Beware illiquid markets: lessons from COMP price action

And some observations on market manipulation and the oracle problem

When COMP launched a week ago on Uniswap, most people were shocked to see it quickly reach $100 (implied network value of $1B) and flabbergasted when it ranged around $300 (implied network value of $3B) a few days later.

Almost everybody thought this was “overvalued,” including the Compound team that said in a governance proposal something to the effect of “we never expected COMP to be anywhere near $200 and didn’t design our distribution with that high of a price in mind.” Nowhere was this disbelief in the high price more evident than in the derivatives market where traders thought it was so overvalued that they were paying between 5-10% a day in fees to short the asset.

But leading up to the Coinbase listing, going long COMP was in my mind the clear winning strategy. Why and what can we learn from this? Beyond that, what are the implications?

Supply and demand

At first, I agreed with everybody else. COMP was overvalued and would need to come down. Why? Because of the way it listed. Like UMA just a few weeks ago, the spot market for COMP was created on Uniswap by the team (to my knowledge).

Here’s how it works:

The team creates a new pool, in this case COMP/ETH by depositing x number of COMP and y number of ETH, the ratio x/y determines the “listing price” for the asset and the USD value of all the assets deposited x + y determines the “depth” of the market—the deeper the pool, the less the price moves when somebody makes an order.

Like in other markets, if somebody buys COMP, the price of COMP goes up. And if somebody sells COMP, then the price of COMP goes down.

Because the market was created by the team, the price will not go below the “listing price.” As a result, traders are motivated to be the first buyers of COMP in this pool. Their trade is “risk-free” because if they are first, the lowest the price can go is back to the “listing price.” Because the pool is relatively illiquid, the price moves a lot with relatively low volumes. So these traders seeking to scalp some risk-free profits will rapidly push the price to an elevated level. We saw this happen with just a few traders putting millions into the market pushing the price up I think 5x (need to confirm) to ~$100.

Why does this listing method suggest declining prices? It doesn’t necessarily on its own (see UMA), but because COMP was getting issued every block to users of Compound, there was continuous selling pressure. And if the sentiment of the market was that COMP was overvalued, then the likelihood of COMP earners selling was high.

Additionally, the team and investors hold ~46% of the total tokens compared to the paltry ~0.03% of tokens that were getting issued every day. Details are sparse on when those tokens unlock, so some traders were fearful that this massive pool of tokens would lead to overwhelming sell pressure.

This line of reasoning (combined with beliefs that the asset was overvalued on fundamentals) is why the market was so short the token.

Why were bears wrong?

They overestimated sellers and underestimated buyers.

Most important on the the sellers side, team and investor tokens did not hit the market. Things would have gone upside down if those tokens moved. The chances of those tokens hitting the market before the already announced Coinbase listing were close to zero because the team knows that allowing even one investor to sell a fraction of their shares into such an illiquid market would cause a panic. But not everybody believed that (I saw some comments circulating that investors had no lockup and could sell at any time, so there was some fear in the market).

The buyers side is harder to gauge. I didn’t expect many organic buyers wanting exposure to the token (a) because market sentiment was so negative and (b) the token was earnable through “yield farming” on Compound. I thought there were some parties with an incentive to offset the selling (namely locked up token holders and high volume yield farmers), but felt low conviction here.

Because of this, I expected the price to stay relatively flat. Some buyers, some sellers, and the market doesn’t really price these assets on fundamentals. With the launch of the FTX Perpetual Swap, the way to play the market was to go long the perp and collect the funding payment (if the market is short, the shorts pay longs and the market was very very short). Spot price might go slightly down but the 10% per day in funding should make up for it.

Illiquid markets

What I didn’t consider was the incentive to manipulate the market to squeeze the shorts. Teo shared a snippet from a post on the topic here:

Matteo Leibowitz @teo_leibowitz

manipulating COMP spot while holding outsized long perp swap position is the real yield farming

June 23rd 2020

10 Retweets57 Likes

The long and short of it (haha) is that because of the relatively large size of the COMP Perpetual Swap market, it would be profitable to buy the Perp and then buy spot in significant enough size to move the price, amplifying gains in the Perp and squeezing the shorts.

I am not a securities lawyer but based on this wikipedia page I believe that this is illegal in traditional markets? (If you know more please comment and I’ll edit your input into the post). But in a world of permissionless and pseudonymous systems trading “not-securities” this probably doesn’t apply.

It’s not even clear whether anybody actually manipulated the market, but the spot market did start going up after the listing of the perp all the way until trading started on Coinbase (which was when all the reasoning above expired and one would want to get out of their long position).

I’m not suggesting that any manipulation occurred here but this is a very clear warning of what’s to come.

This form of manipulation is made possible when spot markets are relatively illiquid relative to derivatives markets. If teams continue to release tokens “Compound style” with an illiquid AMM listing (Balancer has just done the same) combined with yield farming, then other teams have an incentive to release derivatives products to capture some of that attention. The tokens will list with elevated prices and some of the market will want to short.

If you are one of those people wanting to go short, heed these warnings. If teams want to protect these market participants, they should add more liquidity to the spot markets and delay listing derivatives until spot liquidity is sufficient. I wouldn’t count on either of these things though. Incentives are a hell of a drug.

Oracles

There is one final lesson from this that could be of far far greater significance: the critical importance of oracles.

In crypto, oracles bring information from the outside world onto the blockchain. When you create a synthetic asset on Ethereum that tracks the price of a real-world asset—let’s say gold—then the blockchain is constantly asking the Oracle for the price of gold and using that information to price the ERC20 version.

We are starting to see an explosion of assets and markets that rely on information from Oracles. Until now, the market has mostly understood these dynamics in abstract. We know that Oracles need to be difficult to corrupt in order for us to want to use the asset and markets that rely on them. But there weren’t significant volumes purchasing these oracle-reliant assets or trading these oracle-reliant markets.

This is changing fast and this COMP Perp vs COMP Spot offers a concrete example of how things could go wrong.

You can think of the COMP Perp as relying on the COMP Spot price as an Oracle. The difference between the COMP Perp price and the COMP Spot price determines the funding rate. And the funding rate is what keeps the COMP Perp market tethered to reality.

As we now know, if the spot market is relatively illiquid, then the cost to “corrupt” the “oracle” (COMP Spot market) goes down, which means an attacker can profit more through exposure to the COMP Perp.

Now imagine a world where DeFi has succeeded beyond our wildest expectations. There are tokens and markets for everything from US equities to local weather. All of those assets and markets that reflect the real world are using an oracle of some sort. Those oracles need to be—for all intents and purposes—impossible to corrupt in order for this new permissionless financial system to be healthy.

It’s going to be a bumpy road. Just as investors lost their shirts buying into the outlandish promises of the ICO era, so too will investors in the DeFi era lose their shirts under-appreciating the risks of illiquid markets and corruptible oracles.