The Perpetual PvP Ponzi

So a couple weeks back, I got into a debate with wireless anon on perp vAMMs (particularly Perp v1 and Drift).

Midway through the endless twitter thread, I realized neither wireless nor I shared any data to back our claims. And then the next day, this happened….

This isn’t the first time tuba nerd snipped me into researching a mechanism, see the thread Vertex and I wrote on UNI v3 vs Sushi’s Trident.

But I couldn’t resist a good nerd snipe, so 0xDosa and I went digging.

Why are we writing this article?

In our research, we studied the incentives + historical data of vAMMs and came to the following conclusions:

1. Most volume comes from arb.
2. Arb Traders are playing a PvP Ponzi.
3. Funding rate sign (+/-) is determined by the starting pool price
4. Funding magnitude is bounded by trading fee or fee pool size.
5. Protocol’s cannot withdraw profits without blowup risk.
6. At some point, this PvP game could blow up the Insurance Fund.

Over this article, we will support these claims with data.

Well what are vAMMs?

vAMMs (virtual AMMs) are UNI v2 pools (xy=k) without any LPs. All liquidity is virtual and (often) hardcoded by the protocol.

Basically the protocol says: “We have a UNI v2 pool with 100 vETH and 200,000 vUSDC” (w/o holding any ETH or USDC). And that is all it takes to make an ETH vAMM perp.

Who is trading if there are no LPs?

On Perp v1, we broke down top 20 accounts by volume and open interest:

Source: Perp v1 Dune Dashboard

The top 20 accounts by volume and open interest account for between 85–95% of volume and open interest. We suspect the same to be true for Drift. On Drift, 300–800 DAUs generate between $50-$100M in daily trading volume (Drift Dune Dashboard).

The data supports the claim that most trades come from a handful of accounts. We suspect these accounts to be running two arbitrage strategies:

1. High Volume = Price Arbitrage
2. High Open Interest = Funding Rate Arbitrage

Are price arbitrageurs playing a PvP ponzi?

We know that 20 accounts dominate Perp v1 and that there are no LPs. Essentially, these 20 accounts are trading against each other. If all 20 accounts closed their position, then the price would revert back to the terminal price (starting price) of the vAMM.

This is because vAMMs are path independent. For example:

• Initial Price = $2,000
• Alice Opens 1 ETH Long. Price = $2,500.
• Bob Opens 1 ETH Long. Price = $3,000.
• Alice Closes 1 ETH Long. Price = $2,500.
• Bob Closes 1 ETH Long. Price = $2,000.

After closing all positions, the price returns to its starting price. Alice made a profit and Bob made a loss. But their PnL didn’t depend on the true price of the asset, only on the price of the vAMM.

In short, yes all arbitrageurs are playing PvP ponzis. Nobody wants to be the last one holding the bag.

If they are playing PvP, how does price arb work?

At any given point in time, the price arbitrageur is always doing one of two things: (1) opening positions or (2) unwinding positions. Arbitrageurs open positions when they spot a price difference (see below).

But because vAMMs are path independent and there are no LPs, arbitrageurs must strategically unwind their position (against other arbitrageurs). For example, if an arbitrageur controls 20% of Open Interest (OI), then for them to unwind another arbitrageur must take over their position. This means:

1. An arbitrageur can only unwind if there are other arbitrageurs.
2. No arbitrageur wants to be the last one in the pool.

Okay I get the PvP, but ponzis have high APYs?

Every good ponzi needs a high APY, and in vAMM perps that APY comes from funding rates.

The first thing to note about funding rates is that they are largely paid out by the protocol. In that sense, earning funding is another PvP game (Trader vs Protocol).

To understand funding rate arb better, we must answer:

• How are funding rates determined (sign and magnitude)?
• How does the protocol pay funding?
• Does the protocol make a profit (after funding)?

How is funding sign and magnitude determined?

Because vAMMs are path independent:

• If price > terminal (start) price. Traders are net long.
• If price < terminal (start) price. Traders are net short.

Net Position as Price moves from Terminal Price

These traders play a major role in determining (1) the sign and (2) the magnitude of funding.

Sign (Positive or Negative Rates)

Net traders control the sign of the rates in the following way:

• If traders are net long, then rate < 0 || rate < CEX Funding Rate
• If traders are net short, then rate > 0 || rate > CEX Funding Rate

The easiest way to understand why this is true is to look at the counterfactual.

• If traders are net long and rate > 0 || rate > CEX Funding Rate.
• Then traders would be paying the more expensive rate.
• Traders would close their positions and push prices down.
• Traders will do this till mark <= index and rates are in their favor again.

Drift Funding Rates on SOL and LUNA

You can see in the above diagram that Drift rates for SOL and LUNA are either negative or positive based on whether terminal price is greater or less than current price.

Magnitude of Rates

Because most volume comes from arbitrage, vAMM prices tend to lag the index price by at least 0.1% (i.e. the cost of a trade). This lower bounds funding rate:

• In Perp a 0.1% lag implies a lower bound of 36.5% APY (0.1%*365 days)
• In Drift, the lower bounds is MIN(36.5%, Fee Pool)

But in practice we notice rates can go much higher. Why so?

This happens when price >> terminal price || price << terminal price. When there is a large difference between terminal price and a few traders control the open interest, they can manipulate/amplify rates as they unwind their positions (pushing mark away from index).

How does the protocol pay funding?

The protocol pays out funding from the revenue it generates via trading fees.

Date: March 21, 2021

From the above we can see that both protocols make more revenue from trading fees than they pay out in funding.

Does this mean the protocols are profitable? Well, not exactly…

Are the protocols profitable?

Short answer: no.
Long answer: both protocols have profitable insurance funds, but they cannot withdraw these profits without putting traders at risk. Let’s look at the reasons why for each protocol.

Perp v1

In Perp, for pools where price >> terminal price, the protocol is paying more in funding than they are earning in trading fees. The best examples of this are ETH (Terminal Price = $500) and BTC (Terminal Price = $19,000). You can notice in the dune bar graph for ETH and BTC, the total funding paid (gray bar) > fees earned (orange bar).

As prices deviate from terminal price, the Insurance Fund (IF) would pay more in funding than it earns in fees. This would lead to the IF blowing up.

Drift

In Drift, the same insurance fund depletion could occur but there are some design subtleties to consider first:

• Funding is capped by the fee pool (Fee Pool = 0.5 * Trading Fees)
• IF pays for repegs and k adjustment when there is mark-oracle divergence

So if the fee pool is capped how can the IF blowup? It can happen in a 3-step process:

Step 1: As price diverges from terminal price, the fee pool gets depleted. We can see this in ETH, BNB, and BTC pools on Drift. Notice how the the fee pool remaining decreases as price change increases below (see data here):

March 21st Drift Data

Step 2: As the fee pool gets depleted, volume decreases and potential of oracle-mark divergence increases. Notice how 24hr volume decreases as fee pool decreases below (see data here):

March 21st Drift Data

Step 3: The reduced funding increases the likelihood that arbitrageurs leave (like when a ponzi reduces APY). When arb’ers leave it creates a deviation b/w mark and oracle price. This deviation is corrected by an expensive repeg, which the insurance fund must pay for.

Currently, BTC has 250k in its insurance fund. A 1% repeg costs $26k and 10% costs $260k (calculated using this script). A 10% repeg could deplete the entire insurance fund for BTC on Drift.

Repeg Cost of BTC on Drift

Conclusion

Kudos to Perp and Drift for designing one of the coolest PvP games in our industry. Ultimately, this is a great game for arbitrageurs if they play their cards right. You can earn high funding rates and price arb — you just don’t want to be the only arbitrageur left playing the game.

But these games do face blowup risk and they are not exactly sustainable or profitable for their operators. In both cases, it isn’t safe for the protocol to withdraw profits from the insurance fund due to the blowup risk.

Despite this, I see no reason why the PvP Perpetual Ponzi can’t continue going indefinitely with a token backstopping the insurance fund.

Credits

This post was authored/researched by Noodles and 0xDosa

Peer reviewed by: Tuba, Brian Pendleton, Nate, Jib, and Rebecca