Why did Terra Collapse? Confessions of A Market Maker

Posted by:
Jerry Celegans
May 13, 2022

Why did Terra collapse and UST depeg? Could it have been totally avoided or at least prevented? In this article, we answer those questions, and look at how Terra could have done better to prevent collapse.

A Market Maker’s Take on the Collapse of Terra

We answer the above question from my point of view as a professional market maker. I've been working on Wall Street for over a decade, and make markets for both equity and crypto markets. With my deep understanding of market microstructure, I can reveal the secrets of the Terra collapse.

TL;DR: Terra could have most effectively prevented the collapse by announcing publicly a standing massive bid for LUNA. They did this crazy shit instead(read on for more!)

Collapse could never be 100% avoided: it was a reflexive asset

First, please understand that the way Terra was structured as of this month, collapse could have never been totally avoided. The Terra ecosystem (LUNA/UST) is a reflexive asset -- if everyone believes it's worthless, it becomes worthless. This is not true for most other stablecoins: for the Maker system, even if you believe MKR/DAI is worthless, your DAI is still overcollateralized for ETH.  USDC is also overcollateralized by external assets.

Now Terra was going in the right direction by buying Bitcoin and other cryptos to back UST. If Terra's external (non-LUNA) assets exceeded its UST debt, then only in that case could collapse be 100% avoided.

By the time Terra started buying BTC, they already had about $18 bln of UST debt, and only $1.5 bln of BTC assets -- it was too little, too late.

If collapse couldn't have been avoided, could it have been prevented?

After all, Terra worked being a reflexive asset without enough external collateral for many months. Could Terra have much reduced the risk of a collapse?  As a professional market maker, I would say the answer is a resounding yes -- Terra could have done more to prevent the collapse.

The system should have been able to handle stress many times the amount that caused it to fail. The rest of this article will explore why.

How LUNA / UST Works: A Quick Review

To understand the rest of the article, we'll have a quick refresher to the main moving parts of the Terra ecosystem. The ecosystem consists of two tokens, LUNA and UST.  Both are not in fixed supply. UST is intended to be pegged to the dollar, where LUNA is a speculative coin. The core mechanism is:

  1. You can burn 1 UST to mint $1 worth of LUNA
  2. You can burn $1 worth of LUNA coins to mint 1 UST

If UST sells for less than a dollar, I can buy one, convert it to $1 worth of LUNA, and sell that LUNA for a profit. If UST buys for more than a dollar, I can buy $1 worth of LUNA, convert it to 1 UST, then sell that UST for a profit.

Action 1. above boosts UST value if it is worth less than $1. Action 2. above decreases UST value if its worth more than $1. These two actions peg UST to $1. Hence, algorithmic stablecoin.

Note that for these mechanisms to work, LUNA must have nonzero value. This assumption is nontrivial; most crypto projects fail at this step: they don't have a token with substantial value to begin with. But before the crash, LUNA was worth $32 bln so they were clearly making the system work.

Primary and Secondary Market Makers are Key to the Peg

To aid with stability, Terraform Labs had contracted a number of primary market makers, including Jump Crypto. The term “market maker” sounds fancy, but in reality is just a participant, like me, who makes money providing liquidity. In this case, it is an entity that executes the burn and mint mechanism. Primary market makers enforce the UST peg conditional on positive LUNA prices, by explicitly and directly burning UST.

Faith in these primary market makers allows other secondary market makers to provide liquidity as well. Secondary market makers may not know or care about the burn and mint mechanism, but will put out a bid at 99 cents, trusting that Jump Crypto and co. will prevent larger dislocations. In fact, the mere existence of Jump Crypto and co. can lead to secondary market makers providing most of the liquidity. It’s a system built on faith but backed by arbitrage.

How the UST Peg Breaks: Step by Step

Knowing the above, we have all the tools for me to illustrate to you how Terra broke this week. Some people think the straw that broke the camel's back was Terraform Labs moving $85 mm of UST to Curve's 3pool. Some people point to the $2 bln UST unstake from Anchor. Others accuse Citadel and Blackrock of orchestrating the crash. While I have no opinion on the actors, I know from the markets this is how the unravelling happened mechanistically.

First, UST deviates only a little bit, say 99 cents. LUNA is at $100. Jump wants to arbitrage UST with LUNA.

Step 1: Low LUNA Slippage Means Closer UST Prices

Jump buys one UST for 99 cents then burns it to mint $1 worth of LUNA, which is 0.01 LUNA. When it goes to sell the LUNA to lock in 1 cent of profit, it realizes that the bid-ask spread is larger than expected. The highest bid is $99.8 instead of $100. So it incurs some slippage and makes 99.8 - 99 = 0.8 cents instead of 1 cent of profit. So far so good -- with tiny LUNA slippage, arbitrage profits are maintained. Some LUNA bidders notice that an arbitrage is going on, but they don’t care because the traded volume is so small.

Step 2: Medium LUNA Slippage Leads to the Tipping Point

Now there is a massive sell order for UST that blows through bids, putting UST at 90 cents. The UST secondary market makers are surprised, but they don’t back out yet, believing either that Jump will save them or that there’s no more selling coming.

The LUNA bidders, anticipating arbitrage selling, start pulling their bids. Not only will there be selling pressure, but new LUNA coins are minted, diluting the value of each token. LUNA liquidity is thinning out. The initial arbitrage doesn't look as juicy anymore. Jump executes the arbitrage but suffers a slippage of 3 cents, making only 7 cents per UST burned instead of 10. With medium LUNA slippage, arbitrage profits are teetering and precarious.

Meanwhile, there’s still a massive sell overhang on UST at 90 cents.

The secondary market makers holding UST are now afraid of more whales selling UST. After all, as secondary market makers, they're not equipped to directly arb this 90 cent UST against LUNA. They do what I've seen in markets a thousand times: they puke. What this means is that these market makers were getting long UST as the price dropped, and at this point they get scared enough that they reverse course and dump their long positions at a loss.

This adds fuel to the downward spiral of UST. At the same time, LUNA bidders realize they aren’t bidding low enough due to all the sales of LUNA.

The expectation of more selling begets more selling.

So LUNA liquidity thins out even more. As the price of LUNA falls, slippage in LUNA increases, UST arbitrage becomes less profitable, and every UST burned mints more and more LUNA, accelerating inflation. The spiral of doom has begun.

At some point, the slippage increases to 15 cents. The arbitrage now only makes sense at UST lower than 85 cents.

With secondary market makers gone, UST liquidity is also very thin. Previously calm UST holders now try to rush out the door, fearing the worst.

Step 3: Collapse

UST drops to 75 cents.

LUNA bidders pull more liquidity. Hyperinflation. Arbitrage slippage rises.

The most fervent LUNAtics now start dumping UST.

Just like the hierarchy of market makers, there is also a hierarchy of LUNA bidders. Short term speculators, hedgers, long term investors, etc. But there is no backstop for LUNA. LUNA is a pure speculative asset with no hard reserves. So they all run for the exits. Arbitrage slippage rises again.

But what backs the value of LUNA the coin?

It is an error to think that as long as the market cap of LUNA was greater than that of UST, the peg would be easily re-established. This viewpoint is wrong because market cap alone has little relevance to how much it can be impacted.

Look at any company’s surprise earnings as an example. At market close (4pm US Eastern time) on April 19, Netflix (NFLX) was trading at roughly $330. Upon a terrible earnings release, it ended up trading at $259 by 8pm, a 21.5% drop. That’s close to $34B in market cap eliminated. Do you think it took $34B worth of trades to push the price that low? Nope. Between 4pm and 8pm, NFLX traded less than $3.0B, which is less than 10% of the market cap loss. In fact, during the immediate 16% drop, only $63M traded, which is less than 1% of the market cap loss.

One could respond, ‘Well, that’s because it was after hours and there was no liquidity.’ Sure, anyone with $63M could have probably pushed the price down 16% after hours. But speculators would have bought that dip immediately if earnings were actually good. NFLX dropped further the next day.

The point is that how much LUNA moves has very little to do with how much LUNA is worth. If the market thinks LUNA is worth nothing, bids will immediately move down with no trades needed.

The depegging of UST was essentially a super bad earnings report for LUNA, made worse by an inflationary mechanism designed to backstop UST at the expense of LUNA holders.

In my view, Terraform Labs should have swooped in to backstop LUNA, not UST. Terraform Labs had a $1.5 bln BTC reserve that it amassed for precisely this kind of depegging, but it seems they used it to buy UST instead.

By announcing a stabilizing bid at $50 LUNA for example, the entire hierarchy of LUNA bidders would have stayed or even grown. Even without any tangible reserves (since we’re in crypto after all), by simply announcing buy backs, it could have emboldened an entire army of hodlers to defend LUNA.

Moreover, UST already had a backstop: the arbitrage mechanism. The mechanism works if and only if LUNA has value. Backstopping the value of LUNA implies the backstopping of UST. Without a stable LUNA, the arbitrage bound for UST inevitably falls farther and farther away from $1 in a death spiral as described above.

During the collapse, there were many times at which UST shot up, only to eventually tumble back down. Whether it was Terraform Labs buying UST with its BTC reserves or not, it did not work.

Terraform Labs should have announced a swift and overwhelming LUNA backstop to early signs of a UST depeg instead of poo-pooing it like Do Kwon did.

Doing so would have maintained the peg and restored enough confidence to prevent a stampede for the exits, perhaps without even dipping too much into its reserves. Terraform Labs would have been able to survive a bit longer, building up more reserves to withstand an even bigger stress down the line.

Whether this collapse was due to a bad actor is largely irrelevant. The system did not incentivize the correct behaviors under stress, and under this stress, all it took was a loss of confidence to kick off a death spiral.

--Did you enjoy this article?--

If you enjoyed this article and want to support a good cause, we're supporting Cryonics research this month with the Cryonauts NFT.