ETH-UNI Strategy (IL)

[YFI-Cent]

I know this was posed before by @mattdw but I want to re-introduce this where I’ve gone through this more thoroughly and added some modifications. I’ve quantified the impermanent loss risks below so people can better understand them. I think we should develop IL strategies and over time improve the IL hedging perhaps with options. For now I present the below:

Proposal yETH Strategy:
This is a ETH/DAI Uniswap farming strategy using the yETH vault although this could also work for the yWBTC vault. As you can imagine the flow is:

1. User deposits ETH into yETH vault:
2. Some % of ETH goes to MKR vault (Based on Strategy Collateralization Ratio – I am proposing 250%)
3. DAI is generated in MKR vault subject to Strategy Collateralization which should equal the value of ETH not deposited into the MKR Vault. See the example below using 10k ETH deposited at a $400 ETH/USD price:
4. All farmed UNI is converted into ETH is held separately and is incorporated in any subsequent rebalance event. If no rebalance, just accumulates as ETH profits.

As mentioned above this strategy clearly has impermanent loss risk. There are a few constraints that we have to work with:

1. We have DAI debt that can’t breach the MKR collateralization ratio floor of 150%. Once we get close to that level a rebalance in required.
2. The other is that we would prefer to not rebalance as that crystalizes impermanent loss. We need to set the Collateralization high enough that it gives us room where the ETH/DAI ratio can float without being forced to crystalize impermanent loss (particularly to the downside). I am proposing to rebalance on an upside move of equal magnitude, but that is an (open question). Based on a 250% collateralization ratio we can incur a 40% decline (and 66.6% upside) and not have to crystalize impermanent loss. See the downside example below. This would imply a $240 ETH price down and a $666.66 ETH price up. The impermanent loss is 3.17%, but only on the Uniswap notional amount (not the full vault size). This equates to a $58,065 or 241.94 ETH ($58,065 / 240 USD/ETH price) permanent loss once rebalanced.

Once the MKR vault breaches the 150% threshold (or we can set how close we are comfortable) we now have to rebalance. A rebalance to 250% will look like the below. The total position is calculated in terms of a Net ETH position and the process is started again as if that amount of ETH was originally contributed to the vault. The net changes are made below to effect the rebalance. In this example 172.81 ETH is withdrawn to from the MKR vault and converted to 41,064 DAI which then pays down the DAI debt. 901 ETH and 213,989 DAI is withdrawn Uniswap. The 901 ETH is converted into 213,989 DAI. The combined 427,978 DAI is also used to pay down DAI debt and then we are fully rebalanced.

Returns
Based on a $3.00 UNI price and $50k in daily fees in the ETH/DAI pool and a 2% DAI annual interest rate we should generate approximately a 15% annualized return before accounting for impermanent loss.

Given we have a 40% downside and 66% upside range where we aren’t crystalizing impermanent loss that should give us plenty of room to earn before having to rebalance. We can make the range bigger but at a cost of lower annualized return (due to higher collateralization ratio hence smaller Uniswap pool) and a risk of a larger crystalized impermanent loss at given a wider range. I think 250% strikes the right balance. At a 40% downside we would crystalize a loss of 241.94 ETH which is approximately 2 months of farming. At a gain of 66% we would lose 145.16 ETH which is approximately just over 1 month of faming. We could have some combination of 6 downside rebalances or 12 upside rebalances where we would lose money break even. The main risk would be a massive gap risk to the downside where we couldn’t rebalance at our designated 40% downside but ended up being much lower than that.

DAI Liquidity
The only time DAI liquidity becomes an issue is when we have to rebalance to the downside we need to sell ETH to generate DAI to pay down the debt. Just a matter of how much price impact there is. We can mitigate this by making sure we are not bigger than a certain percentage of the Uniswap pool. For example if we had 1,000,000 ETH ($400,000,000) in the vault initially at 400 USD/ETH, on a 40% downside event we would need to sell 107,335 ETH for DAI to rebalance. This equates to ~25.5mm DAI. Our existing uniswap position (not counting the rest of the pool) would have ~67mm DAI so the DAI would be there its just a matter of the price impact of the rebalance. Maybe someone can help model that to determine what % of the uniswap pool we should be to not generate an adverse price impact on the rebalance.

OPEN QUESTIONS:

1. What should the MKR Collateralization ratio be? I propose 250% since it gives room to float down 40% and up 66.66% from initial ETH price before any action is required.
2. Do we want to rebalance to the upside and if so at what increments? I propose yes at 66.66% upside relative to last rebalance.
3. When we rebalance from a 150% MKR vault collateralization do we want to rebalance back to 250% or to some lower number? I propose 250%
4. In order to allow continuous deposit and withdrawals I believe you need to unwind a slice of the vault meaning unwind their percentage of the uniswap LP pool, pay down their proportion of debt and return their NET ETH position. Likewise contributions would need to generate DAI in the MKR vault to maintain the current (not initial) collateralization ratio and then contribute to uniswap proportionally. I think this needs to be done to make sure impermanent loss is not shifted between different yETH vault contributors. Can someone confirm this? Is there a better way to handle deposits/withdrawal?
5. Should we limit the strategy to some percentage of the total Uniswap pool? I’m not opposed to being no more than 20% of the Uniswap pool to mitigate price risk on downside.
6. Happy to have someone check my calcs as well.

Let me know what you all think.

[ejbaraza]

I think that this would be a great idea if we pair the strategy with options from Opyn/HEGIC. I don’t know how it could be automated, but would like to see what others think about the merits of such an idea.

[milkyklim]

Question, answer to which I didn’t find in the OP, why would people use this vault with ~15% APY if nearest competitor offers 38.54% APY (or ~29% if we don’t account for $PICKLEs)?

Or more generally, how one can get better performance on (IL) strategies where extra tokens are distributed by other projects?

[YFI-Cent]

@milkyklim To answer your question in an open source world where any strategy can be copied, economically, any strategy that incrementally provides a token of any value will likely outperform those that don’t unless there is a specific moat that gives one team an advantage that can’t be replicated. Those advantages in our case are access to MKR OSM (which ostensible could be replicated), a highly skilled team which give people faith in security of funds and first mover advantage to give some scale and save of aggregate gas fees. To be fair I’m not sure this would compensate for the current 9% pickle yield, but that yield is not sustainable in perpetuity.

Also the pickle jars are a little different animal (although technically they could just copy the code for this and add pickle rewards to it). Two points:

1. We are actually earning the same amount as the pickle jar (less pickle rewards) because we are putting in 56% of our notional (since our notional is 100% ETH and we need to keep 250% CR at MKR). However, this strategy allows people with 100% ETH exposure to participate as oppose to those with only 50% ETH / 50% DAI positions. If I own 100 ETH and want to farm, but want to maintain my 100 ETH exposure either you would have to manually re-create the strategy I propose (and earn 29% on 56% notional or about 15%) or sell half of your ETH for DAI to do this.

2. Pickle does not manage impermanent loss at all. To walk through an example. Let’s say you start LP’ing in the ETH-DAI pool when ETH was at 400 and DAI was at 1.01 if ETH drops to 240 we would automatically rebalance the pool. The drop to 240 would cause a 3.17% permanent loss for us once rebalanced and a 3.17% impermanent for pickle jars. If ETH further dropped to 144 we would only suffer another 3.17% (im)permanent loss, but the pickle pool would suffer a total of 11.76% impermanent loss. It gets exponential as you go further down. Remember this loss is on the Uniswap notional not the total vault value. If ETH dropped to 100 our (im)permanent loss would be 8% vs. 20% for pickle. To be fair our loss is crystalized and if the ETH price recovers back to 400 pickle jars would suffer none. Same issue to the upside. ETH to 1600 would be a 20% loss for pickle holders, but only 8% for us. I believe it is better risk management. That’s why we need to set the band wide. We really don’t want to have to rebalance if we don’t have to as increased volatility dents returns. Also pickle jars dont have a MKR CR to manage.

[undefinedza]

It’s good to see Strategies being discussed that take on financial/IL risk. It’s definitely worth modelling this Strategy with hedging instruments (related Impermanent loss insurance).

The cost of hedging against IL will obviously eat into yields, but it may also allow us to be more capital efficient (e.g. less sensitivity to maintaining a high Maker CR in order to avoid crystallising IL) which will boost yields.

[YFI-Cent]

I would actually prefer to run this strategy with IL hedging, but don’t have a sense for pricing and we don’t have a solution yet so in the absence we can run it without. Eventually we can layer IL hedging on top of this to make it much better and more efficient. I really don’t want vault investors to take on any material ‘loss of principal’ risk.

Aside from managing IL, does anyone else have any comments or suggestions. Maybe as it pertains to contributions/distribution methodology, rebalance intervals, MKR CDP ratio, sizing limitations?