There have been a few excellent articles and Twitter threads highlighting the current situation on Compound Finance where a trader has opened a massive loan using 7% of the supply of Augur’s REP token. In this article, we’ll take things one step further and explore what could happen next thanks to the unique features offered by Compound. First, here’s a selection of what’s been written so far:
- 12/4 Max Bronstein’s Twitter thread
- 12/4 Eric Conner’s Twitter thread
- 12/5 Hassen Naas’ “What’s going on with Dai on Compound.Finance”
- 12/6 Matteo Leibowitz’s “A Tale of Three Dapps”
To summarize, a large trader, who we will refer to as “LT” moving forward, sent 7% of the total REP supply to Compound as collateral for a large DAI loan. LT actually kept another 3% of REP’s total supply in a separate address, so in total LT controls over 10% of the REP supply. LT also appears to own over 1% of the supply of MKR tokens, and has a large Maker CDP open with about 1/1000th of the total ETH supply with a value over $8 million today.
Thanks to excellent analysis and investigation by the authors of the above threads and articles, we know that LT used the DAI loan to purchase ETH which they added to their existing Maker CDP. In effect, LT used their REP tokens as collateral to increase their margin long position on ETH.
Matteo, from Crypto Chat, outlines several possible actions LT could have taken instead of using Compound for a loan. These include selling REP for ETH or borrowing ETH directly from Compound, and helps us draw some implications about LT’s motives. However, since Matteo’s article, LT has added more REP to the loan from Compound which seems to support the idea that LT simply wants to increase exposure to ETH at these prices without reducing their long exposure to REP.
The impetus for this article came from a debate around a point Matteo brings up in his article: if LT defaults on the REP loan, will Compound’s liquidation mechanism function properly and prevent losses for lenders? After all, as Matteo mentions, the protocol only offers a 5% incentive to market participants for repaying the loan. If liquidation fails to proceed in an orderly money and there is a run by lenders to withdraw capital, many lenders, who are currently supplying over 2 million Dai to earn the lucrative interest rate, could be at risk of principal loss.
After analyzing the contracts, I believe the incentive mechanism is likely to function smoothly, but the outcome may augur poorly for the price of REP (sorry… had to). Let’s look at the Compound liquidation process to analyze why.
If the value of LT’s collateral (REP tokens) ratio falls below 150% of the value of LT’s DAI loan, anyone may then call the liquidateBorrow() function on the Money Market smart contract. Under this scenario, Compound basically offers up a portion of LT’s REP tokens at a 5% discount to incentivize market participants to return LT’s loan to a 150% collateral ratio. Now as mentioned, Matteo expressed concern that a 5% discount may be insufficient incentive given the transitory holding risk of an illiquid asset like REP. This is a legitimate concern as REP’s price could easily fall more than 5% before you manage to send it to an exchange and sell it. However there are two important points to highlight:
- You earn 5% not once, but on each and every liquidation transaction. Given the significant size of LT’s loan, a small trader with a little scripting could run iterative liquidations and repeatedly compound their capital by 5% each time.
- Compound has altered the overall ERC20 market structure by giving us a large, liquid market to borrow and short REP, allowing one to neutralize the risk of holding REP while liquidating LT’s loan.
Let’s consider combining these two ideas into a low risk trading strategy. You could split your funds 60:40 into two sleeves. The larger sleeve will be used to borrow REP short on Compound. The 60:40 ratio was selected so that given the %150 collateral ratio, the REP we borrow will be equal in value to the second portfolio sleeve. Realistically, you would want to split closer to 70:30 so that your loan has some margin of safety from liquidation itself. Here are the steps assuming a starting portfolio of 100% DAI:
- Take the larger sleeve, use Compound to deposit DAI and borrow REP. Your wallet now contains two sleeves of REP and DAI of equal value.
- Send the REP to an exchange of your choice. We need to sell REP for DAI to turn this into a proper short position.
- Go ahead and sell the REP while also liquidating LT’s loan with the DAI in your wallet. You are now equal parts long and short REP, although you earned 5% extra REP from the liquidation.
- Send REP from your wallet to the exchange, and the DAI from the exchange to your wallet. Rebalance the two sides if needed.
- Return to step #3 and repeat. Continue until LT’s loan is back to a %150 collateral ratio
Thankfully this can be scripted and automated without too much difficulty. Let me know in the comments if you’re interested in seeing some sample code. Now it may be apparent, but each liquidation grows your total capital by 2% (5% reward x 40% of portfolio), which you would split and add back into each sleeve of the strategy. Automating this strategy, you could begin to rapidly grow your initial capital with each successive liquidation. Upon reviewing this article, Calvin, the strategy lead at Compound, also pointed out that you are lending the original 60% of your initial DAI and borrowing REP, earning carried interest for the lifetime of the strategy.
This strategy, while potentially lucrative, is not fully risk-free. There is some risk incurred between steps 1 and 2 when the trader is net long REP. From that point forward, the strategy remains market neutral to the price of REP. However, the other risk to keep in mind is that if the price of REP spikes, the loan you took out in REP could be liquidated regardless of your net exposure. Remember that LT is well capitalized and could attempt to squeeze your short position, or they may simply want to defend the price of REP when they observe their loan being liquidated. However, it may be difficult for LT to pull this off quickly enough to cause a liquidation since Compound sources price data from three exchanges and uses the median value as the protocol price. Either way this risk can be mitigated by monitoring the price of REP and rebalancing collateral into your loan as needed.
Ideally, REP could be sold directly through a DEX such as Uniswap, eliminating the need to hold REP or open a short position at all. This would also greatly increase the frequency of liquidation transactions since the existing strategy is bounded by the deposit and withdrawal times of the exchange you select. Unfortunately, there is not yet a highly liquid DEX market available for REP, so the strategy described is probably the optimal approach until that changes.
The impact of this strategy would typically face a haircut at the very end due to the need to buy enough REP on exchange to cover your short. If you amass enough profits, the slippage from buying back REP could be substantial. However, the beauty with Compound is you can simply cover your short off exchange using the liquidateBorrow() function to draw out more of LT’s REP to close your position. Some have speculated that LT used Compound to avoid crashing the price of REP, but as you can see, it can be used by short sellers just as effectively in the opposite direction.
So to quickly summarize, we have a relatively low risk way for traders to quickly earn a modest but continually compounding return on their capital. There is some initial risk taken to borrow REP, but the lucrative returns outweigh this risk. The Compound team has also mentioned on their Discord channel that their partners, including hedge funds and other financial institutions, are engaged with the protocol and prepared to take on the liquidation process.
With all this being said, LT’s loan currently remains overcapitalized at +300% at the time of writing. REP would need to fall over 40% from its current price to close to $3.00 for liquidation to become lucrative. LT also retains substantial REP and ETH reserves (in their CDP) they could mobilize to defend their margin long plus a large stack of MKR they could potentially sell. However, as several prior authors have noted, the current implementations of decentralized finance apps like Compound, for better or worse, leaves user activity pseudonymous but open to scrutiny. LT is clearly an experienced operator with a large war chest and the technical savvy to programmatically manage these positions, but I do wonder if they have considered the potential Schelling point they have created by using 7% of REP’s supply as collateral for a margin loan. Large traders in traditional markets are forced to be quite careful not to reveal their positioning lest someone hunt their stop price in order to liquidate their position for a quick profit.
This story reminds me in some ways of the infamous London Whale trader at JP Morgan who, like LT, continued to pile more and more leverage into a massive margin short position on a credit spread between investment grade and LIBOR rates using CDS swaps. A number of hedge funds, ironically including a separate JPM trading desk, deduced what was going on and took the other side of the trade to drive prices up, liquidating JPM and forcing a $6+ billion loss.
No single trader had the scale to bring down the London Whale, but once his positions were inferred, the huge potential profits of liquidating JPM created a spiral even a massive war chest could not overcome. Could we see a similar dynamic here? Time will tell, but I suspect if we get anywhere near LT’s liquidation price (REP $3.20), we may see a rapid and reflexive collapse in the price of REP. Leverage has a way of making you look like a genius or a fool. Buffet said it best: “Only when the tide goes out do you discover who’s been swimming naked”.
Note: After reviewing this article, Calvin from Compound also mentioned that the team has been closely been monitoring this situation and is exploring possible techniques to obscure user transactions, perhaps using a protocol like AZTEC.