Saffron Weekly (11/12/23)

Dingo
saffron.finance
Published in
7 min readNov 12, 2023

What’s new for the month of November.

During our work on Saffron Fixed Income Vaults we’ve accumulated a wealth of knowledge from our research efforts. Today I’d like to check in with the community to give everyone an update of how progress has been going and delve into the research we’ve been conducting in order to bring DeFi more tools and more options.

As we’ve explained in the past, Saffron Fixed Income Vaults consist of a two-sided contract where users exchange the risk of a variable outcome from providing liquidity to a Uniswap Pool for a fixed-rate income paid upfront. You can find a complete recap of what SFIV is here.

Today we’ve collected extensive data on over 26 different vaults, and we’ll explain a few of the ways we’ve spent the past six months analyzing their performance and outcomes to optimize them.

When it comes to SFIV, to make this contract (or vault) a few parameters must be set:

  1. The Uniswap Pool
  2. The price range of the liquidity will be placed.
  3. The length of the contract
  4. The fixed rate of the contract

Our ultimate goal with this research is to find the parameters that make an optimal vault for both kinds of end users that SFIV serves — the fixed and the variable side. Essentially what we mean by this is that we aim to find the right set of parameters to identify “fair” vaults, which, by definition, would set these parameters with the best knowledge possible for both sides to profit on average.

However, while our Saffron Fixed Income Vaults are a sophisticated product they ultimately have a limited market unless we can set these parameters carefully. This prompted us to take advantage of the stagnant market and spend a significant time researching how we can make the most profitable vaults.

We gathered Uniswap pair candidates and downloaded their historical activity using Dune. Then, we built an internal engine that effectively simulates the vault performance and searched for suitable parameters that would allow us to create profitable vaults.

Methodology

Using the historical data, we calculated values for vault parameters such that, on average, a vault would bring enough profit to the variable side to cover the risk they’re taking.

However — as is the case with SFIV — the impermanent loss is somewhat more permanent for the fixed side due to the nature of their capital being locked in a vault for its entire duration. This removes the option of an exit strategy that would normally help to mitigate impermanent loss.

Bearing that in mind we tested new vault strategies to maximize profitability. First, we looked to limit the length of the contract. This would diminish the variable side’s capacity but helps decrease impermanent loss statistically (see Figure 1 below). However with this strategy there was still loss — on average — for the fixed side.

Figure 1: Average impermanent loss calculated as the value of the final portfolio compared to holding the initial portfolio using historical data for ARB/ETH (0.05%) Uniswap pool (Arbitrum chain). Simulations were calculated with the following parameters: Range was set as the current price and 5% above the price of ETH in ARBs, which makes a liquidity provision of ETH alone. We randomly selected 250 starting points for each measurement between 3/25/2023, 5:40:29 PM, and 7/18/2023, 10:13:47 AM.

SFIV as a call option for the fixed side

Another use case of the vaults for the fixed side SFIV as a call option to trade tokens with a specific price. This is achieved with a single-sided liquidity provision.

We started to test vaults that would place the liquidity at a range set between the current price and below. If the price went down, the assets would swap, and the fixed side would end up with their fixed income — and would have paid a smaller price than the spot price at the beginning of the term but higher than the end of the term. If the price goes up, the fixed side would keep their original position and fixed income, leading to an overall profit that equals the fixed rate of the contract.

We found that — with the right parameters — this strategy has an acceptable risk for both sides when we set vaults for high-volume pairs such as ARB-ETH and USD-ETH.

SFIV with limited IL for the fixed side.

Our next strategy involved using LSD token (wstETH specifically) to generate yield with much less risk of incurring impermanent losses.

We discovered we could predict the impermanent loss with a certain degree of variability of the wstETH-ETH pair on Arbitrum. The price of wstETHETH follows the APR from the Lido protocol. Using this knowledge we predict the most likely price of the wstETH at the end of the vault, and as a consequence, we could know the expected impermanent loss. Next, we could set the fixed income to beat the impermanent loss and add some gain for the fixed side. So far so good. The remaining question was if the variable side would be profitable with these parameters.

We found that there are a range of parameters that indicate profitability to the variable side (for example, setting the price range between the current price and +0.02% of the current price). Based on historical data, we know that the wstETH price should go up by 0.01% per day, which aligns with the 4.5% APR from LIDO. In most cases, within two days, the LP position should be entirely ETH.

So we got to simulating this data. We simulated 5000 contracts with random starting dates from 6/15/2023, 12:01:13 AM to 9/8/2023, 9:46:59 AM using historical data of the wstETH/ETH 0.01% Uniswap pool on Arbitrum. In our simulations, because of the impermanent loss, the fixed rate of 7% in ETH would lead to a 3.6% APR on average for the fixed side (median of 5.5% APR) against holding wstETH but a 7.6% APR on average (median of 7.6% APR) against holding the portfolio in ETH.

Our results indicate that with these parameters the variable side would be profitable and earned 11% APR on average. Setting the fixed rate so high makes the variable side lose significantly more than usual, in this instance the variable side lost money 43.7% the time. 50% of the time the variable side would lose at least 10.35%, Figure 2.

Figure 2: PnL of the variable side in 5000 simulations of ETH/wstETH pool.

The variable side was still in profit on average because around 20% of the vaults had outcomes with greater than 100% variable side profit, Figure 3.

Figure 3: Cumulative distribution of PnL outcomes for the variable side in the simulations of ETH/wstETH pools.

These results are summarized in Table 1 below.

Table 1: Summary of PnL analysis of 5000 simulations of ETH/wstETH pool.

Why haven’t we published more vaults?

As of right now we’ve had three vaults live on Ethereum to the public with another waiting to launch once the Fixed/Variable side are filled to 100%. However in order for these vaults to operate effectively, as outlined in this post back in Q2, there ideally needs to be a significant amount of volume making this financial instrument shine the brightest as there are more people interacting with the underlying Uniswap LP positions. This is also why we placed such an emphasis on privately testing and simulating data in the meantime. However, now that the market appears to be picking back up again it may an opportune time if you have a particular vault in mind that you would like to see launched — all you need to do is request a Vault to be opened with a specific pairing and parameters on our Telegram and we will create one for you.

What are we working on now?

Simulation results accounting for the variable side outcome as a function of percentage of time in range.

Above you can see the collection of our simulation results for the variable side of SFIV over the past several months. Now our focus is refining our research protocol and finding more use cases for this instrument. In the meantime, we are developing SFIV in other underlying yield-bearing instruments in which I’ll briefly elaborate on today. The product we’re building next uses the concept of the fixed income vault’s upfront payment to receive yield from an ETH staking position up front. That future yield can be used for anything, but we’ll have an option to automatically fund a position on dYdX or GMX to long ETH. This facilitates a long ETH position without risking losing any of your ETH principal, because you’re using all of your future yield up front to enter the long position. This boosts APR on ETH staking as long as the ETH price trends up. The product is thesis-compatible with staking ETH because staking ETH is — well, by definition — a long ETH position. We’ve done research and backtesting on optimal configurations and they all help boost yield as long as ETH is trending up even a little bit.

In addition, we’ve also been building a bitcoin wallet and exchange that facilitates trades for Ordinals and BRC20s. This will be completed and launched soon as we have been quietly working on this over the last few months. There’s a lot going on and there is still more yet to come of what the Saffron team has been doing behind the scenes, but those details will have to wait for a later date. Starting in January, we will be switching to a monthly release cadence going forward to keep you all abreast of our newest developments. If you have any questions about what you’ve read here today please feel free to reach out to myself and the team through Telegram or Discord. We are — as always — excited to hear your feedback.

Community resources

Fixed Income Vaults: https://vaults.saffron.finance/
Web3 app: https://saffron.finance
Governance: https://gov.saffron.finance
Telegram: https://t.me/saffronfinance
Discord: https://discord.gg/pDXpXKY
Twitter: https://twitter.com/saffronfinance_
Github: https://github.com/saffron-finance/saffron

--

--