Summary:
This proposal is to update parameters that dictate the interest rate curve in the CRV-long, USDe-long, and pufETH-long LlamaLend markets. Rationale is to better protect lenders and improve performance of the markets.
Abstract:
LlamaRisk recently published a methodology for optimizing the Semilog monetary policy used in LlamaLend markets. Based on this system, we determine a target utilization rate for each market and assign optimal parameters for the min and max rates that determine the Semilog interest rate curve.
| Market | Current Params (min/max) | Avg. Utilization (6 month) | Proposed Params (min/max) | Target Utilization |
|---|---|---|---|---|
| CRV-long | 1%/80% | 59% | 0.03%/466.562% | 65% |
| USDe-long | 0.5%/40% | 55% | 0.00001%/88.072% | 85% |
| pufETH-long | 2%/45% | 52% | 0.00005%/92.874% | 84% |
In general, optimized values are more extreme than current values. This produces a more defined inflection point around the target utilization and substantially increases the borrow rate above target, strengthening the assurance that liquidity is always available.
Motivation:
There have been concerns raised recently about the current Monetary Policy configuration in many LlamaLend markets. In particular, the Secondary MonPolicy used in some markets was at full utilization recently, prompting a vote to revert these markets to Semilog. There is a general concern that the max rates, even in Semilog markets, are currently not set high enough to protect market liquidity in all scenarios.
LlamaRisk has developed a methodology for optimizing Semilog markets (and alternative IRMs) based on historical withdrawal demand in the market (to assign a target utilization) and accounting for trade-offs in mean-squared error, time above threshold, and rate/utilization volatility. This allows us to use quantitative models for assigning parameters that are suitable for the target market’s behavior. These optimizations will better protect lenders with a stronger assurance of available liquidity, allow us to monitor the maturity of the markets and calibrate accordingly, and parameterize these markets for maximal efficiency for the benefit of all market participants.
Semilog is a somewhat inflexible IRM, and research is underway by LlamaRisk and Swiss Stake to implement an improve IRM model that allows more finely tuned optimizations. This proposal is rather for additional lender protection in the short term until an improved model is ready to be implemented.
Specification:
Thorough details of the optimization methodology and results for the markets in this proposal are in the full report linked above. Provided here are the results for the CRV-long market as an example.
Target utilization is determined by identifying volatile periods defined as 2x the daily standard deviation and determining the 10th percentile of crvUSD withdrawals during those periods. We identify the 10th percentile as -35%, meaning that 90% of all observed withdrawals during volatile periods were less than or equal to this value. From this, we identify 65% as an appropriate target utilization.
We have run simulations across different parameter configurations to arrive at an optimal Semilog, and optimal Piecewise Linear (as a potential alternative IRM), and compare to the current (default) Semilog configuration. We then compare their performance in key criteria: Mean Squared Error (how closely does the market operate at target utilization), Time Above Threshold (how much time does the market spend above the target utilization), and rate/utilization volatility of the market. Notice that both the optimized Semilog and Piecewise Linear exhibit superior performance in all categories except rate volatility. In general, the tradeoff is that with improvements to market stabilization around the target utilization, rate volatility worsens. Some amount of increase volatility is acceptable in exchange for liquidity assurances that protect lenders and superior capital efficiency.
As a specific representation of the IRM curves, notice the optimized Semilog exhibits more pronounced curvature, substantially increasing the rate above target utilization. (note: the IRM curve here is capped at 80% to focus attention on the target utilization range.)
Notably, the existing Semilog Monetary Policy has hardcoded limits on
MIN_RATEof 0.1%. This requires a redeploy of the Semilog contracts. An example of the redeployed contract for the CRV-long market is here with min/max params assigned as 0.03%/466.562%. This contract can be introduced to the market by callingset_monetary_policy()to the CRV-long market Controller contract at0xEdA215b7666936DEd834f76f3fBC6F323295110A.
For:
Stronger liquidity guarantees that protect lenders. Optimize market operation around the target utilization for improved capital efficiency.
Against:
Introducing additional rate volatility in the market is undesirable, may increase potential for rate manipulation in the market.
