of its former self. I don’t miss my iPhone.


If you enjoyed the post, please consider liking the post down below ❤ and sharing the newsletter with a friend ➦. We’d also love to hear your thoughts in the comments 💬 and any suggestions or ideas for improvement too!