AVv3 Launch Announcement

Huacayachief
Alpaca Finance
Published in
7 min readJul 27, 2023

--

Dear alpacas,

The long-awaited moment has arrived. After much R&D, we’re finally announcing the launch date for Automated Vaults V3! (AVv3)

This article will go through the launch plan and all the relevant details for the upcoming release of this newest product line in our ecosystem.

🚀Launch Details

Timeline:

AVv3 will go live on 3 August 2023 at 11 AM UTC

Vault Details:

The first instance of AVv3 will be a BNB Savings Vault (L-USDTBNB05-PCS1), which you can invest in here when it goes live.

Underlying Farming Pool: PancakeSwap BNB-USDT (0.05% swap fee tier)

Planned Vault’s Capacity: $250k at start, increasing by $500k / week.

Note: the actual capacity increase could be accelerated or delayed depending on any potential issues discovered after production deployment and/or uncontrollable market situations (e.g., low availability of assets to borrow due to Binance Launchpad.)

Once we are comfortable with the operations, we will start scaling the capacity and add additional vaults (e.g., market neutral). We expect ~2–3 weeks of live production time before adding an additional vault.

Invest & Withdraw: As a user, you can expect the same simple experience when investing in the new AVv3. You will be able to freely invest and withdraw from the vault at any time without lockup. The vault will accept base tokens for investment; i.e., BNB for BNB Savings Vault and Stablecoins for Market-Neutral Vaults. However, when withdrawing from the vault, you will have the option to choose whether to convert all assets into base tokens or to minimize trading. For the minimize trading option, you will receive assets proportionally to the current asset composition in the vault.

💡AVv3 Strategy

The AVv3 strategy is based on the one of AVv2. The Vault Manager leverages many data points including historical prices and various economic factors to develop signals which can determine, with a high probability, which market regime we are currently in (trending or range bound; will there be a reversion, etc). Based on this information, the hedging algorithm would adjust its behavior (i.e., “operating mode”) based on what’s optimal at the time. For more detailed information, please refer to our Docs.

The Hedging operation in AVv3 works similar to repurchasing in AVv2. The LP position is not affected / changed during hedging. The hedging transaction only involves swapping of debt; i.e., borrowing more of one asset and repaying another to achieve the desired exposure. However, the main difference is that with hedging in AVv3, the swap is done directly on a DEX. This change will create higher transparency, while not sacrificing the benefits from swapping on CEX, because CL DEXs already offer low enough fees and price impacts that are competitive with CEXs.

With AVv3, the additional logic and output that the vault manager needs to provide will be the liquidity price range. At launch, the price range will be set at roughly +/- 10%. This range is based on the historical asset volatility. The range is also set to be slightly narrower than the average LPs’ settings — higher capital efficiency and higher yields for our positions. Over time, as we gain more confidence in the strategy, we will look to further concentrate the range to increase yield generation.

📈Backtest

In this section, we will discuss the backtest setup and results

Data and setup

Backtest Period: August 2022 — July 2023

Data Sources: To run our backtest, we collected the data of every swap transaction on the PCS BNB-USDT pool from the blockchain to build up historical states of the pool over time from scratch. Please note that the backtest was run on the PCSv2 actual data for the period of August 2022 — April 2023, because the PCSv3 pool data did not go back far enough. Then for the period of April 2023 — July 2023, we switched to using the PCSv3 data. However, using this data does not have any impact on the validity of the test as we only used the swap transactions information to build up the test; this is further described below:

Assumptions and methodology: Below are the key assumptions in running the backtest:

  • CAKE rewards: we used the current CAKE price and the current emission rate when calculating farming yields so as to not over-estimate the potential yields in the live environment, because the historical price and emissions rate of CAKE was higher in the past.
  • Swap Fees: we assumed 0.05% swap fees for the entire backtest period — as this is the fee tier for the BNB-USDT pool we will be deploying in. This is a conservative assumption because the actual swap transactions in the back test (Aug 2022 — April 2023) were done on a 0.25% fee pool (rate for PCSv2.) One could argue that with the lower swap fees, you would expect higher volume to be routed through the pool.
  • BNB Price: BNB price at any point in time was derived directly from the swap information and the composition of assets in the pool.
  • Liquidity: Liquidity information of the pool at any given time was derived directly from the swap transaction size and the price impact related to the swap. Combining the liquidity information with the CAKE rewards and Swap fees, the % yields from these two sources could be calculated
  • Borrowing Fees: we assumed a constant 6% APR borrowing rate for both BNB and USDT
  • Price Range: + / — 10%
  • Hedging: we ran the algorithm to hedge 50% of the delta exposure daily. For example, if at the end of one day we had an exposure of $500, $250 of the exposure would be hedged. While this methodology might seem relatively simplistic, it provided a sufficiently accurate, conservative, and unbiased output for our purpose. A more complex model is not necessarily better here and could lead us into various pitfalls such as overfitting, look-ahead, and other data mining fallacies.

Results

The backtest results showed a Sharpe ratio of 2.33 and base return to investor (no leverage) of 23% APY. We expect the strategy to be run at average of ~2x with estimated APY of 48%, and an estimated maximum drawdown of ~16%, giving a RoMaD (return-over-max-drawdown) of 3.

📅Parameters

Borrowing Interest Rate

As previously mentioned, we have designed AF2.0 with the ability to have multiple borrowing interest models. To create predictability and operational stability, we will be implementing a flat borrowing interest rate model of 6% for AVv3.

Note: For simplicity and ease of implementation in the initial stage, we will use a flat slope for all utilization levels. However, in the future, we might create a positive slope curve at high utilization to also incentivize managers to return some debt. Given the vault’s controlled capacity at launch, this should not be an issue.

Fees

We will institute the same fee structure for AVv3 as we used for AVv2. The distribution of fees will also follow the same structure as AVv2. (Ref)

  • 2.0% Management Fee
  • 15% Performance fee which will be assessed on the $CAKE farming rewards and the trading fees rewards
  • 0.3% Withdrawal Fee

💂Safety Guardrails

As mentioned in the previous article (AVv3 Progress Update), AVv3 is modular in design and provides the vault manager with more freedom to operate. The role of the smart contract is to act as a container to facilitate the execution of strategies, and check that the managers’ actions fall within the acceptable risk parameters and guardrails. We list these guardrails below:

Net Exposure Tolerance:

  • Description: the smart contract would check the vault’s net exposure before vs. immediately after the manager’s action and not allow any hedging transaction that would make the net exposure of the vault worse off. For example, a market-neutral vault would want a 0 exposure on BNB. If the current exposure is slightly long BNB, then a manager would not be allowed to make a hedging transaction that would increase the BNB exposure.
  • Value: 0%

Note: this check does not apply to the “borrow more” transaction where the vault manager increases the leverage of the vault.

Swap Price Impact Tolerance:

  • Description: we will set a price impact tolerance on a swap (e.g., hedging, etc.) If a swap transaction has a price impact greater than the set percentage value, the transaction will be reverted.
  • Value: 1.0%

Max Leverage:

  • Description: We expect normal operation to run between 2–3x leverage. We will set a leverage limit for the vault, beyond which would not allow for the manager to borrow more assets.
  • Value: 8x

Equity Change Tolerance:

  • Description: The equity tolerance acts as a final safety check that the transaction performed by a manager is safe. It will check the % change in equity before vs. immediately after the transaction and will revert the transaction if the % loss is too high.
  • Value: 0.25%

🔒Security

As with any major product release at Alpaca Finance, we practice utmost caution to prioritize safety. For that reason, we had the AVv3 code audited by Inspex. The final audit report will be shared soon once it is completed.

Thank you for taking the time to read this article and for your support. We’re excited for the launch and hope you are too! Please follow us on our social media channels below to stay updated on future developments.

Website · Telegram · Twitter · Discord

--

--