On Decentralizing Prediction Markets and Order Books

We propose techniques for decentralizing prediction markets and order books, utilizing Bitcoin’s security model and consensus mechanism. Decentralization of prediction markets offers several key advantages over a centralized market: no single entity governs over the market, all transactions are transparent in the block chain, and anybody can participate pseudonymously to either open a new market or place bets in an existing one. We provide trust agility: each market has its own specified arbiter and users can choose to interact in markets that rely on the arbiters they trust. We also provide a transparent, decentralized order book that enables order execution on the block chain in the presence of potentially malicious miners. 1 Introductory Remarks Bitcoin has demonstrated that achieving consensus in a decentralized network is practical. This has stimulated research on applying Bitcoin-esque consensus mechanisms to new applications (e.g., DNS through Namecoin, timestamping through CommitCoin [10], and smart contracts through Etherem). In this paper, we consider application of Bitcoin’s principles to prediction markets. A prediction market (PM) enables forecasts about uncertain future events to be forged into financial instruments that can be traded (bought, sold, shorted, etc.) until the uncertainty of the event is resolved. In several common forecasting scenarios, PMs have demonstrated lower error than polls, expert opinions, and statistical inference [2]. Thus an open and transparent PM not only serves its traders, it serves any stakeholder in the outcome by providing useful forecasting information through prices. Whenever discussing the application of Bitcoin to a new technology or service, its important to distinguish exactly what is meant. For example, a “Bitcoin-based prediction market” could mean at least three different things: (1) adding Bitcoin-related contracts (e.g., the future Bitcoin/USD exchange rate) to a traditional centralized PM, (2) converting the underlying currency of a centralized prediction market to Bitcoin, or (3) applying the design principles of Bitcoin to decentralize the functionality and governance of a PM. Of the three interpretations, approach (1) is not a research contribution. Approach (2) inherits most of the properties of a traditional PM: Opening markets for new future events is subject to a commitment by the PM host to determine the outcome, virtually any trading rules can be implemented, and trade settlement and clearing can be automated if money is held in trading accounts. In addition, by denominating the PM in Bitcoin, approach (2) enables easy electronic deposits and withdrawals from trading accounts, and can add a level of anonymity. An example of approach (2) is Predictious. This set of properties is a desirable starting point but we see several ways it can be improved through approach (3). Thus, our contribution is a novel PM design that enables: • A Decentralized Clearing/Settlement Service. Fully automated settlement and clearing of trades without escrowing funds to a trusted straight through processor (STP). • A Decentralized Order Matching Service. Fully automated matching of orders in a built-in call market, plus full support for external centralized exchanges. 4 http://namecoin.info 5 http://www.ethereum.org 6 https://www.predictious.com • Self-Organized Markets. Any participant can solicit forecasts on any event by arranging for any entity (or group of entities) to arbitrate the final payout based on the event’s outcome. • Agile Arbitration. Anyone can serve as an arbiter, and arbiters only need to sign two transactions (an agreement to serve and a declaration of an outcome) keeping the barrier to entry low. Traders can choose to participate in markets with arbiters they trust. Our analogue of Bitcoin miners can also arbitrate. • Transparency by Design. All trades, open markets, and arbitrated outcomes are reflected in a public ledger akin to Bitcoin’s block chain. • Flexible Fees. Fees paid to various parties can be configured on a per-market basis, with levels driven by market conditions (e.g., the minimum to incentivize correct behavior). • Resilience. Disruption to sets of participants will not disrupt the operations of the PM. • Theft Resistance. Like Bitcoin, currency and PM shares are held by the traders, and no transfers are possible without the holder’s digital signature. However like Bitcoin, users must protect their private keys and understand the risks of keeping money on an exchange service. • Pseudonymous Transactions. Like Bitcoin, holders of funds and shares are only identified with a pseudonymous public key, and any individual can hold an arbitrary number of keys. 2 Preliminaries and Related Work 2.1 Prediction Markets A PM enables participants to trade financial shares tied to the outcome of a specified future event. For example, if Alice, Bob, and Charlie are running for president, a share in ‘the outcome of Alice winning’ might entitle its holder to $1 if Alice wins and $0 if she does not. If the participants believed Alice to have a 60% chance of winning, the share would have an expected value of $0.60. In the opposite direction, if Bob and Charlie are trading at $0.30 and $0.10 respectively, the market on aggregate believes their likelihood of winning to be 30% and 10%. One of the most useful innovations of PMs is the intuitiveness of this pricing function [24]. Amateur traders and market observers can quickly assess current market belief, as well as monitor how forecasts change over time. The economic literature provides evidence that PMs can forecast certain types of events more accurately than methods that do not use financial incentives, such as polls (see [2] for an authoritative summary). They have been deployed internally by organizations such as the US Department of Defense, Microsoft, Google, IBM, and Intel, to forecast things like national security threats, natural disasters, and product development time and cost [2]. The literature on PMs tends to focus on topics orthogonal to how PMs are technically deployed, such as market scoring rules for market makers [13,9], accuracy of forecasts [23], and the relationship between share price and market belief [24]. Concurrently with the review of our paper, a decentralized PM called Truthcoin was independently proposed. It is also a Bitcoin-based design, however it focuses on determining a voting mechanism that incentivizes currency holders to judge the outcome of all events. We argue for designated arbitration in Section 5.1. Additionally, our approach does not use a market maker and is based on asset trading through a decentralized order book.