Biblio

Many blockchain applications use decentralized oracles to trustlessly retrieve external information as those platforms are agnostic to real-world information. Some existing decentralized oracle protocols make use of majority-voting schemes to determine the outcomes and/or rewards to participants. In these cases, the awards (or penalties) grow linearly to the participant stakes, therefore voters are indifferent between voting through a single or multiple identities. Furthermore, the voters receive a reward only when they agree with the majority outcome, a tactic that may lead to herd behavior. This paper proposes an oracle protocol based on peer prediction mechanisms with non-linear staking rules. In the proposed approach, instead of being rewarded when agreeing with a majority outcome, a voter receives awards when their report achieves a relatively high score based on a peer prediction scoring scheme. The scoring scheme is designed to be incentive compatible so that the maximized expected score is achieved only with honest reporting. A non-linear stake scaling rule is proposed to discourage Sybil attacks. This paper also provides a theoretical analysis and guidelines for implementation as reference.