Restoring the Peg Mechanism

As we restore reweights and direct incentives, we should take the opportunity to make adjustments to the mechanisms based on learnings from the first week of the protocol being live before everything was paused.

FIP-2 created a price floor for FEI at $0.95, which can help range bound the peg when there are large deviations either in demand for FEI or the price of ETH.

Observations on original mechanisms

Direct incentives:

  • The spread between the burn penalty and mint reward is a “no trade zone” where traders prefer secondary markets or not trading.
  • Having high potential burns makes it hard for oracles and aggregators to accurately determine the price of FEI on incentivized Uniswap pools.
  • Mint rewards can occur without a sell penalty if the oracle price of ETH goes down
  • Do not directly incentivize peg restoration, but rather facilitate deflation which reduces the supply

Reweights:

  • Incentive parity is a complex trigger condition dependent on time since peg restoration, distance from peg, trading volume, and ETH volatility
  • Arbitrage with a “true price” of FEI between the mint and burn prolonged the time to incentive parity

Fixed Cadence Reweights

One solution to solving the issues with reweights is to have them occur regularly with a predetermined time interval of every few hours for example. This allows traders to estimate the amount of time before which they can sell at a profit for supporting the peg.

This has been suggested in Reweight Without Reweights (using cumulative deviation)

An issue with this approach is that reweights can be sandwiched by arbitrageurs up to the minimum distance from the peg required for a reweight (currently 1%).

However, if the sandwich provides a guaranteed profit, then traders are incentivized to buy the price up to the minimum distance from the peg before the reweight occurs. They will then take profits on the arbitrage in the spread between the peg and minimum distance. This should bring the price close to $.99 closing the gap towards $1.

The primary question is what time interval is best for the reweight cadence. Something on the order of 2-8 hours feels appropriate. Having a predictable and frequent reweight will help bring confidence to traders in supporting the FEI peg.

Capped Direct Incentives

Capping the burn incentive to a smaller value (1-3%) treats it more like a trading fee and less like a spread that widens to unpredictable sizes further and further from the peg.

This makes it easier for trades to occur, for traders, oracles, aggregators to estimate and model for the worst case outcome even under extreme conditions.

Larger values provide additional deflation to FEI indirectly helping support the peg and collateralization ratio, while smaller values reduce trading inefficiencies and integration issues.

We should add a cap even if it is much higher like 5-10%. The number chosen would also be informed by the decision to remove the mint reward discussed below, which would better suit a value between 1-2%.

Removing the Mint Reward

The mint reward’s primary function is to close the gap between the burn penalty and the spot price. If the gap is capped to a small enough value, then the mint reward’s utility is diminished.

Removing it would also remove potential economic attack vectors and inefficiencies. An example is minting inflation caused by the price of ETH going down. The price of FEI would be diminished without the necessary burn to ensure deflation. Therefore the mint reward caused net inflation in the FEI price.

Looking towards FIP-3

Code for fixed cadence reweights can be reintroduced separately from the changes to direct incentives over the coming days. Here is a pull request containing this code: Fixed Reweight Cadence by Joeysantoro · Pull Request #96 · fei-protocol/fei-protocol-core · GitHub

The direct incentives changes can be included in a subsequent proposal which also fixes the vulnerability disclosed through Immunefi.

12 Likes

I support fixed cadence reweights, and since reweights are part of the whitepaper no vote should be needed. Just implement it ASAP. We’re already pro-reweights by buying in originally. Voting on it again is redundant and a time waster.

6 Likes

i think the 1% sandwich arbitrage is a low price to pay, given where we currently are at now.

to the time interval, in my experience as a prop trader, I would lean on the side of a shorter interval for reweight cadence, so I guess in terms of the range proposed, 2 hours. here are my thoughts on why:

First, if fei has had more time to depeg, so is the amount of size available to be executed in the arb. as the size grows, slippage grows exponentially and becomes a larger risk factor. This problem is only compounded by the fact that the longer the interval, the more participants, and therefore even more risk to get destroyed on slippage, as everyone is going the same way you want to.

Second, In order to perform the arb, I’ll have to use a good amount of cash. this is not very capital efficient, and given that yield farms are still paying well. if the sizes are too large, it won’t be worth if for a large majority of traders to move their money out of the farm in order to arb FEI. as a result, only big players with large amounts of excess capital will be able to perform the arb.

Both of these factors contribute to a more unstable FEI price, as they push smaller traders away from wanting to do the arb, and gives those with more capital more of an advantage. Ultimately, the FEI price, if enough traders are pushed away, could be at the whims of a few traders, who if they chose to stop, the price would collapse.

This being said, reweighing more often does have a set cost to the PCV along with gas prices. What is nice is that both concerns I have with too long of a reweight time are quantifiable, as the amount of different participants, theoretical arbitrage size available, etc, etc, are all things which can be monitored and analyzed. Along with this, as more money comes into FEI and liquidity in defi in general improves, the amount of time between reweights can be increased with little to no consequences.

I think it is best to, therefore, start with a lower reweight time than what may be needed, and then increase or decrease the time dependent on the realized cost of reweights and performance of arbitrageurs.

3 Likes

I agree, we can also consider a few other parameter adjustments like:

  1. lowering the FEI reward for triggering a reweight, currently 500 FEI (could even be 0 tbh as the arb is enough of a reward)
  2. Moving 150k ETH back into the Uni pool since the stabilizer is not being fully utilized. This would make the reweights more effective as they operate on a larger pool
  3. reducing the min distance to something like .5% instead of 1%
5 Likes

interesting points! here are my thoughts

  1. the arb theoretically should be enough of a reward yeah, however if a trader or trading desk attempts a reweight and loses money from the arb, there is a good chance that they will avoid FEI for at least a while. A reward that allows traders to scratch easier on a failed re-balance is a small price to pay for continued future liquidity

  2. that’s a very good point. I wonder how the gas costs on a contract that drips funds back and forth between uni pool and stabilizer based off stabilizer utilization would look, or how much such a product would help

  3. same as point 1

Another possible parameter/idea i’ve had is to force reweights based on the current (calculated) realizable impermanent loss that the pcv will suffer, like a stop loss per say. At the very least, that’ll be the number I’ll be checking once all these reweights start happening

2 Likes

Given the passage of FIP-2 and the release of most sell pressure; the most likely situation where FEI would deviate from the peg rate would be when ETH fluctuates wildly.

Barring ETH fluctuations, 8 hour reweights intervals are more than adequate. But as observed in previous market crashes, ETH can very well crash more than 5% within say, 30 minutes. Given the very large LP depth, such movements will open up a massive arbitrage window; as ETH price plunges buyers can buy in tens of millions of dollars without bring FEI back to $1, and simply wait for the PCV to cover their exit in the form of reweights.

Therefore I would advocate for another additional measure, at least during the period of timed reweights; an oracle based trigger for reweights. IF the protocol detects a very large ETH movement (say 3%+) between each oracle refresh, that should instantly trigger a reweight; and ideally on the same block to front run arbitrageurs.

In this long run, this is perhaps more elegantly achieved by diversified PCV assets and exponentially faster reweight timers based on distance to PEG. (and other fine methods discussed in the previous threads) Seeing that there are considerable debate regarding the future of burn and mint rewards, it could take a considerable amount of time between the implementation of FIP-3 and a full restoration of mechanics and parameters. Therefore, there should be a stop gap measure to reduce potential losses to the PCV.

2 Likes

Please convert all eth into usdt and peg usdt fei on uni price always up 1.01 that’s great don’t make simple task into complicated :grinning:

USDT is still a little shady. Better to peg it to USDC or GUSD. They are audited and backed 1:1 for sure.

I agree that there needs to be both a time-based and a price-based trigger for reweights with the price-based mechanism as an ‘emergency release valve’ as price swings can be very volatile.

2 Likes

FEI reward for triggering a re-weight can be reduced to 250 FEI. Don’t think we need 500 FEI to have someone to pull the trigger.

2 Likes

Depends on the transaction costs, right? If I understand it correctly even lower should do the job on L2, with cheaper gas

It seems better to reduce the min distance to .5% and even to .1%. This helps to stay closer to the peg. The product we have is FEI at $1. When we have sth different of it, we could execute the reweight. What would be the disadvantages of choosing a min distance of .1%?

I like the idea of 4 hours minimum time interval as it brings confidence to support the peg, but also give some time to market forces find a price equilibrium. Less than that seems to be just a $1 redemption available all time.

I agree, the incentive to buy FEI below $1 is the expectation that will be back to $1. I always prefer simplicity when is possible. Less is more.

I agree, it is like a trading fee to access the huge liquidity below the peg. If min distance is 1% and trading fee is 1%, the equilibrium FEI price before reweight will be maximum of 0.98. That`s why seems more interesting .1% min distance for reweight and .5% - 1% trading fee, to have FEI price closer to $1.

The biggest penalty for sellers below the peg could be the opportunity cost of not selling it at $1. But due to AMM Uni logic, it is difficult to sell at the peg. Assuming the majority of ppl cannot sell exactly after reweight, everyone will be paying this fee. And the price on secondary mkt will converge to FEI - trading fee. As we aim to be very close to the peg, I think this fee need to be low.

1 Like

Totally agree with @cozeno; well diversified PCV assets is likely help maintain FEI price…
Perhaps it will be wise to diversify first, minimize reweight time & percentage distance as @joey suggests and then insert new bonding curves immediately for other cryptos . I not in favor of incentives/penalties for a stable coin which create multiple prices; an efficient arbitrage mechanism will do the work…

Arbitrageurs could front-run the oracle updates so this wouldn’t really help. Ultimately arbitrage would lead to more profit for PCV than loss for the following reason:

  1. Assume traders always buy up to the minimum distance from peg or the min distance for profit (2 x 30bps uniswap fee)
  2. The profit from arbitrage is the distance from the peg is at reweight time minus the 60bps in fees

The arbitrageur makes 40bps and pays Fei Protocol 60bps AND helps restore the peg before the next reweight.

Having faster reweights at further distances from the peg is an interesting and intuitive idea although it would require a rearchitecture and reaudit of the contracts so we could save it for a future update if we observe issues with fixed cadence

2 Likes

The amount of FEI burned from a reweight is approximately equal to:
distanceFromPeg * totalPCVFeiLiquidity

So lowering the minimum distance would make it take longer to reduce the supply to a given level that will meet demand at $1. Reweights also cost the protocol ETH paid to remaining LPs, so having too frequent of reweights would be bad. I think we could consider .5% for this proposal, anything less than .6% is almost the same because that is the point where arb is profitable while all of the liquidity is on Uniswap. Once we get on other AMMs with lower fees we could lower this number more

2 Likes

Hello,

Thanks for responding Joey; I am specifically addressing the situation akin to the crash of April 17th, or potentially greater crashes. If ETH goes down to only 2500 within 90 minutes, Then FEI will trade at $0.83 more or less, until whenever reweight kicks in. If reweight does not happen imminently, the current liquidity pool can support over $100M of buying, at an average price of ~$0.92 before hitting the “minimum distance for profit”.

You are correct that FEI protocol will make 60bps in this process, but the PCV will ultimately have increased its net liability by the orders of tens of millions of dollars.

Given the average realized volatility of ETH, the event that I have described will inevitably happen, and most likely before FIP-4 is implemented. I think without question FIP-4 will need to have some kind of mechanism that pushes for faster reweights at greater deviations to the PEG. I understand that any new code beyond reweight on a timer will call for auditing; therefore my statement primarily serves as a reminder to proceed swiftly to FIP-4 as to avoid outsized reductions to the collateral Ratio.

Best

1 Like

Hmm it seems to me that minimum distance works more in a scenario of low sell pressure. If we have high sell pressure, FEI price can be lower (close to the floor if we still have it) and reweight can burn more FEI, reducing the supply when we need most. The supply reduction seems not so different in this options of min distance .1%, .5% or 1%.

Can we estimate how much is this cost?

Yeah, I forgot about the .3% pool fee. The .5% makes sense for now.

I agree with cadence reweights, arbs need certainty to execute a trade.

I’m hesitant on reintroducing penalties, as this can also stifle use, as we saw with the previous big penalties. A smaller penalty might help, but if you also reduce the mint reward, what incentive does a buyer have to want to buy fei?

There is not enough integration within the defi space, and the potential to instantly lose 1-3% on your trade may be a major deterrent in defi users adopting fei.

I would suggest removing DI, having the scheduled reweights and wait for more integration of fei before reinstating dampened DI.

This is an interesting point.
To rephrase what @cozeno said, the PCV doesn’t just hold ETH, it makes market with it. So when ETH crashes, PCV loses extra from the impermanent loss, since it will be selling ETH at the pre-crash price. And he is suggesting that PCV be a more active market maker, and change the price of FEI by reweighting in times of ETH volatility.
I think this is an important point that may need to be addressed, but I haven’t yet wrapped my head around how exactly IL interacts with reweights. Uni v2 AMM (xy=k) is the simplest market maker that never reweights; ironically, it works well because of this simplicity. In Uni v3, we are seeing sophisticated market makers that “reweight” all the time. They also work well if they have a good reweight strategy. The PCV is an in-between market maker that reweights deterministically to keep the peg. Does this make it a very bad market maker that makes losses when it shouldn’t? Can price fluctuations around a constant long term trend hurt the PCV? How would savvy traders trade against a market maker who follows a deterministic reweight schedule?

Reweight without reweights implicitly assumed that ETH remains stable. The assumption may have been fine when the volatility of FEI itself was the biggest concern. But once we start considering ETH volatility, reweight is no longer just a pegging mechanism; it is an active market making strategy.

3 Likes