Introduction

Sommelier’s Real Yield USD Cellar, operated by 7Seas Capital, is the final evolution of organic stablecoin yields in DeFi.

By “organic yield” we mean yield that results from trading activity (fees) rather than resulting from rewards or incentives. The primary sources of organic yield exist on lending platforms like Aave and Compound, and decentralized exchanges like Uniswap. These protocols comprise both the largest source of locked liquidity (TVL) as well as user activity.

Given the diversity of organic yield opportunities in DeFi, there exists the unsolved problem of how to allocate stablecoin capital to maximize yield (we call it the stablecoin allocation problem). A baseline approach is to monitor the DefiLlama Yield boards and allocate accordingly, however this approach is tedious, inaccurate, and fails to account for effects such as market impact.

The Real Yield USD Cellar is an effort to solve the stablecoin allocation problem.

How does the strategy work?

Solving the stablecoin allocation problem is non-trivial and requires some component of off-chain computation. Sommelier’s novel infrastructure enables active management of an ERC-4626 vault (guided by off-chain computation) while remaining non-custodial, transparent, and decentralized.

The Real Yield USD Cellar is the first product to launch on the new V2 Cellars architecture which was recently audited by Macro.

The V2 architecture deserves its own post, but the TLDR is that it’s a vault architecture that can interact with any arbitrary DeFi protocol, provided that an “adaptor” exists for that protocol. In addition to the main Cellar (vault) contract, the contracts for adaptors to Aave, Compound, and Uniswap V3 were audited as well. This means that at present Sommelier can support any strategy that makes use of Aave, Compound, or Uniswap V3, such as long/short trading strategies or Uniswap V3 liquidity provisioning optimizers.

The Real Yield USD Cellar makes use of these three adaptors (Aave, Compound, Uniswap V3) to tackle the stablecoin allocation problem. In particular, the Cellar simultaneously allocates capital to Aave and Compound lending pools and Uniswap V3 LP pools in order to maximize yield.

The optimal allocation is determined by a numerical optimization procedure that accounts for swap fees and market impact due to position size, and makes use of various simple time-series forecasting methods to estimate (future) base yields.

One important reason that the Real Yield USD Strategy is able to achieve superior yields is that it optimizes Uniswap V3 tick ranges. Picking a lending position on Aave or Compound is relatively easy (ignoring factors like market impact which are actually important) because there are no degrees of freedom - it simply boils down to the decision of whether to lend a certain token or not. Providing liquidity on Uniswap V3, on the other hand, is complex because the choice of tick range determines both fee revenue and impermanent loss. Our optimization procedure accounts for all of these factors.

What are the risks?

Risk is mitigated in the Real Yield USD Cellar at several levels.

First, the Cellar smart contract can only interact with reputable, battle-tested protocols that are approved by Sommelier Governance - Aave, Compound, and Uniswap V3. This is actually more broadly a feature of the V2 smart contract architecture - a trust-minimized approach to scaling DeFi integrations.

Second, the Cellar limits the choice of stablecoins to DAI, USDC, and USDT. Limiting the choice of stablecoins is a core design consideration - allowing riskier stablecoins would boost yield .

Third, there is no leverage at all in the system. The Cellar only has the functionality to lend on Aave and Compound and LP on Uniswap V3 (as well as swap between stablecoins). As is the case with existing ERC-4626 vault products, Sommelier Cellars are fully non-custodial and transparent, meaning users can deposit/withdraw assets freely at any time and are able to see all rebalance activity on-chain.