Semi-automated funds allocation

Summary

Transition from a weekly funds allocation vote to a programmatic calculation of the optimal allocation given a set of pre-defined rules. Retain human oversight and execution of the reallocation.

Background

Since late 2021, a weekly vote has been conducted to determine how OUSD’s underlying collateral (DAI, USDC, USDT) should be allocated across the various yield-generating strategies. Four voters participated in the first proposal. Over a year later, eight voters participated in the most recent proposal. Voting each week is time-consuming, and voters are not aligned on a framework for determining the appropriate allocation.

What’s not working

OUSD holders want to earn the highest possible risk-adjusted yield, but their funds are managed in a way that often doesn’t maximize risk-adjusted yield. Voters may skew the allocation to lower-yielding, potentially higher-risk strategies simply because they happen to show up and vote during a week when participation is low.

Voters lack the information necessary to make optimal decisions. Moving funds in and out of different strategies involves non-trivial gas costs and slippage in the case of Curve/Convex. Allocations also often involve swapping the underlying collateral. It’s not currently possible for voters to account for these costs to know the real impact of their decisions.

Our current process is also inefficient and slow. Even if all voters have sufficient information and are aligned on the desired outcomes, it’s a waste of time for each of them to go through the process of calculating the best allocation. If they all agreed in advance on a set of rules, the calculation could be run once and executed according to their collective preferences. With the current 48-hour voting period and a 7-day wait in between each reallocation, the protocol is unable to react to market opportunities that would bring higher yield.

Perhaps most importantly, prospective users of OUSD are concerned that there is no predictable way that their funds will be managed and they don’t know who will be making the decisions from one week to the next. While the consequences are relatively low with allocations being limited to only a handful of heavily-audited, whitelisted strategies that are constantly being monitored for risk, it’s reasonable for them to wonder if all of the collateral will be converted to the least desirable stablecoin and deposited into a single protocol, for example. They want to know that there is a well-informed, objective procedure for reallocations.

Proposal

Let’s work our way toward automated reallocations that can be triggered more frequently without compromising security. In the past, we’ve contemplated an automated rebalancing feature at the smart contract level that would seek to maximize yield whenever a public function is called. Exposing this functionality could be dangerous and would lead to increased code complexity as our strategy count and ruleset continue to grow. But there’s nothing stopping us from creating an off-chain script that can be run to calculate the optimal allocation for our strategists to execute. This would allow for more frequent, effective reallocations while maintaining human oversight.

Rather than vote weekly on how funds should be allocated in the moment, voters would prescribe rules in advance that would be incorporated into the script. For example, one rule might be that no more than 50% of the funds should be held in any one stablecoin (except during extraordinary circumstances like this past weekend when holding 100% USDT would have been preferred). This would become one of the constraints used to mitigate risk.

We’ve already started to scratch the surface with informal constraints, such as the limit on our Convex LUSD allocation. Voters agreed on a maximum allocation of $100k for that strategy and we added a simple formula to the strategists’ tooling to handle any excess votes. In the future, every strategy could have a maximum percentage or dollar allocation that voters would set and update from time to time.

Example logic

‣ Get the yield rates for all strategies ‣ Rank DAI, USDC, and USDT according to the rates available from the strategies ‣ Calculate an optimal Convex OUSD allocation of X ‣ Fund the Convex OUSD strategy up to X using the stablecoin with the lowest opportunity cost ‣ Iterate through each strategy from highest-yielding to lowest ‣ Deposit up to Y of the relevant stablecoin into each pool not exceeding a limit of Z

Other potential considerations:

‣ Minimize costs ‣ Maximize MORPHO rewards ‣ Allocate a minimum amount to each protocol ‣ Maintain a certain level of diversification