System Health & Liquidation Mechanisms in the Nereus Protocol
The liquidation mechanism implemented by a protocol like Nereus ensures that ‘bad debt’ positions are promptly unwound and liquidated, preventing the protocol from having to assume the burden of potentially undercollateralized positions. A ‘bad debt’ position constitutes a position where the loan to collateral value ratio falls below a predetermined limit called the ‘liquidation threshold’. Although, a ‘bad debt’ position is not always natively undercollateralised, meaning the protocol has an additional layer of safety even after a position falls below its token specific ‘liquidation threshold’.
Liquidators are compensated for unwinding ‘bad debt’ positions within a protocol. Protocols must ensure that this compensation adequately reflects the market risk of the collateral asset within the bad debt position which is being liquidated. Providing liquidating agents with an economic benefit can ensure that ‘bad debt’ positions within the protocol are liquidated before they become undercollateralised. This improves the protocol’s health and helps it withstand the large liquidation volume customary to the cryptocurrency markets. This is our intention for Nereus.
Liquidations
Protocol participants interact with a DeFi protocol primarily for the purposes of lending and borrowing. Collateral must be posted to borrow from a protocol. A borrowing position must be overcollateralised, meaning the participant must deposit a greater value of assets than they borrow. The initial ratio of the borrowed tokens value over the deposited tokens value must be below the maximum required Loan to Value (LTV). Given the maximum LTV at 70% for WBTC, a 1000 USDT deposit would let the depositor borrow up to 700 USDT worth of WBTC. The 30% remainder provides the protocol with a layer of safety in case the posted collateral value decreases and/or the borrowed tokens value increases. Ideally, the maximum LTV and the liquidation threshold should decrease with the volatility of both tokens involved and should increase with the correlation between both tokens.
For instance, a 1 WBTC depositor could borrow 0.7 WBTCs with a very low risk of liquidation as the correlation between the deposited and borrowed assets, both WBTCs, is naturally at 100%. As the values of the deposited and borrowed tokens move together, the actual LTV at 70% would hardly fluctuate, with the only change being the LTV slowly deteriorating over time as the variable debt interest accrues. This accrued interest is added to the borrowed 0.7 WBTCs to compute the actual LTV. Hence, over time the accruing interest can push the position towards the WBTC liquidation threshold of 75%.
In Nereus V1 and all standard protocols today, the maximum LTV and liquidation threshold are not optimised for correlation risk. Instead, they are conservative choices that are calibrated in a discretionary manner, by the protocol’s decentralised autonomous organisation (DAO). If the position holder deems it not conservative enough according to her risk profile, she can post more collateral to make sure the actual LTV remains near her targeted level.
At the very least, a lending position must conform to the token’s specific liquidation threshold. Consequently, the position holder must ensure that her position remains above the Liquidation threshold of the specific asset they have deposited. In the situation where the value of the collateral asset posted drops below its pre-specified Liquidation threshold, the position is now up for liquidation, and the position holder would lose a minimum of the ‘liquidation fee’ earned by the liquidator.
In practice, any agent can utilise the LiquidationCall() function that lets them acquire and close a portion of the undercollateralised position. A maximum of 50% can be liquidated by a single agent. The collateral of the ‘bad debt’ position is essentially bought at a discount, which constitutes the liquidation fee. The discount varies based upon the volatility and risk profile of the collateral token in liquidation. The initial position holder loses her right to redeem her position when it falls below its specific liquidation threshold.
In the extreme scenario where liquidators are not incentivised enough to acquire the ‘bad debt’, the positions can become a burden to the protocol. In this situation, the protocol itself must assume the risk of the bad debt position, with its treasury first, and then with the help of its community. The protocol’s overall risk is exhibited in the ‘health factor’ of the protocol. A protocol that must assume bad debt positions which have not been liquidated will have a low health factor and can be susceptible to large losses during extreme market volatility.
Nereus will follow the market norm by providing liquidators with the economic incentive of a discount upon the collateral assets they purchase whilst unwinding & liquidating an undercollateralised position.
How Do Protocols maintain accurate Asset Pricing?
The dilemma of how a DeFi protocol such as Nereus maintains accurate asset pricing is a very relevant discussion within modern DeFi, as blockchains must find a way to communicate with events taking place outside their network, and outside the decentralised world. Oracles can be bridges between the centralised and decentralised worlds, or among various blockchains, protocols and Dapp’s. They set the market price and volume standards needed to accurately estimate the value of a position.
There are two established types of Oracles — Centralised and Decentralised Oracles.
Centralised Oracles can provide asset pricing data to a protocol such as Nereus. Although, utilising a centralised data source would represent a single point of failure for the protocol’s ecosystem. A centralised Oracle represents a significant security risk to a protocol, as its centralised nature means it is susceptible to manipulation via a single attack vector. A centralised Oracle would seek to mitigate this risk by diversifying its sources (e.g., exchanges, OTC desks). The original Chainlink oracle is structured with several nodes, each representing a centralised data source. The ultimate price fed to decentralised DApps on the other end is in essence a median of these prices that is trimmed for outliers. Because of the centralised nature of the nodes, and despite the decentralised structure of the Chainlink contracts, this version of Chainlink can still be considered as centralised.
True decentralised Oracles could provide protocols such as Nereus (subsequent versions) with an alternative to centralised Oracles, as unlike centralised Oracles they are not reliant on centralised data providers . Decentralised Oracles rely on one or multiple DEXes (e.g., Uniswap), increasing the independence of the pricing data they provide.
Utilising a Decentralised Oracle within a protocol keeps within the roots of DeFi, providing a transparent and reliable source of pricing data. An example of an Oracle is Chainlink’s Decentralised Oracle Networks, which provides aggregated pricing data from a diverse range of premium decentralised data sources. Additionally, Chainlink hosts this service on a transparent network that is secured by node operators running security-audited software that has been rigorously validated to operate at scale without downtime or corruption.
Nereus V1 will implement the Chainlink pricing oracle. The Chainlink oracle is a best-in-class oracle and will help Nereus to maintain a strong protocol health factor in the situation of market volatility. Accurate and swift pricing of tokens will ensure that positions instantaneously become available for liquidation when the underlying collateral asset value fluctuates.
Recursive Borrowing in DeFi
As the market norm in DeFi Protocols is overcollateralised borrowing, Traders employed the traditional method of recursive borrowing to leverage their initial position. Recursive borrowing consists of opening a collateralised position, then posting the borrowed amount as collateral to open an additional position. This can be repeated multiple times to achieve significant leverage. Employing a recursive borrowing strategy increases the user’s credit risk significantly. As the borrowed amounts are also at risk, the user would be locking her liquidity. If one position is liquidated, she might not be able to increase her collateral, or redeem her debt, and could eventually lose everything.
Recursive borrowing can also cause complications for Protocols such as Nereus. The credit risk born by the protocol is higher, and correlated liquidation events would hurt the protocol’s health factor. Furthermore, the protocol distributes its native tokens to incentivise participation in its early days. These distributed rewards could be farmed by one user with an extreme risk profile. As mentioned in prior articles, Mercenary Capital participants utilise Recursive borrowing to leverage their positions on asset markets which distribute incentives to borrowers and lenders. This allows the mercenary capital participants to farm protocol native token incentive rewards. Behavioural analysis will help us estimate the risk of recursive borrowing for the protocol.
Nereus V1 has implemented conservative risk parameters to discourage excessive borrowing on any asset market. The recursive borrowing strategy will not be economically viable when the utilisation rates on specific asset markets reach suboptimal levels.
