Upcoming COMP farming change could impact the Dai peg

Farmers would pretty much be able to do the strategy by redepositing whatever they borrow to Aave and borrow DAI and redepositing again to Compound. You would get leverage that way and execute the 1 and 2 strategies above.

2 Likes

I think this means that you can’t do a flash loan so that you borrow $30m REP, poke the refreshCompSpeeds contract, then undo the flash loans, in an atomic trx. This would mean you’d earn way more COMP from Rep until someone else pokes the contract.

2 Likes

I think that means that only interactions from an external wallet will get COMP rewards any smart contract won’t get any rewards and that includes flash loans contracts

I was looking at what the real answer is here about this compound farming but have not taken on the risk of the new compound governance change here to DAI and maker.

@cyrus Your analysis looks correct.

The risk is of taking on the trade in/out over the time frame expected and prices of the assets. USDC has less of a pricing risk but ETH could be used to Mint DAI and swapped for USDC. There is a time/pricing on DAI risk here. How much of a premium that can be paid see numbers below. The other issue becomes if supply return rates rise enough to pull in more compound supply as the supply side of the equation that puts a rate and comp drag on the rest. I don’t think it is big enough to offset the leverage available here and the rate of return given comp drip and price in USD. So once this starts to move it will max out available supply depending on the DAI/USDC pricing risk over the time of availability.

It is entirely unclear how the farming mechanics will play out in detail but given the returns available below based on the new comp metric about to go into effect is significant and the analysis above pretty correct.

Here is what I have as a basic financial analysis on this.

$1B compound supply+borrow assets under deposit (doesn’t matter exactly HOW it got there - it IS there). The yield it is currently seeking is .5COMP/block or 2880 COMP/day. I think they lowered this slightly to .44COMP/block. We should check and post numbers for the calculations here.

Release 2880COMP/day valued at 240USD and compound is offering approx .7M USD/day for $1B of assets under depost (all that is changing is the value of comp and the amount of deposits seeking these 2880 COMP) that is all.

$1B earning .0007 or .07%/day available at current comp prices and current drip rate (doesn’t matter the model) works out to 29%/yr - on the entire $1B however it is portioned out (some deposits will earn more some less)

This is going to screw with whatever interest rate model you are dropping on top of it and changing the nature of how the comp will be rewarded just shifts the 29% yield around in those $1B in assets.

Given the size of this yield currently it is causing unusual behavior driving defi liquidity availability to extremes. Fortunately with BAT this has not seemed to cause a price problem. Given the recent increase in reserve requirements (50% on BAT/ZRX/etc) and interest rate models will drive assets into the lower yield spread instruments (USDC and DAI as well as possibly USDT to some extent) and if leverage can be applied to lever up this return it will.

I completely agree that something needs to be done, but it is unclear what can be done. Add more DC and all that will happen is that it will be soaked up - entirely.

The problem here won’t be the DC, but the PEG mechanics will be completely broken.

I do NOT think we can up the DC enough to satisfy this and the question becomes how do we do this safely with respect to assets. I know this is counter intuitive and may cause even more pressure on DAI/USDC PEG price in the short term but I would be prepared to soak up some of this problem with a somewhat higher SF (now rather than later) and allow the system to expand into this with a high DAI/USDC PEG rate against not just an across the board SF increase when the DC needs to be expanded but also a preemptive SF rate increase to build some surplus reserve incase things go whacky. Ideas to expand borrowing into B,C asset classes to expand the DCs with higher rates and higher LR as well as higher Liquidation Fees in particular may help.

Remember we still have a somewhat liquidity constrained liquidation system and we don’t want to stretch this. Right now the only deposit system that DOESN’T have an auction system is the USDC ones This needs to be fixed ASAP before Maker gets trapped in some wierd USDC/DAI price swing and can not auction the USDC position. We need USDC people to manage their LR’s against liquidations like normal. Whether to lower the LR there I’d say no. Honestly I would want to increase the LR on a new facility so that we can sop up some of the USDC looking to find DAI and reduce leverage available. People using ETH to borrow DAI should be ok. Be prepared to expand the DC’s into this as demand is driven into it and make bloody sure to increase the SF (now) rather than later. We already know the PEG price mechanics are going to be screwed due to this and there will be nothing we can do about that. Literally may not be able to expand borrowing into this to provide enough liquidity and anyone trying to take the DAI USDC short here could get really hurt in multiple ways until compound fixes the problem.

I honestly would consider governance minting ‘out of thin air’ DAI to exchange for USDC or other assets during this price rise to try to manage the increase and PEG somewhat. We literally are going to have to, by force and temporairly reduce the assets backing DAI, to make it LESS attractive to own it. Playing the liquidity game directly within the protocol and hazarding (to some extent) our DAI holders I think is the only way to have any hope of managing this situation so that the DAI USDC price doesn’t get totally out of whack with no way to bring it back.

I do have an idea for how compound can mitigate this issue that i have mentioned in discord #governance channel, but need to formally write up, but each time I have brought up pieces of this idea it has been rejected.

Conceptually the problem here as illustrated above is the high $$ comp value. The solution is basically to take the remaining 4.4M being portioned out and doing the following:

D1_D Democratically distribute N comp to all wallets that so that D1=wallets*Ncomp
a) Have at least one deposit tx to compound AND
b) Still have at least $100 worth of cTOKENs OR $100 worth of debt currently.
c) Drip it out at a rate determined by age of (a) the older the wallet the faster the drip of N to that wallet. or do it all as an air-drop based on a chain snapshot at some time T (preferred)

D2_I Use the total interest earned and paid on each wallet (i.e. real compound useage) prior to this debacle to calculate a air-dropped distribution to all the wallets. We voted at compound some time ago on what assets to add to the protocol based on interest/paid earned so I know this is doable. It is a fixed distribution based on the snapshot time T

D3_C Use the rest of the Comp to distribute based on a $$ value model of comp with a cap of D3_C/4years so that the distribution model does not distribute more than C% $$ value of comp for $$ Total value of assets so that the interest earned on the $$ value of the comp is no more than 1%/yr in USD value.

Additionally taking some restrictions off the holders of the first 5.5M comp distributed so they can sell would help but only after D1, D2 are distributed. (give your users with comp first shot at these markets as a reward for being there before everyone else)

D1+D2+D3=4.4M and D1+D2 should be >= D3 to deal with comp liquidity and price problem.

The point with the above is that there simply is not enough COMP distributed yet to drop the price to realistic levels. One could allow the insiders to sell it all but why not simply air-drop COMP based on PRIOR use and performance to reward your users and help depress the price of COMP with supply. D3_C is the critical component to eliminate this COMP farming though. If one sets the yearly return on $$ value COMPed to be a small % in $$ value relative to the Total $$ supplied - adding assets will NOT increase the value relative to assets until the price of COMP drops so that the D3_C supply cap for the 4 year distribution works out. Basically this becomes ‘solved’ when the price in USD of COMP drops so that the max D3_C drop of .44COMP/block has a value < 1% yrly return on the supplied $$ value of assets.

Note 1% is arbitrary but needs to be reasonably low. IN this low rate environment 1% is literally double the current returns on most assets. It was picked to literally be as low as reasonably possible but still drip out some in D3 until comp prices normalize much lower. The other key component is to get liquidity out there based on some snapshot for the D1, D2 air drops of COMP. I honestly don’t know why the team there did not consider doing this so the markets instantly get a good chunk of liquidity so price wouldn’t get out of hand. The final piece is not allowing COMP not matter what the price to be driped out at a rate higher than 1% $$ return on total assets supplied seeking this return. I consider even 1% to still be somewhat high but tolerable as compared to the current 29% return rate on $1B of assets supplied.

3 Likes

I don’t think it’s crazy to mint out of thin air DAI to help ensure the peg. The only difference would be that I think it would be better to issue new MKR and purchase DAI with it. The new MKR will be a reserve similar to the surplus to address the DAI peg through arbitrage.

The fact that MKR is deflationary should not imply that when there is a critical necessity we can’t go to the market for funding. Anyway, it’s a personal opinion, could be very well wrong with this.

Ref increasing SF although it is necessary due to the short time frame between the event and our current situation it will likely be a non material event in the grand scheme of things.

2 Likes

I think one tends to forget or not realize that COMP farming at this point is only profitable for Institutional players, and/or “Whales” – you need serious money, above 10,000,000 DAI to make COMP farming worthwhile/profitable. If we are afraid that a bunch plebs like myself will break the PEG of DAI substancially, even leveraged at 4X–then we’re not seeing the full picture IMO.

Let’s not forget that Institutions prefer to settle in USDC, as oppose to DAI. Let’s not forget that COMP bag holders also own bags of MKR–after all, Compound is still a plutocracy, now matter how you slice it…

@ElProgreso I beg to differ regarding COMP farming being profitable. Literally it is the tx costs against capital here and I as a non-whale was able to apply as little as $10K additional to my own compound assets to get a boost of return that returns my tx fees in no more than a couple days (I am still looking at the details of my situation). I agree that whales have it better here because their return to tx cost is low but they are taking on a shitload of asset pricing risk. It is going to be interesting to see the BAT situation unwind after loading up 200K BAT of accumulated interest a day to the compound system. I have this at a couple million BAT in interest alone that will need to be actually paid to shift these assets.

So no - you don’t need serious money.to take advantage of this. But you do need something north of $1K (more like $10K or above) to have any hope of capitalizing.

So it’s only profitable north of 10k? Can we simulate the potential impact by having the total amount of CDPs open over 10k + current excess capacity (DC-amount already minted)?

I do not think it is a good idea to sell MKR to buy DAI. In effect all the system needs to be able to do is just Mint DAI - MIP14 the DAI out to another account to buy assets (how and who does this is a real question but it will be needed eventually). MKR by definition is already pledged to make up the difference. Running a negative surplus literally means the assets backing DAI are NOT 1:1 but if we buy say stable coin assets it will. Just the surplus will read negative -$XM DAI while another account with have surplus +$YM stablecoins (USDC and the like). This literally IS a DAI direct injection mechanism controlled by MKR through MKR governance backed in the end by MKR and probably the only solution to manage this.

As I look at it unless the stablecoin values really change this would still create roughly 1:1 DAI backing but still create DAI to inject via the asset buying mechanism. I would limit buying only to stablecoins though as we don’t want to drive asset prices but simply DIRECT INJECT DAI liquidity based on minting DAI running a negative DAI surplus against a + stable coin account. Not as a decentralized model or DAI facility for everyone to use but directed by governance the MKR holders that are going to back this with their MKR in the end.

The reasoning here is that governance won’t have to pay a stability fee or worry about liquidations doing this. The issue is that governance may not be fast enough to do this. What we really need is a elected team to do the fast paced management, guided by governance and risk teams on this but I don’t see it happening fast enough. Literally I don’t think Maker will be able to respond fast enough nor has the tools to act properly to try to rebalance the DAI market. The problem will be measuring what is needed and then trying to manage providing it. We literally are going to have to leave things alone until we see how bad the DAI PEG gets and then try to figure out based on how much DAI liquidity was soaked up by compound to see what the first pass looks like.

My suggestion is that we raise the SF to 1% now rather than later. Deal with the inevitable PEG hit until compound gets their shit together with a real tangible plan that stops this. Right now nothing I see in prop 10 (enacted) and prop 11 (soon to be) is going to change anything except drop this problem right into USDC and DAI laps. Will USDC with stable prices soak this up or will it drive against DAI.

I am with @cyrus it is probably going to drive against DAI until the PEG compensates, I think from a risk perspective we have to raise SF since lowering it later won’t do dink (DCs will be maxed out already) and we will quickly find ourselves back at my mint DAI run a negative DAI surplus and buy stablecoin solution to direct inject DAI liquidity to manage this. My problem is what if this cranks DAI up to $.5-1B. But if we throw a SF on this of at least 1-2% we can capture the added liquidity SF fees as a kind of risk compensator. In the end once this comes under control governance can use the stablecoins to unwind the DAI short stablecoin or stablecoin long positions and bring the surplus back to positive and then decide what to do with the extra DAI generated.

I have not had much time to look at the exact details on this but I am pretty sure the DAI PEG is probably going to see north of 1.03 for some sustained time if the markets fall we are going to have a hell of a problem with DAI liquidity for ETH or other collateral auctions. I think we need to warn borrowers to start paying back, which again is a reason to raise the SF. Try to reign in the system now and put a risk premium on all of this now before it gets out of control until compound reigns this in.

1 Like

The above is a rough estimate based on my own situation. I literally can’t even grok the scenarios from a how to run this computationally because there are too many moving parts. Compound being the biggest one.

I just know that the $10K I added basically will yield the tx fees in 1-3 days depending and then the added return on capital runs north of 30%/yr currently. I was looking at .15-.2%/day return on assets here but this is going to change with prop 11. Until I see this in my own account I won’t have a good handle. I have a rough estimate that it will only change me by 20% until the BAT unwinds and then my returns will drop accordingly. I will still get the comp but no longer the 13% additinal supply return on the BAT, ZRX. The comp return is sufficient to drive rates to 8-10% supply/return diff and still give a solid 10-20% return on top of this whatever asset is applied.

No matter how I turn this it is the $$ value of the comp being distributed against this $1B in assets seeking this 290M US equivalance of comp looking to slowly drip out over the past year.

I am hoping someone sees something other than me or has some good different ideas. Because the above is all I got here atm.

@MakerMan Yield Farming is a complex math problem, but with $10K or more–your allocation of COMP on a Daily basis (2880 per day), will be minimal. You’re better off buying the Token COMP on the open market as oppose to leveraging 4X, Yield Farming a speck of that 2880 per day and driving yourself into financial ruin IMO

Allocation of comp as a % is the same at 10K or 10M - this doesn’t change until prop 11 comes into effect but then follows the borrow. All leverage does is make more tx costs and for same capital ups the return. I actually am not using much leverage. Other players are. Biggest reason is that leverage here is strictly controls by Liquidation Ratios or Collateralization Ratios on assets where one can borrow.

It also takes two wallets and double the tx’s to basically do something like Deposit DAI, borrow ETH, sell ETH for DAI and repeat this or some variation of it with collateral to ‘lever’ up.

Literally the COMP return on capital if the return model was strictly based on a capital allocated scheme would be straight across all assets. Because they don’t take the total comp dripped and split is based on capital allocation vs. splitting this on asset classes they will skew again the return into the asset classes. @cyrus is correct in that the lowest rate spread assets will be the ones that will be racheted up. So BAT should drop and USDC, DAI will probably crank up. But there is no diff in someone levering $10K to 30K than someone levering $10M to 30M except in the details of how the COMP value is spread across the asset classes (which is varying). The total COMP from drip still remains the same and hence the problem remains the same depending on total COMP distributed and $$ COMP value.

I agree the problem is complex and it also is a moving target right now. What is NOT complex is the math on the available return here. Number of COMP * COMP dollar value / Total $$ in the system (borrowed/supplied) is the pure driving force here. The portioning out is only relevant from the asset classes affected and compound is not applying an across the board same approach. This literally is wreaking havoc in their own system as well but with 50% reserve fees at least they are picking up some risk value. This is a real problem Maker is going to be forced to reckon with and it is a fast changing dynamic given compound own governance process that is simply taking the same $$ amount rather than changing the allocation amount so the drip no more than a 1% $$ yield on the total assets.

Until this changes the problem continues and I as a market participant will take advantage like the rest.

@MakerMan Hi, did you consider ETH collateral? With ETH collateral it can make more sense to COMP farm ETH instead of DAI.

I posted some analysis above.

1 Like

ETH is doable but given the low SF on Maker I think it would be preferrable to depo ETH here and take the DAI to compound. I literally don’t have the time to spend to completely wrap my head around which collateral will end up being the best. Because of their useage based rate curves it is really unclear to me which one will win or what governance will do in another 3 days to jerk everyone around with this massive COMP return being offered.

Personally I wish compound governance would just pull the plug on COMP distributions entirely and revisit this more carefully as I don’t think the rest of the DeFI space (Maker in particular with DAI) is going to be able to handle the demand. It is going to skew the PEG pretty hard I think with no real way to address this with tools available. (SF, DC’s).

[x-posted from rocketchat]

As was suggested over the past few days, I think we should make plans to (a) lower the USDC liquidation ratio, and (b) add an ETH-B collateral type with a higher stability fee. However, I think that it’s actually unlikely that the COMP distribution change is going to result in much DAI getting sucked up, for the following reasons:

  1. post-patch, COMP yield will be independent of the interest rate earned/paid, and instead becomes proportional to

    $totalBorrowedInX / ($totalSuppliedInX * $totalBorrowedInAll)
    for suppliers on market X and proportional to

    1 / $totalBorrowedInAll
    for borrowers on market X. At current COMP yields, and in light of the popular supply/borrow recycling trick, this is purely a coordination/schelling point game where the goal is for everyone to supply/borrow as much as possible to the same market, and due to the high COMP yield the interest earned/paid on the borrow itself is less relevant.

  2. given (1), the best potential schelling point in the short term is one which can bring in the most supply, i.e. the asset that is the most liquid to borrow. I believe this means that the likeliest candidates will be ETH, USDC, or USDT (in decreasing order of likelihood). I believe ppunktw also already made the point on the forum post about ETH, that it will make more sense for farmers to directly supply ETH into compound (and then do the recycling dance to reborrow it, etc.), rather than mint DAI against ETH to supply into compound which reduces their yield by their collateral ratio and introduces liquidation risk. Buying DAI off of the market to supply to compound is more capital efficient, however quickly runs into liquidity issues which IMO puts a cap on the size of that market and makes it an unlikely schelling point (as opposed to ETH, USDC, or USDT where the amount that can be sucked in is virtually unlimited).

  3. as we have seen with BAT, both the theoretical and practical equilibrium is for farmers to crowd into the same asset, due to the COMP yield being proportional to the market’s share of borrows. Therefore, unless DAI turns out to be the schelling point, it is likely to be little affected.

  4. in the long term, if COMP yields decay (due to growing assets in Compound or the price of COMP coming down or both), then margins for farmers become squeezed, and only then do the compound interest rates start to become significant in the calculation again. It’s true that in that case the shape of the DAI interest rate curve might make it an attractive schelling point, however I think lack of liquidity will stop it from catching up with the more liquid assets which could have hundreds of millions of $ in outstanding borrows at that point.

Overall I think it’s very hard to predict what will happen with any of these incentive changes, and I’m not proposing any specific outcome with high certainty. However, I also don’t think DAI becoming the main farming asset with supply getting sucked into compound is the most likely scenario. I think we should prepare measures (a) and (b) for that case anyway, however I don’t see a reason to panic yet.

EDIT: an earlier version of this post had an error in the heuristic formula for COMP yields in point (1), which I have attempted to correct. I think the conclusions from it remain the same though.

13 Likes

Agree, from what I have seen on Compound so far, larger whales determine markets that receive most of COMP rewards and its more probable ETH becomes dominant assets that gets recycled and less likely DAI, despite its favourable rate curve.

For instance BAT dominates Compound now not just because of steep rate curve (current COMP compensation is based on rate levels) but also because few BAT whales showed up and recycled (borrowed and resupplied) its BAT assets 4x+ (yes actual BAT deposited on Compound is less than 25% of what you see).

I think its more probable we see similar behaviour with ETH than DAI because there’s simply more ETH whales than DAI whales. As Lev said I also don’t believe ETH whales are willing to buy DAI (slippage issues) or mint it (liquidation threats). Potentially they could though swap ETH for USDC and mint DAI, but then you have to ask yourself why ETH whales would want stable exposure just because of COMP rewards. I am worried though that lower LR for USDC might be used by USDC whales to mint DAI and supply and recycle at Compound. So maybe lower LR USDC vault is not the best idea?

Again, it is hard to forecast anything and I think its more probable DAI won’t dominate unless some weird dynamics (increased refinancing from Maker to Compound) boost DAI markets on Compound and overtake ETH. I would hope thought that ETH whales simply dominate from day one.

4 Likes

My assumptions and solutions on how new COMP farming might result in.

With the new COMP vote going live tomorrow. All the other markets both lendings & borrowings will start to shift to DAI. Just like the BAT which went from $3,000,000 supply to $300,000,000 within 2 weeks. As most of the BAT were the same BAT going in supply and borrow but the market difference of borrow and supply went to more than $30,000,000 meaning 10times more BAT ended up locking in Compound which also increased the BAT price because of higher demand in the market. Graph attached from coinmarketcap of BAT prices over the course of the last 2 weeks.

Now with all the market shifting to DAI includes BAT, USDC, USDT primarily will boost the DAI demand in the market. It can also result for Vault owners to move their Vault to Compound which wasn’t relevant till now but after this new vote, it’ll be.

Problem: an increase in DAI demand will result in DAI peg making DAI > $1.

Solution: Vote to make USDC-A liquidation ratio at 110% which will help users to take arbitrage advantage up to 10x and instant increase in DAI supply. This looks like most reliable to me for the short term DAI stability.

1 Like

Have you seen Lev’s argument here? Upcoming COMP farming change could impact the Dai peg

You can with a different address.

Dharma has submitted a new proposal to update the DAI interest rate model:

This might take some of the pressure off DAI by widening the spread between borrow and lend rates.

1 Like