[For Discussion] A plan for interest model adjustment

Background:

When we first launched Alpaca in March of 2021, the market was in a strong bull run.

It was a different environment with high speculative demand and higher yields. BNB Chain yields on DEXes (without leverage) were in the range of 50% - 100% APR for most pools. However, these yields dropped over time as more capital entered the market and farm tokens like CAKE depreciated in price.

Today, all the major farms have yields below 15% APR (as of 29th Mar 2022)

PCS BNB-USDT: 15.01%

PCS BNB-BUSD: 13.94%

MDEX BNB-BUSD: 10.31%

MDEX ETH-USDT: 13.31%

What does this mean for our interest model? The 20% rate in slope2 is too high for the current yield environment and there will be very few borrowers at 20%. It’s just not profitable to do leveraged yield farming at that borrowing rate.

As a result, lending pools’ utilization rarely rises above ~40% (borrowing interest = 15% @ 40% utilization) This was evidenced in the major pools where utilization was hovering between 30% - 40% prior to Automated Vault’s launch.This leaves a lot of unborrowed capital on the sidelines, which also means lower APY for lenders.

We believe a more efficient “equilibrium” could be achieved by lowering the borrowing interest at slope2, which will create higher utilization while still maintaining the same, or even higher revenue, to the lenders.

Example:

State Supply (USD) Borrowed (USD) Borrowing Interest Rate Annual Interest paid (USD) Lenders APR
State1 100 30 10% 3 3%
State2 100 50 9% 4.5 4.5%

As seen in the simple illustrative example above, with higher utilization, borrowers enjoy lower borrowing interest (10% → 9%) while lenders also earn higher yields (3% → 4.5%)

The question then is: how do we ensure there will be an increase in borrowing demand to increase utilization, compensating for the lower interest rate?

Challenges:

While the core team has discussed this issue internally in the past, the question above is one of the reasons we have not lowered Slope2 borrowing interest until now. There was no guarantee that with a shift in the interest model, there would be a corresponding increase in the borrowing demand.

With no new borrowing demand, lenders and our protocol would end up with less revenue, which could lead to lending deposit withdrawal, potentially creating a vicious cycle where the protocol lost significant TVL.

In fact, we tried with this back in Q3 of 2021 where we moved the Slope2 starting point from 50% utilization to 60% utilization to try to incentivize borrowers to open more positions, but it had no perceptible impact, and we didn’t pursue it further.

Rationale:

With the introduction of Automated Vaults, the situation has changed. We have seen a strong demand from the community for Automated Vaults, with the TVL cap for each 8x pool reached within a few hours of launch.

So with Automated Vaults, we now have high confidence that we could move our system to a higher utilization state where all the participants in the ecosystem will benefit.

Adjusting the interest model would achieve five major benefits:

  • Higher APY for lenders
  • Lower borrowing interest rate for borrowers
  • Higher protocol revenue
  • Additional capacity for our Automated Vaults
  • Higher lending utilization which provides more platform stability

Proposed Implementation:

We propose the following target interest models for each asset:

Slope2 Kink: utilization at which Slope2 begins

Slope2 Interest: the borrowing interest rate for Slope2 (the flat part)

Asset Slope2 kink Slope2 interest Lend APYs at Slope2
BNB 85% 12.5% ~ 10% - 12%
BUSD 85% 12.5% ~ 10% - 12%
USDT 85% 12.5% ~ 10% - 12%
BTC 85% 7.5% ~ 6% - 7%
ETH 85% 7.5% ~ 6% - 7%
USDC 85% 10.0% ~ 8% - 10%

As seen above, if we can raise the utilization to Slope2 portion of the model, lenders would also benefit from higher lending APY%

We would not make this implementation in one go, but rather adjust the slope gradually for each token each time we add a new instance of an Automated Vault. This way, we can ensure that lenders and platform revenue will not be adversely affected by a change. The core team will have the discretion to adjust interest models within the boundary above. For example, slope2 kink for BNB will never be beyond 85% and the slope2 interest rate will not be set below 12.5%.

The diagram below illustrates how the interest slope will change. The specific parameters will be based on various parameters at the time - e.g., TVL, APR, etc.

14 Likes

Agreed that this is a great enhancement that benefits everyone especially with neutral strategies attracting new capital into the protocols.

2 Likes

Great rationale! Happy farming herd!

Great job!
What’s our target utilization rate? Between 60% and 80%?

Since this has allowed the flexibility to set the i/r model to the maximum utilization kink and minimum i/r, it should be allowed. This would boost the usage of delta neutral even further, and provide more supply to the already growing demands.

Also as a clarification, I assume slope 2 (the flat part, ie: 85% to 90%) would be the optimal utilization? As anything above 90% would follow slope 3, which is a high gradient and assumed would cause the delta neutral vaults to be unprofitable.

85-90% over a number of iterations

1 Like

“Also as a clarification, I assume slope 2 (the flat part, ie: 85% to 90%) would be the optimal utilization?”

Correct

1 Like

I like this change and will gladly vote in favor but I wonder what this means for the BNB based AVs performance on the long term.

Historically, a good portion of lent BNBs exit the platform whenever there’s a launchpad on Binance.com. Assuming the BNB pool is at the optimal 85% utilization before the launchpad, it would be at near 100% utilization almost inevitably during the launchpad.

Do you think this is a problem at all?

It depends why the BNB leaves. If it leaves for higher APYs, then if we can get Alpaca’s BNB lending APYs up to be competitive, it wouldn’t have a reason to leave.

If the BNB leaves to get exposure to launchpad tokens, then it’s out of our control. But by increasing utilization, we can at least decrease the amount of BNB that can leave on a whim in mass. Also, even when BNB has left in the past, utilization has typically stayed around 90%, because above 90%, new BNB lenders are attracted to plug the gap.

2 Likes

When is the new interest model coming on production?

As mentioned in the post, we will gradually make adjustment as we add capacity to new Automated Vaults.

We are planning to put up a BNB-BUSD vault this Thursday.

2 Likes

This discussion is now a proposal. Any additional feedbacks / comments can be provided here: [AIP-5] Interest Model Adjustment