Practical Applications of Decentralized Prediction Markets

How cryptocurrency-powered markets could disrupt betting, finance, forecasting, scientific research, and governance

Introduction

In early 2016, I speculated that decentralized prediction markets could give us the power to easily “Google” the probability of any imaginable event occurring.¹ Now that the first market has been running for seven months, it’s time to re-evaluate the potential of what was at the time a theoretical and unproven technology.

Prediction markets allow users to buy and sell shares in the outcome of specific events. After the event occurs, users holding shares of accurate event outcomes are rewarded while users holding the inaccurate ones lose their money. Share prices fluctuate between 0 and 1 based on supply and demand from traders until the event’s outcome is determined, and can act as reliable estimates of the event’s likelihood of occurring. If I purchase “yes” shares of “Will Space X send humans to Mars by 2024” for 5 cents each, the market estimates a 5% likelihood of this happening and I will be able to sell my shares for $1 a piece if it does.

Decentralized markets use cryptocurrencies and smart contracts to differentiate themselves from their centralized counterparts in a few key ways. Instead of users having to register with centralized services that restrict access and manage market creation, no registration is required and anyone on the planet can create or participate in any type of market. Instead of relying on markets run by closed-source code hosted on private servers, markets are automated by open-source code running on public blockchains. Instead of having to trust a centralized third party to hold their money, users maintain control of their funds.

Allowing billions of people to easily speculate on the likelihood of an infinite number of events occurring can democratize access to markets and advanced forecasting techniques. If decentralized prediction markets manage to achieve mass adoption, society will have the power to tap into and leverage the “Wisdom of the Crowd” at unprecedented scale.²

Decentralized Prediction Markets

Augur

Augur’s User Interface

Augur is the world’s first fully featured decentralized prediction market to launch and is built on Ethereum’s blockchain. Augur’s reputation token (REP) holders serve as the system’s decentralized oracle by reporting on the outcome of events. They are a critical part of the platform because blockchains rely on oracles to feed them information about events taking place in the real world. Reporters are rewarded with a percentage of trading fees for accurately reporting, and punished with a loss of REP for lying. A detailed description of the project’s goals and technical architecture can be found in the Augur whitepaper.³ Dr. Robin Hanson (“father of modern prediction markets”), Vitalik Buterin(Ethereum founder), Ron Bernstein (former Intrade CEO), and Elizabeth Stark (Lightning Labs Founder, Coin Center Fellow) have all served as advisors to the project.

Augur was one of the first projects to run a crowdsale on Ethereum, pre-selling the REP token to early adopters in order to fund development of the project. Augur’s novel use of REP to achieve full decentralization and perform a critical function within the network helps it stand out in a sea of projects whose attempts to create valuable utility tokens appear to be scams, money grabs, or misguided attempts to shoehorn a token into a product that doesn’t need one. Applied cryptography consultant and Ethereum skeptic Peter Todd commented that “Augur is one of those rare cases [where a utility token makes technical sense]”.⁴ Cryptoasset analysts Chris Burninske and Jack Tatar wrote, “Augur is one of the clearest uses of cryptotokens, and its potential success could set the stage for even more implementations of crypotokens in the future.”⁵

Gnosis

An example market from the Gnosis Olympia prediction market tournament

Gnosis is an Ethereum-based prediction market protocol that is still under development. Gnosis’ team is focused on creating open-source tools to help developers easily build specialized prediction market applications for different use cases instead of building a single application for all types of markets.⁶ The team is also making the effort to acquire the licenses required for running a regulated exchange in the United States.

Gnosis operates using a dual-token model; the GNO token can be locked up to generate OWL tokens which are each worth one dollar worth of fees on the Gnosis trading platform. Gnosis does not include its own oracle and will be compatible with a variety of external centralized and decentralized oracles. The Gnosis Dutch Exchange is a decentralized trading protocol that will be used to trade event outcome shares.⁷ A detailed description of the project’s goals and technical architecture can be found in the Gnosis Whitepaper.⁸

Gnosis CEO Martin Koepellman created and sold Fairlay, the largest Bitcoin-based prediction market. Vitalik Buterin, Dr. Robin Hanson, and Ethereum co-founder Joseph Lubin are advisors for the project. The team is also working with Wedbush Securities on a stock fundamentals forecasting application.⁹

Applications

Betting

Betting on popular events like sports or elections is the most obvious use case for prediction markets. Some of Augur’s most popular markets apart from those on cryptocurrency price speculation have been the 2018 U.S. Midterm elections, the 2018 World Cup, the 2019 Oscar Awards, and the 2019 Super Bowl. Decentralized markets can compete with traditional betting services by offering greater access to market creation and participation, lower fees, and user custody over funds.

Financial Markets

Decentralized prediction markets allow users to gain exposure to a stock, commodity, or other type of asset without owning the underlying. This could make a much broader selection of financial instruments available to a significantly larger portion of the population than is currently possible. Anyone can participate in markets without anyone’s permission or relying on a centralized intermediary to hold/transfer their funds.

A farmer in the developing world with a mobile phone, internet access, a healthy business and some savings, but no access to banking, the U.S. stock market, or advanced financial services could “add” Google, Tesla, or Apple to their portfolio from a web or mobile app by betting on the direction of the stock’s price movement.

A college student in the U.S. could gain exposure to their favorite stocks in a few clicks, without having to register for a brokerage service. Instead of needing to purchase an entire share of Alphabet for over a $1000 which they might not be able to afford, they could purchase a small fraction of a “share”, a feature that isn’t available on popular investing apps like Robinhood.

An investor who wants to profit from the fluctuations in housing prices of a certain city could do so without having to invest in property directly or through a fund, instead they would bet in a market on average house prices in the area. Individuals and organizations can create decentralized insurance applications by allowing users to hedge against the possibility of a disaster occurring by purchasing shares saying that it will happen.¹⁰ Nearly any category of financial derivative products could theoretically be built using a decentralized prediction market.

Forecasting

A report published by the CIA in 2007 states that prediction markets can be used to enhance U.S intelligence capabilities.¹¹ The report describes how Google, HP Labs and Caltech, pharmaceutical giant Eli Lily, and The Iowa Election Markets (IEM) have used prediction markets to forecast product launches, office openings, product sales, drug trial outcomes and U.S. election results.¹² These markets often outperformed official company forecasts and traditional polling. For example, IEM predictions of U.S. Presidential election outcomes from 1988–2004 were more accurate than national polls 74% of the time.

Caltech researchers found that a prediction market based Information Aggregation Mechanism (IAM) was more accurate in predicting Intel’s future sales than the company’s official forecasts 75% of the time.¹³ Internal prediction markets at Ford Motor Company and Google also outperformed expert forecasts.¹⁴

Most research on prediction markets until now has come from organizations with deep pockets and significant technical resources: national governments, universities, and large corporations. Decentralized prediction markets will allow startups and other small and medium sized organizations to more easily take advantage of a powerful tool with proven utility.

Scientific Research

Prediction markets have been used to effectively estimate the reproducibility of psychological studies, evaluate the replicability of economic research, and to forecast the outcome of professional research evaluations.¹⁵

They can also be used to fund scientific researchers “by rewarding an instantaneous, honest and unbiased disclosure of research findings.” Allocating funds based on actual contributions to research problems instead of based on past performance could help overcomes issues like publication bias and delayed information disclosure.¹⁶ Opening up the creation of and participation in prediction markets to the entire world will significantly increase the ability of researchers and their funders to leverage their capabilities.

Governance

A group of economists that includes four Nobel Laureates and Google’s Chief Economist has argued that prediction markets can improve social welfare by helping businesses and governments make more accurate forecasts and better corporate governance and national policy decisions. They caution against over-regulation that would limit their potential benefits to society and advocate for “safe harbor” status for certain types of institutions that wish to run markets.¹⁷

One radical proposed application is Futarchy, a governance system where society’s goals are chosen through democratic voting while prediction markets are used to determine which policies will be pursued in order to achieve those goals.¹⁸ We’re just now starting to see the first serious experiments with futarchy systems built on decentralized prediction markets. Aragon has awarded a $120,000 grant to fund the research and development of a futarchy application for their Decentralized Autonomous Organization (DAO) management platform.¹⁹

Ralph Merkle, co-inventor of public key cryptography and creator of the Merkle Tree has written a paper on DAOs, Democracy and Governance.²⁰ Merkle describes a system of governance for nation states where prediction markets would be used to determine which policies should be pursued to improve a “Democratic Collective Welfare” score that is calculated using a poll of individual citizens’ self-reported well-being. Public platforms like Augur and Gnosis that are open for developers to build on top of will make it easier for us to put these types of ambitious concepts to the test.

Some might be concerned that futarchy would limit participation to those with disposable income. Governments or organizations that determine futarchy is a legitimate decision making tool could easily subsidize participation costs, paying people to participate in prediction markets in order to reap the benefits of the crowds wisdom.

Barriers to Adoption and Potential Issues

User Experience and On boarding

These platforms must be accessible to non-cryptocurrency enthusiasts in order to achieve their full potential. Applications and exchanges like Coinbase and Cash App have made the process of purchasing, storing, and using cryptocurrencies easier, but there is still a lot of product development and consumer education that needs to be done. Augur users have to download and synchronize a desktop application, Augur Node, in order to use the platform. Mainstream adoption will require web and mobile applications that are as simple as the ones most people use on a daily basis.

A new in-browser version of Augur Node, “Augur State” is being developed. Veil and Guesser are centralized platforms that provide user-friendly interfaces for creating and interacting with Augur-based markets, and Predictions.Global allows anyone to easily browse markets. Helena is an application built on Gnosis that aims to provide an easy way for investors to stay up to date on blockchain events and trends related to business, technology, and legal topics.²¹

Market Liquidity

Low liquidity across Augur’s markets has contributed to underwhelming levels of trading activity since its launch. At the time of writing there is only about $2 million at stake across all markets on the platform. This makes it difficult to use for anyone interested in betting large sums of money. It also severely limits Augur’s potential utility as markets will require significant amounts of people betting in them in order to average out people’s biases and provide reliable estimates.

This issue can be addressed over time by making the platform easier to use, on-boarding dedicated market makers, and concerted marketing efforts once the platform can easily handle larger amounts of users.

Blockchain Scalability

The lack of mature blockchain scaling solutions will act as a barrier to adoption in the short to mid-term. If large numbers of users tried using Augur today it could start to clog Ethereum’s blockchain, leading to slow transaction times. Off-chain order books, state-channels, and sidechains, can be used to increase performance in the near-term until long-term solutions like sharding are available.

Fees

Augur co-founder Joey Krug has detailed the current fee situation and how it might evolve over time. Fees are incurred when: purchasing Ether (~1.5% — 4%), creating markets(~1–2%), paying Ethereum gas fees (~3%), and dealing with Ether’s daily volatility (~5%). Joey expects that market creator fees will trend towards 0, gas fees will drop as Ethereum scales, and volatility issues will be addressed by supporting stablecoins like Dai in Augur V2.²²

Oracle Reliability

Augur’s decentralized oracle mechanism has functioned effectively so far, but it could prove to be unreliable in the long-term. Centralized oracles represent a single point of failure and potential vulnerability. With the recent resolution of a poorly worded market on the U.S. midterm elections, we have some early evidence that Augur’s oracle is flexible and has the ability to favor the Spirit of the Market over the Letter of the Market.²³

Market Manipulation

If these markets gain widespread adoption there will be significant financial and social incentives to manipulate them. A number of attack vectors are outlined in Augur’s whitepaper.

Illegal and Unethical Markets

People might choose to create morally dubious markets. One example that has caused some concern is assassination markets.²⁴ The first and best line of defense against this is relying on Augur’s reporters. If a majority of REP tokens are in the hands of people who don’t approve of the creation of and participation in such markets, then reporters can declare them invalid.

If REP holders fail to shutdown clearly unethical markets, law enforcement has a wide array of tools at their disposal to discourage proliferation of assassination markets. Google, Chainalysis, Neutriono, and the SEC are developing tools for conducting in-depth analysis of blockchain-based activity.²⁵ The DHS is even looking to acquire tools for monitoring anonymous cryptocurrencies.²⁶ In light of the availability of these tools and network analysis showing that usage of cryptocurrencies for illicit purposes has declined over time, concerns surrounding these types of markets may be overblown.²⁷

Regulation

Heavy-handed regulation could discourage users from participating in markets, especially if governments decide to implement outright bans. Veil is currently unavailable in the United States, because of “legal and regulatory considerations.”²⁸ A recent Supreme Court ruling that cleared the way for states to allow commercial sports betting indicates that there may be an opportunity to create a more friendly regulatory environment through legislation or the judicial system.²⁹

Foolish Crowds

Researchers have discovered several ways in which the Wisdom of Crowds can be undermined. One study found that highly confident individuals and critical masses of lay people sharing similar views both serve as “attractors of opinions” that can have significant social influence on crowds and drive the group away from the truth.³⁰ Another found that knowledge about estimates made by others in the group can undermine the wisdom of the crowd through “social influence”, “range reduction”, and “confidence” effects.³¹While these dynamics can impact the usefulness of prediction markets, the features of decentralized prediction markets could give them some resistance to these issues.

Recent work has shown that the dynamics of group accuracy change with the group’s network structure. In centralized networks where individuals can dominate the collective estimation process, group estimates become less accurate. In contrast, with decentralized communication networks, group estimates become reliably more accurate as a result of information exchange.³² It seems reasonable to expect that the utility of decentralized prediction markets as general forecasting tools will be dependent on their size, diversity, and the extent to which participants communicate with each other and are aware of how their peers are betting.

There are a variety of potential systematic biases (ie: difficulty understanding large numbers) that might be found within a crowd leading it to be wrong. Knowing what the crowd thinks is ultimately just one of many available tools that can be used to make more accurate predictions.

Conclusion

We are in the early days of experimenting with decentralized prediction markets. We’re seeing glimpses of their potential utility in forecasting and financial markets, but haven’t yet seen any serious efforts to use them for scientific research or governance. The current version of Augur is a technically impressive but inefficient betting tool, incapable of supporting the hundreds of millions, or billions of users that would make it a truly world-changing platform.

Still, the fact that the first market has been operating for seven months without any significant hitches is remarkable. Further research is warranted to explore the extent to which decentralized prediction markets can offer more benefits to larger segments of society than their centralized predecessors. We may be witnessing the birth of a revolution in financial technology and the emergence of a class of applications that can play a significant role in facilitating mass adoption of cryptocurrencies and open blockchains.

References

[1] Shingai Thornton, “Practical Applications of Cryptocurrency: Decentralized Prediction Markets”, Medium (blog), January 12, 2016, https://medium.com/@shingaithornton/practical-applications-of-cryptocurrency-decentralized-prediction-markets-95a15be69a76

[2] Sheng Kung, Michael Yi, Mark Steyvers, Michael D. Lee, Matthew J. Dry, “The Wisdom of the Crowd in Combinatorial Problems”, Cognitive Science 36, issue 3 (January, 2012): https://onlinelibrary.wiley.com/doi/full/10.1111/j.1551-6709.2011.01223.x

[3] Jack Peterson, Joseph Krug, Micah Zoltu, Austin K. Williams, and Stephanie Alexander, “Augur: a Decentralized Oracle and Prediction Market Platform”, Forecast Foundation, July 12, 2018, https://www.augur.net/whitepaper.pdf

[4] Peter Todd, Twitter, October 3 2017 https://twitter.com/peterktodd/status/915254130985156609

[5] Chris Burninske and Jack Tatar, Cryptoassets: The Innovative Investor’s Guide to Bitcoin and Beyond (New York: McGraw Hill, 2017), 64

[6] “Gnosis Apollo”, Read the Docs, https://gnosis-apollo.readthedocs.io/en/latest/index.html

[7] “DutchX”, Read the Docs, https://dutchx.readthedocs.io/en/latest/

[8] Gnosis, “Gnosis Whitepaper”, Gnosis Ltd., December 22, 2017, https://gnosis.pm/assets/pdf/gnosis-whitepaper.pdf

[9] https://www.pepperlaw.com/resource/30462/14C2

[10] Etherisc, “Creating Prediction Markets for Insurance”, Etherisc.com (blog), Dec 8, 2017, https://blog.etherisc.com/creating-prediction-markets-for-insurance-5759484621a

[11] Puong Fei Yeh, “ Using Prediction Markets to Enhance US Intelligence Capabilities”: A “Standard & Poors 500 Index” for Intelligence, Central Intelligence Agency, last updated: Jun 26, 2008, https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/csi-studies/studies/vol50no4/using-prediction-markets-to-enhance-us-intelligence-capabilities.html

[12] Bo Cowgill, “Putting crowd wisdom to work”, Official Google Blog, September 21, 2005, https://googleblog.blogspot.com/2005/09/putting-crowd-wisdom-to-work.html;

Charles R. Plott and Kay-Yut Chen, “Information Aggregation Mechanisms: Concept, Design and Implementation for a Sales Forecasting Problem”, California Institute of Technology (unpublished), 2002, https://authors.library.caltech.edu/44358/;

Joyce E.Berg, Forrest D.Nelson Thomas A.Rietz “Prediction market accuracy in the long run” International Journal of Forecasting 24, issue 2, (April, 2008):285–300, https://www.sciencedirect.com/science/article/pii/S0169207008000320

http://www.forecastingprinciples.com/files/Berg_Nelson_Rietz_2007.pdf

[13] Benjamin J. Gillen, Charles R. Plott, Matthuew Shum, “Information Aggregation Mechanisms in the Field: Sales Forecasting Inside Intel”, MIT Economics, (October, 2012), http://economics.mit.edu/files/8299

[14] Bo Cowgill, Eric Zitzewitz, “Corporate Prediction Markets: Evidence from Google, Ford, and Firm X”, The Review of Economic Studies, Volume 82, Issue 4, (October, 2015), Pages 1309–1341, https://academic.oup.com/restud/article-abstract/82/4/1309/2607345?redirectedFrom=fulltext

http://www.restud.com/wp-content/uploads/2015/03/MS14671manuscript.pdf

[15] Anna Dreber, Thomas Pfeiffer, Johan Almenberg, Siri Isaksson, Brad Wilson, Yiling Chen, Brian A. Nosek, and Magnus Johannesson, “Using prediction markets to estimate the reproducibility of scientific research”, PNAS, (December, 2015), https://www.pnas.org/content/112/50/15343.abstract

Colin F. Camerer, Anna Dreber, Eskil Forsell, Teck-Hua Ho, “Evaluating replicability of laboratory experiments in economics”, Science 351_,_ issue 6280, (March, 2016): 1433–1436_,_ http://science.sciencemag.org/content/351/6280/1433/tab-pdf

Marcus R. Munafo, Thomas Pfeiffer, Adam Altmejd, “Using prediction markets to forecast research evaluations”, Royal Society, (October, 2015), https://royalsocietypublishing.org/doi/10.1098/rsos.150287

[16] Johan Almenberg, Ken Kittlitz, Thomas Pfeiffer “An Experiment on Prediction Markets in Science”, PLOS ONE, (December, 2009), https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0008500

[17] Kenneth J. Arrow, Thomas Schelling, Robert J. Schiller, Vernon L. Smith, Cass R. Sunstein, Hal R. Varian, “The Promise of Prediction Markets”, Science 320, issue 5878, (May, 2008): 877–878, https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.320.1811&rep=rep1&type=pdf

[18] Robin Hanson, “ Shall We Vote on Values, But Bet on Beliefs?”, Journal of Political Philosophy,(October, 2007), https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.71.8309

[19] Aragon, “Aragon Nest: Third round of grants”, Aragon blog, October 25, 2018, https://blog.aragon.org/aragon-nest-third-round-of-grants/

[20]Ralph Merkle, “DAOs, Democracy and Governance”, Merkle.com, May 2016, http://merkle.com/papers/DAOdemocracyDraft.pdf

[21] Veil, https://veil.co/;

Guesser, https://www.guesser.io/;

Helena, https://helena.network

[22] Joey Krug, “Fees, Fees, and… Fees?”, Medium (blog), August 8, 2018, https://medium.com/@joeykrug/fees-fees-and-fees-8939c2b5ecae

[23]Benjamin Roberts, Twitter, February 1, 2019 https://twitter.com/benjmnr/status/1091355105822162944

[24] David Floyd, “The First Augur Assassination Markets Have Arrived”, Coindesk, July 25, 2018, https://www.coindesk.com/the-first-augur-assassination-markets-have-arrived

[25] Michael del Castillo, “Navigating Bitcoin, Ethereum, XRP: How Google Is Quietly Making Blockchains Searchable”, Forbes, February 4, 2019, https://www.forbes.com/sites/michaeldelcastillo/2019/02/04/navigating-bitcoin-ethereum-xrp-how-google-is-quietly-making-blockchains-searchable/;

Chainalysis, https://www.chainalysis.com/;

Neutrino, https://www.neutrino.nu;

Securities and Exchange Commission Office of Acquisitions, “Blockchain data”, Federal Business Opportunities, last modified January 31, 2019, https://www.fbo.gov/index?s=opportunity&mode=form&id=c18a03f93cf06df47dab8a1c1a7f87a9&tab=core&_cview=0

[26]Department of Homeland Security Office of the Chief Procurement Officer “The Department of Homeland Security Small Business Innovation Research”, Federal Business Opportunities, last modified Nov 30, 2018 https://www.fbo.gov/index?s=opportunity&mode=form&id=7ed0c5ef2df7e26ffb5c1dee3ceaa171&tab=core&_cview=0

[27] Adam Hayes, Shaowen Lio, Paolo Tasca, “The Evolution of the Bitcoin Economy: Extracting and Analyzing the Network of Payment Relationships”, SSRN, (July , 2016), https://ssrn.com/abstract=2808762

[28] Paul Fletcher-Hill, “Introducing Veil”, Veil Blog, January 8, 2019, https://medium.com/veil-blog/introducing-veil-649036f9d492

[29] Adam Liptak and Kevin Draper, “Supreme Court Ruling Favors Sports Betting”, New York Times, May 14, 2018, https://www.nytimes.com/2018/05/14/us/politics/supreme-court-sports-betting-new-jersey.html

[30]Mehdi Moussaïd, Juliane E. Kämmer, Pantelis P. Analytics, Hansjörg Neth, “Social Influence and the Collective Dynamics of Opinion Formation”, PLOS ONE, (November, 2013), https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0078433

[31] Jan Lorenz, Heiko Rauhut, Frank Schweitzer, and Dirk Helbing, “ How social influence can undermine the wisdom of crowd effect”, PNAS, (May, 2011), https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3107299/

[32] Joshua Becker, Devon Brackbill, and Damon Centola, “Network dynamics of social influence in the wisdom of crowds”, PNAS, (October, 2017), https://www.pnas.org/content/114/26/E5070