The Oracle Problem: How Decentralized Platforms Decide What Happened. Part 2
In the first part, we described the main problems of automatic oracles and concluded that they are too much non-universal to work in the prediction markets (PMs). Since PMs participants can make bets on various types of issues, to determine the winner, oracles should be able to find any kind of information that people can find as well.
Thus, it seems to be fair enough that the oracle should represent a community of reporter people organized in such a way that it would be beneficial for them to give conscientious answers — honest and competent.
There are various methods of motivating reporters for providing honest answers, and, therefore, in this article, we will look at some techniques that were implemented in Chainlink, Augur, Bitcoin Hivemind and Amoveo blockchain systems.
1. Chainlink Solution: a Set of Trusted Feeds and Reputation Monitoring
Even though an ordinary user learned about smart contracts in 2014 with the deployment of the Ethereum blockchain (ETH), these days they are still have not yet been widely introduced due to the difficulty of development. The main idea of the Chainlink blockchain (LINK) is to build these oracles on the Ethereum network. However, despite the considerable development time (comparable to the lifetime of ETH), the proposed solutions turned out to be quite straightforward and potentially very vulnerable. The main type of Chainlink oracles uses so-called “trusted feeds,” which are actually centralized.
The trusted feed can be organizations (for example, stock exchanges) or individual reporters. The specific source of information depends on the smart contract (on the type of data that is needed to close it). In fact, the trusted feed is chosen right before the conclusion of the contract.
If the contract was designed for a large payout for one of the parties, you can not exclude the probability of bribing the trusted feed to give false information. Since the nature of the blockchain transactions is irreversibility, the wrong closure of a contract may lead to theft of funds, even if this error was quickly detected. The probability of bribing is riskier if a trusted reporter is an individual or a little-known organization that can cover its tracks right after a fraud has been committed.
Besides the fact the trusted feed fully depends on the payment for services, this payment, to be fair, should be comparable to the amount for which the contract was concluded. This is perhaps the main problem of the trusted feed systems: oracles are either expensive or not very reliable.
To further motivate these feeds for honesty, Chainlink keeps monitors their reputation. Alongside the higher level of reputation, there is a higher probability this feed will be used in a smart contract. Since the reputation cannot be converted into money in an easy way, there is a chance of a so-called “retirement attack.” The main idea is that while the reporter is involved in the work of the blockchain, his reputation is important, but once the reporter decides to leave the system, then the only way to profit from this reputation — corruption. The higher the reporter’s reputation, the lesser he is expected to trick and the larger the contract can be trusted. But even a long-term and honest work experience do not guarantee at all that the reporter would not like to corrupt this contract.
2. Augur and Bitcoin Hivemind Solution: Many Reporters with a Financially Convertible Reputation
To avoid the retirement attack, in Augur and Bitcoin Hivemind prediction markets (PMs), good reporters get a reward in reputational tokens — REP and HIVE, respectively. They circulate freely on the cryptocurrency exchanges, and any honest reporter can earn on his reputation.
Meanwhile, the oracles of these systems are truly decentralized: many reporters can participate in their work at the closure of any contract, voting which bettor from the PMs has won. Reporters’ reputation plays a key role in the vote weight. Reporters can be rewarded with additional tokens for the correct answer or lose some funds for the wrong, respectively. Given such rules, for reporter, it is almost always unprofitable to lie.
The reputational tokens reward system reduces the bribe problem in comparison with the trusted feed but does not solve it completely. If the bettors’ stakes are high, and the price rate of REP or HIVE is low, then a significant part of their holders may be ready to lie for a larger reward.
Another problem of the above-mentioned systems is their reliance on other blockchains. Augur is just another layer of a network over the Ethereum blockchain, and Hivemind is over the Bitcoin network.
The disadvantage of this dependence we will overview right down below.
3. Amoveo Solution: Reporters Make Bets for Their Answer
The way how the Amoveo blockchain and its oracles work has similarities with Augur: reporters get a reward for the correct answer and suffer losses for the wrong one. However, there are certain fundamental differences as well.
Firstly, Amoveo has a full blockchain network with its own native currency — VEO, while Augur and Hivemind represent an additional layer over the Ethereum and Bitcoin networks. In various situations, they use either REP and HIVE tokens, or ETH and BTC currencies.
Secondly, as smart contract closes, Amoveo reporters do not vote for the chosen variant, but make bets in its favor in the VEO currency. These bets can vary, depending on how confident the reporter is. The correct answer is not the one that has the maximum votes, but the one that achieved the maximum total bets. All the collected bets are distributed in favor of the correct answer supporters.
In comparison with the recently-mentioned systems, this gives the following advantages.
- Reporter’s reward or loss does not rely on the token price rate. Reporters are free to choose the amount they want to risk;
- The non-limitation of bet size makes reporters responsible. The Amoveo participant may vary the bets depending on how competent he considers himself to be in a particular question.
- Thanks to this, the reporters’ winnings can greatly depend on the complexity of the question. If the question is elementary, then with a high probability everyone will answer unanimously and return the bids. But if the question is complicated, then a competent reporter can make a high bid and win a significant part of the opponents’ bets.
When the time comes to close a contract, the Amoveo oracle sends out to the participants a question to determine the winner. For example, “Is a Brent oil barrel worth more than $50 today?” Reporter volunteers choose the answer option: TRUE, FALSE or BAD_QUESTION. They support their answers with cash bets.
In case a barrel of oil is clearly higher than $50, most reporters will choose TRUE, if it is obviously cheaper — FALSE. But the situation may be ambiguous too: for instance, on the last working day of the year, the price of oil fluctuates around $50, and after the exchange’s closure the price on some turns out to be a little higher, and on the others — a little bit lower than the price that was mentioned in question. If the conditions of the bet did not specify from which stock exchange to extract the data, then a significant part of the reporters can answer BAD_QUESTION, since it is impossible to strictly determine the winner in such a formulation of the question.
4. Unhappy Bettors or Fork as the Last Argument
Despite all the measures listed above, bettors can sometimes disagree with the oracle’s decision. For example, if in the oil case some bettors tacitly meant the NYMEX (New York Mercantile Exchange) exchange as the well-known in the industry, they would not be satisfied with the BAD_QUESTION result with the return of bets to the participants.
Dissatisfied bettors can fork the network, dividing it into two branches. The new branch will consider that the oracle made the right decision, while the new one — that it was wrong. Sure, there are no guarantees that the new branch will definitely be the better one. Participants and especially miners will “vote with their feet”. If the overwhelming majority considers the oracle’s answer to be correct, it is unlikely to support the new branch.
In the case of truly massive disagreements, the blockchain can fully fork into two branches, as it happened in 2016 with the Ethereum blockchain. After the DAO Foundation incident, the network split into the Ethereum itself (ETH) and Ethereum Classic (ETC). In ETH, all transactions remained unchanged, while on the ETC network they “rolled back” into the past before the theft.
It is worth mentioning that not all the prediction market blockchain-based networks support the fork solution. Among the previously mentioned, only Amoveo (VEO) support this feature. A fork of Augur or Hivemind would require a fork of the parent network, which is difficult to organize, because most of its users are not related to these projects. As for Amoveo, it is a fully independent network.
Image by icons8.
