Stablecoins are designed to address the volatility of crypto assets by maintaining a peg to a non-volatile currency such as the US Dollar. Decentralized Stablecoins that maintain their collateral on-chain need a pricing oracle to determine the current market value of the collateral. They also employ a decentralized governance system to make policy changes. In this paper we analyze the inner-workings of the pricing oracle and the decentralized governance mechanism employed in the MakerDAO stablecoin, one of the largest and fully developed on-chain stablecoins. We study the accuracy of the pricing oracle over time, as well as disagreements between pricing reports received by MakerDAO. We also study the robustness of the decentralized governance system. This work sheds detailed light on the practical operation of a pricing oracle and a decentralized governance mechanism in a large deployed system. We make a number of recommendations for improvements based on our findings.
“The paper empirically studies the performance of of the MakerDAO Pricing Oracles and Governance Mechanisms and proposes improvements.”
“The paper analyzes the MakerDAO oracles and governance using empirical, on-chain, data. On oracles, after a presentation of the existing and upcoming technologies used, the two successive oracles mechanisms for MakerDAO are analyzed. The data fed to these oracles was recovered and compared to external sources of price for ETH-USD. These comparisons lead to recommendations to strengthen the oracle mechanism.
Different variants of the existing algorithm used to condense price feeds into a single price are tested but none seem to provide any advantage compared to the existing one. A statistical test for price feed outlier detection is proposed that would have eliminated past erroneous prices fed to the oracle.
On governance, the voting mechanisms are outlined and voter turnout is measured. Various attacks are presented and their costs explained. The evolution of the supply distribution is measured using external sources and shows that such attacks are becoming more difficult to execute by the top holders. Finally, recommendations are made to simplify the voting process and improve its monitoring.”
“The work studies the pricing oracle and the governance of MakerDAO. The main goal of the system is to keep Dai pegged at $1. For that, MakerDAO uses ETH as collateral and 14 data sources to aggregate pricing data for ETH-USD. The paper does measurement studies to analyze the disagreement between the data sources. Moreover, the paper studies the governance of MakerDAO.”
“The paper examines the stability of the pricing Oracle and governance mechanisms used for the MakerDAO stable coin, and in particular looks at performance between the new V2 mechanisms in comparison with the issues seen in the previous one.”
“The paper is well written, and presents a very interesting history of the MakerDAO stablecoin and some of its competitors. The analysis of price deviation is interesting in and of itself since it provides clear empirical evidence of issues that can be encountered by some of these mechanics, as is the exploration of the implications of the governance mechanisms and their practicality.”
“The paper is based on a solid empirical foundation and the mechanisms studied are well described. It gives a good overview over the improvements achieved for MakerDAO and proposes a number of ideas for further progress.”
“The analysis of price deviations is insightful. Overall, the paper provides good explanation of the current system and pertinent analysis.”
“The impact of stale data on the Oracle’s pricing might have been a cause as well as a result of market volatility”
“The paper makes only a few conclusions from their measurements. The measurements are interesting in itself, but it does not go much beyond it. The fact MakerDAO V2 fails when
bar = 14 looks more like a bug in the contract rather than a design decision. If it is a design decision, what is the reason behind it?”
“Some of the analysis could have been improved upon: price deviations are more usually measured in basis points rather than in absolute amounts (especially important when the price is very volatile). [Additionally, the] discussion of the staleness of feeders doesn't consider an explanation by possible congestion of the network: is feeder 8 using a smaller fee than everyone else leading to its quotes being more often stale?”
“The winsored mean probably doesn't truncate many values given the low number of feeders and its 10% ([0,5%[ and ]95%,100%]) exclusion of measures.”
“There are [many] concerns about the
bar feature, potentially making the contract vulnerable to DOS attacks. It looks like we only need two price feeders to go offline to shutdown the price oracle.”