Gnosis is an open-source, Ethereum-based platform that enables users to aggregate information via prediction markets, a mechanism that incentivizes information sharing and rewards participants for sharing predictions on issues ranging from sports betting, political polling, financial forecasting, to risk assessment. A smart contract ecosystem is used to reward users who successfully predict the outcome of a given event, with buy-in costs disincentivizing guessing and deception. Gnosis also offers developer tools for Dapps that build on the platform by integrating information generated from a prediction market, and suggests several use cases. Gnosis operates on a three-token model. One token, OWL, is used as a coupon to pay fees, while the other, GNO, generates OWL when locked. A third token, Magnolia (MGN), accrues on use of Gnosis’ DutchX exchange and confers discounts to holders.
Gnosis believes prediction markets are a potentially game-changing new tool for effectively aggregating information. Compared to similar existing tools, such as polls and surveys, prediction markets differ in that they utilize market forces to incentivize participants to ‘share information’ via buying stakes in possible outcomes, with the aggregate outcome price reflecting participants’ various dispersed knowledge. By opening up decentralized prediction markets via the blockchain, Gnosis aspires to provide ‘the World’s Most Efficient Forecasting Tool’, a tool that it hopes will ultimately become ‘the Google of Customized Information Searching.’
Gnosis’ efforts are led by co-founders Martin Koppelmann and Stefan George. Co-founder and CEO Martin Koppelmann is the creator of the Bitcoin prediction market Fairlay. Co-founder and CTO Stefan George leads development and the architectural design of Gnosis. He has developed a number of games and apps. Chief Strategist Matt Liston was a Co-Founder and CEO of the competing Augur project, then known as Dyffy.
Additionally, Gnosis is advised by multiple blockchain innovators. Joe Lubin, an early supporter of the Gnosis project and the founder of ConsenSys, partnered with the Gnosis team to make it the first “spoke” in the ConsenSys ecosystem of Ethereum dapps. Two members of the ConsenSys executive team serve as members on the Gnosis Board of Directors and one more is serving as a general advisor. Founder of Ethereum, Vitalik Buterin is also serving as an advisor and has supported the team’s 2017 token sale, which launched successfully via a reverse dutch auction, a method of selling where sellers compete for buyers, typically through undercutting each other’s prices. This broke with the then convention of incentivizing early participation by offering bonuses that gradually decrease over time.
Gnosis enables users to build decentralized prediction markets. In a high-level overview, a prediction market works as follows:
- A question is posed, such as “Will the price of Bitcoin reach $2,000 USD by the end of the year”. Gnosis allows markets for questions with categorical or scalar outcomes, so a question such as “What price will this artifact from the Baghdad Museum sell for at Sotheby’s Spring Auction?” can also be posed.
- With Gnosis’ markets, users can stake for a position by purchasing a complete set of outcome tokens, which correspond to the range of outcomes (here ‘yes’ and ‘no’) and selling the undesired tokens. The conversion of collateral into outcome tokens comes with a fee.
- Provided outcome tokens have been generated by users and are on market, a user can buy a position in either the ‘yes’ or the ‘no’ side of the market at any time by directly purchasing that outcome token from the market. Gnosis has developed a wallet and dutch-auction modeled token exchange to facilitate trade.
- Trade in tokens representing outcomes influences each outcome shares’ price, which provides the market’s ‘forecast’. An outcome’s share price is thought to reflect users’ aggregate confidence and knowledge; shares of ‘yes’ may be priced higher than shares of ‘no’, suggesting the view that a ‘yes’ outcome is more probable.
- When the event in question occurs, an oracle, chosen on market creation, determines what the actual outcome was, thus resolving the market. Here, the oracle might report that BTC reached a price of $2,000 on January 1st. Gnosis offers a marketplace for oracles, and allows prediction market creators to choose among various oracle services.
- When the market resolves, a participant receives a payout if their prediction was correct. Winning shares, outcome tokens corresponding to the Oracle reported outcome, are redeemable for $1 each. Losing shares become worthless.
Besides providing users tools for launching prediction markets, the Gnosis platform will also offer supplemental services and support Dapps designed to integrate prediction market data. Gnosis’s platform will consist of three layers: Gnosis Core, Gnosis Services, and Gnosis Apps, all built on Ethereum.
Gnosis Core: The core layer provides components essential to creating and running Gnosis style prediction markets. Key components include event contracts, smart contracts used to create and redeem outcome tokens, and market contracts, smart contracts used to facilitate trade of outcome tokens on a prediction market. Key services, such as oracles and token exchanges, are also integrated with these components. More details on Gnosis’ current approach to prediction market design can be found below.
Gnosis Services: Gnosis allows third-party providers to offer oracle services, such as verifying the results of an election, the weather in a given location, or stock prices on a particular day. Such services are necessary for prediction markets to function and integral to prediction market integrity. Other proposed services include chatbots, stable coins, wallets, and token exchanges.
Gnosis Apps: Gnosis plans to develop its own in-house apps that integrate information generated from the platform’s prediction markets, and also offers developers tools for designers looking to build their own Dapps on top of the platform. The GnosisX challenge is a contest sponsored by Gnosis designed to encourage developers to build Gnosis based prediction market applications, offering cash prizes to winning designs.
Cryptoeconomics of Gnosis
Gnosis was initially developed around a novel two-token model, while a third token, Magnolia, (MGN) was introduced in early 2018 as part of Gnosis’ token exchange. In Gnosis’ primary token model, one token, GNO, is staked to produce OWL, another token that functions like a currency in the Gnosis ecosystem.
OWL functions similar to a payment token. Gnosis charges prediction market participants fees when participants convert collateral tokens into a set of outcome tokens. Such platform fees, such as a fee for 0.5% of the collateral used in creating event tokens, can be paid in OWL tokens, with one OWL being equivalent to $1 USD in fees. OWL can also be used to pay for up to half the transaction fee charged by Gnosis’ DutchX exchange, which supports outcome token trading. The network also accepts other forms of payment. OWL is designed for stability: OWL’s limited and non-exclusive platform uses should drive down the token’s value, while pegging OWL to the USD helps stabilize OWL’s price. Together, these design choices help cement OWL’s role as currency–– users have little reason to stockpile the token because OWL’s value is unlikely to rise.
GNO is quite different. The GNO token’s only purpose on the platform is to generate OWL tokens. Users generate OWL by locking a certain portion of their GNO tokens in a smart contract for a specified period of time. The longer the period and the greater the number of GNO tokens, the more OWL tokens the user earns. When the GNO tokens are locked, 30% of the OWL tokens are distributed to the user immediately. The remaining 70% are vested out over the duration of the smart contract’s life.
Algorithms applied to locked GNO regulate OWL’s total supply. The target for OWL’s total supply is 20 times the total monthly usage of OWL during the previous three months, denominated in OWL. The effects of such regulation is communicated to users––users are informed how much OWL will be generated by locking GNO prior to locking.
Gnosis’ token sale (ICO), held April 2017, proved extremely successful, raising $12.5 million USD in under 15 minutes. Of the total 10 million GNO supply, 4.2% was sold, with Gnosis retaining the remaining 95.8%. The sale took the form of a reverse dutch auction, with the seller (Gnosis) choosing an initial price ($30) set to decline over time. The intention behind decreasing the price of GNO was to incentivize participants to contribute when the price reached a level representing what they believed was its true value. Every user would pay the same price per GNO as the very last contributor. The sales quick duration is thought to have signaled participants beliefs that the $30 price for GNO was fair or undervalued. These tokens were subject to a one year vesting period and were held with the rationale of supporting further platform and application development, compensating Gnosis employees, and potentially conducting future token sales.
Around Jan 25th, 2018, Gnosis introduced a further token to its platform, Magnolia. MGN is designed to be used internally on Gnosis’ DutchX exchange as both an inflationary mechanism and reward for early, frequent exchange users. One MGN is generated for trading one ETH worth of any token and is automatically credited to users. Exchange fees are reduced in proportion to the percentage of total MGN a user holds. There will be no cap on MGN generated and MGN must be locked, not burnt, to be used. To maintain the same percent discount, users will either need to use the exchange frequently or buy unlocked MGN sold by other users–– MGN will be a tradeable ERC 20 token.
Gnosis is officially spun out of ConsenSys and is incorporated as an international company headquartered in Gibraltar. Gnosis chose Gibraltar for its ‘…farsighted and progressive approach to blockchain regulation.’ ConsenSys holds equity in Gnosis and vice versa, and Gnosis founders and early employees also hold equity in ConsenSys. Gnosis utilizes various ConsenSys services and projects, such as wallet and identity solution uPort, the accounting solution Balanc3 and the back-end infrastructure component Infura.
Prediction Market Challenges and Competition
Regulation: Ostensibly, prediction markets are about generating useful information. Yet sites operate in legal gray areas and several markets have shut down for reasons relating to online gambling, futures trading or money-laundering. The U.S. Commodities Futures Trading Commission (CFTC) shut down Intrade for unregulated futures trading. I-predict, a New Zealand based prediction market, closed when it failed to reach an exemption to anti-money laundering legislation, legislation that created obligations for businesses to check customers’ identities and report suspicious transactions. Iowa Electronic Market (IEM), a University of Iowa based political prediction market, received approval from the CFTC via letters extending no-action to the IEM in 1992 and has been operating since 1998.
Gnosis’ platform provides individuals the tools to create and run prediction markets. Whether such distinction sufficiently absolves Gnosis from existing or future regulation remains to be seen. The platform’s decentralization could make it harder for governments to regulate or shut down.
Understanding the Tool: Current industry focus appears directed at building components for creating prediction markets intended to be deployed in a very wide range of settings. While prediction markets, like polls or surveys, are a promising, perhaps even superior, tool for aggregating information and generating forecasts, providing someone a tool does not alone confer the know-how required to properly utilize it. Pollsters conducting polls or psychologists employing surveys arguably better know how to properly apply these tools for ends of inquiry, and better understand their tools’ limitations. Responsible survey, questionnaire, or poll design is conscious of how various factors influence results, such as how questions are worded, how people are selected to participate, and how the survey is constructed. Sampling bias, questionnaire design, polling error, and statistical significance are all important concepts informing research methodology.
Similarly educating prediction market creators on epistemic issues surrounding market design and implementation could prove instrumental to generating reliable data and legitimizing markets as a source of information. A poorly designed poll does not provide accurate information, yet this does not impugn polls as a tool; helping users distinguish between well and poorly implemented prediction markets can secure their reputation. Future research may also further clarify how current design choices affect the accuracy of forecasts generated using Gnosis’ tools. However, ensuring that knowledge, even if available, informs the efforts of individual users of the platform will likely remain a substantial challenge going forward.
Prediction Markets: Gnosis faces immediate competition from existing prediction markets and prediction market platforms. The most obvious source of competition for Gnosis is the decentralized prediction market Augur. In the short run, Augur and Gnosis are nearly direct competitors.
Augur and Gnosis also differ in several important ways.
- Augur uses their REP tokens to stake to their oracle as a method of ensuring reporting integrity. Gnosis uses a combination of on chain oracles, centralized off-chain oracles, and decentralized off-chain oracles to resolve prediction markets. This gives Dapp developers a lot of flexibility in choosing how to resolve prediction markets.
- Gnosis will use state channels to achieve higher levels of scalability than can be achieved by resolving 100% of transactions on chain.
- Gnosis’ focuses on being a platform for the creation of dapps, meaning that it may eventually offers products and services that Augur does not. On the other hand, app developers could wait to see whose prediction markets generate the more desirable or pertinent prediction data before committing to a platform.
Gnosis also faces competition from other established general prediction markets, such as BetMoose, or Predictious, as well as prediction markets tailored for specific subjects, such as the Hollywood Stock Exchange or the Iowa Electronic Market. These latter, specialized prediction markets might more easily attract the attention of experts with pertinent information, and don’t face the same legal issues affecting Gnosis; the Hollywood Stock Exchange is not played using real money, and the Iowa Electronic Market, as part of an ongoing research project, received the CFTC’s backing.
Options Markets: While Augur is a clear comparison point, in a public medium post entitled ‘The difference between Gnosis and Augur’, Gnosis claims it does not consider Augur a competitor, stating that it views betting exchanges like Betfair, bookmakers like William Hill, and options markets like Nadex as their real competition. This may strike readers as odd, as none of these organizations appears concerned with using markets for the explicit purpose of generating forecasts or allgomating expert opinion. Yet, given the CFTC’s shutdown of Intrade, prediction markets are clearly seen as possible competitors. Should Gnosis attempt to develop into an options markets, it will be interesting to see how they address regulatory concerns or established exchanges efforts to preserve their business models.
Progress and Development
The Gnosis project has been under development since 2015 and the underlying smart contracts, the general web interface, and the Gnosis.js library are already completed. According to the Gnosis whitepaper, several integrations with other projects such as Uport, MetaMask and RealityKeys have been tested. Gnosis has also developed a multisig wallet that is being used by several other projects such as Golem. Gnosis rebranded WIZ token as OWL tokens around January 2018, and the platform introduced a third token, Magnolia, around January 25th 2018 to facilitate its DutchX exchange.
Making Delphi: Challenges and Choices in Prediction Market Design
There a number of subtle complexities to prediction markets, representing both their challenges and opportunities. The constant stream of enthusiastic discussion around prediction markets’ potential within blockchain communities is indicative of the sense of opportunity surrounding them, yet it is not always clear that the reality has matched the larger vision. The following discussion should be of interest to readers interested in the subtler details and design tradeoffs involved in prediction market design.
Prediction markets, it should be stressed, are a relatively new tool for inquiry, not yet understood to same extent as polls and surveys. While blockchain technology offers unique opportunities for perfecting prediction markets, such as through incentivizing participation or decentralizing reporting, optimizing this tool will likely require experimentation and adjustment. This makes the space quite fascinating for those interested in prediction markets’ development, as design choices and comparative approaches to challenges should prove instructive.
Several components developed by Gnosis allow users to design custom prediction markets, like the one described in the above high-level overview. The components include event contracts, market contracts, oracles, and token exchange mechanisms. Gnosis envisions the components organized as so.
Released details on these components offer insight into how Gnosis envisions its markets functioning.
Event Contract. Event contracts are used to create and resolve prediction markets. Each event contract has three properties: an oracle, outcome tokens, and collateral tokens. To participate in a prediction market, a user uses collateral tokens––a currency such as ETH, USD, a stablecoin, etc––to buy a set of outcome tokens. Gnosis also plans to add liquidity to each market through market making bots: automated market participants who buy and sell outcome tokens for a particular event according to some algorithm.These transactions utilize the market contract component, described below.
Outcome tokens, as described by Gnosis, represent all possible wagerable outcomes for an event, which can either be ‘categorical’, like ‘who will win the World Cup’ or ‘scalar’, like ‘what will be the price of Gold in a month’. Users cannot buy a particular outcome token, but rather must buy the complete ‘collection’ of outcomes. Users keep the outcome tokens for the outcome they believe will occur, and sell the outcome tokens for the outcomes that won’t. When the event occurs and the market is resolved by the Oracle contract, outcome tokens corresponding to the actual outcome can be traded back for collateral tokens. Tokens for outcomes proving merely possible become worthless.
Market Contracts. Market contracts facilitate trade in outcome tokens on a prediction markets by providing that market some initial liquidity via a ‘Market Maker’. Gnosis notes that a prediction market doesn’t require a market contract, but claims adding one improves the market’s usability.
A market contract is like an automated market participant who buys and sells outcome tokens for a particular event according to some algorithm. A market creator funds the contract with a pool of collateral tokens, which are used by a Market Maker, a bot that employs an algorithm to price the prediction market’s various outcome tokens given the market’s current state. The bot buys and sells shares in outcomes from other participants at that price. Currently, Gnosis’ market maker is based on Logarithmic Market Scoring Rules (LMSR) developed by Dr. Robin Hanson.
A market creator is able to define an optional fee for such trading, described as a spread between the bid and ask price, and that fee compensates the market creator for providing the market its initial liquidity. According to Gnosis, the fees are also permitted because an automated Market Maker is expected to lose all its funding once the outcome is decided, since, by the time everyone knows about the market’s outcome, anyone can trade against the Market Maker by buying all the winning outcome tokens, leaving the Market Maker with all losing outcome tokens.
Token Exchange. Prediction market participants can buy and sell outcome tokens to one another, or to a market’s Market Maker. With Gnosis, these transactions occur via a decentralized exchange that uses a Dutch auction, rather than traditional order book, model. Gnosis’ offers several reasons for why a decentralized ‘DutchX’ format was chosen, including:
- Centralized exchanges have a higher risk of fund loss.
- The Dutch format is more suitable to the blockchain because it allows sellers to include market orders of any size without resulting in quick price movements, and because it allows traders to choose low gas prices for orders.
- The Dutch auction mechanism results in ‘the right’ or ‘fair’ token price given its applications of game theory.
- The exchange discourages strategic trading and encourages bidders to buy at what they perceive is the appropriate price.
Trading via Gnosis DutchX exchange costs fees, half of which can be paid in OWL, one of the platform’s tokens. Notably, Gnosis does not intend to partake of exchange generated fees, but rather will use them to generate discounts for frequent users in the form of Magnolia, a token that can be held to reduce fees or sold. One Magnolia is generated for trading one ETH worth of any token and is automatically credited to users.
Oracles: The crucial task of resolving a market is performed by an Oracle contract. Each event contract must reference a defined Oracle Contract. That Oracle Contract determines how that particular prediction market gets resolved. Variances in reporting and resolution methods can affect prediction market integrity, security, or status as ‘decentralized’, thus making decisions on Oracle design and integration particularly impactful.
Gnosis’ December 2017 Whitepaper does not contain the details concerning system architecture outlined in the May 2017 Whitepaper, which were described as under heavy development and subject to change, suggesting Gnosis’ final approach to oracles is still evolving. Readers should thus keep in mind that the veracity of released protocol details is in question.
That said, current information suggests Gnosis is oracle agnostic. While Gnosis may itself offer only limited centralized oracle solutions, Gnosis intends to provide market creators choices by opening up a marketplace for oracle solutions. Gnosis allows anyone to sign up to provide oracle services on its interface, and information regarding an Oracle’s reputation can be integrated via a connected Twitter account. Oracle providers can specify their terms, fees, and specializations, and market creators choose among listed oracles, even selecting multiple oracles for added security. Market creators can require oracles to submit a security deposit, which is forfeited should an oracle not agree with the majority of selected oracles.
As such, the actual process and method for determining an event’s ultimate outcome will vary depending on a market creator’s Oracle choice, and can depend greatly on the type of market: a market for local election predictions might use several oracles that report based off several news postings, while a market predicting stock prices could use an oracle contract determining an outcome based off of a single index’s listing, and a third market about natural disaster effects may rely on a combination off IoT sensor data and an on-or-off chain user reporting process.
Presuming Gnosis takes this approach, Gnosis’ claim to offer decentralized prediction markets is misleading when not understood a certain way: a prediction market on the Gnosis platform needn’t be resolved by an on-chain decentralized process, such as when independent and anonymous users collectively vote on-chain to verify what actually occurred. This is not to criticize Gnosis for failing to provide a decentralized oracle, as the decision to incorporate some aspects of centralization is arguably justified on pragmatic grounds––there are major challenges in designing a distributed oracle process that is cheap, timely, ensures honest reporting, and prevents collusion, and whether the benefits warrant the costs is debatable. Rather, this highlights the need for qualification in communicating what Gnosis, or any prediction market, could actually hope to accomplish; Gnosis allows users to create decentralized prediction markets, but markets created aren’t necessarily decentralized in all aspects.
Augur, a competing platform offering users its own version of decentralized prediction markets, has focused heavily on designing a token model that ensures quality distributed oracle reporting. For those interested in better understanding the aforementioned challenges in oracle design, Augur is a good case study; particularly Augur’s concerns with ‘parasitic’ markets, balancing of reporter compensation and platform fees, integration of efficient designated reporters, and design of a forking mechanism.
Augur’s explicit design choices also help clarify several key questions that Gnosis must eventually address:
- What will be Gnosis’ approach to ill-formed questions? It is not clear whether Gnosis currently has a mechanism for incentivizing market creators to design well-formed prediction markets. Markets can be ill-formed in several ways: questions can be ambiguous and subjective, ‘will the release of Crunchy’Os rock investors’ socks’; be answered via a counterfactual lacking a clear oracle method, ‘who would have been president had Comey not reopened his investigation’; or be creator-manipulable to an excessive extent, ‘will I skip class next Tuesday.’ Market creators, presumably responsible for defining the markets possible outcomes, could also omit relevant possibilities: the question ‘Who will win the league match’ could lack a draw outcome.Augur’s approach to this issue is to require both that ‘invalid’ be one market outcome and that market creators fund a bond that is forfeited if the oracle determines that the question was ill-formed and returns an outcome of invalid. So, if a market did ask ‘will the release of Crunchy’Os rock investors’ socks’, participants could stake tokens on the invalid outcome. Whether this will satisfactorily address all the above concerns is unclear, but the choice represents a design conscious of the issue.
- What leeway will market creators have in selecting the currency used for collateral tokens? Prediction markets built on volatile cryptocurrencies seem likely to produce skewed data, since participants decisions are influenced not only by their relevant knowledge relating to the question at hand, but also on their speculative beliefs about the future price of their token holdings. For these reasons, Augur plans on integrating stablecoins, foreseeing their development to be integral to prediction market’s success.Presuming Gnosis’ efforts to drive down and stabilize OWL’s value succeed, OWL could serve as a built-in stablecoin. Gnosis also could leave choice of currency to market creators, but uninformed choice of currency may make a given market a less-than-optimal forecasting tool, potentially leading to frustrating user experiences for those interested in generating accurate forecasts.
- Who can challenge oracles how, and to what extent? While some markets’ results are easily verifiable, oracles still make mistakes. For example, Intrade, a general purpose centralized prediction market, relied on three independent media sources to resolve its market on the 2012 Iowa caucus in favor of Mitt Romney, only for vote certification process to later reveal Rick Santorum the winner.Gnosis ability to prevent or manage such a situation depends on several design choices, each with tradeoffs. A longer reporting period can help ensure oracle accuracy, although this also delays participant payout. In earlier, potentially outdated materials, Gnosis describes a possible solution: ‘With the Ultimate Oracle, a market is created on “What was the outcome of this event?” The original market is then resolved by the outcome of this secondary market. Participants can place bets on either side of the market. After the betting period, a 24 hour timer starts. If at the conclusion of the 24 hours, the outcome remains the same then the market is closed. Participants on the winning side of this market are rewarded and the other side suffers a loss.’