A Visual Explanation of Algorithmic Stablecoins

Haseeb Qureshi
Dragonfly Research
Published in
8 min readApr 27, 2021

Stablecoin usage has exploded in the last year. And yet, fewer and fewer people seem to understand how these stablecoins actually work.

For some reason, stablecoin creators are obsessed with making these designs seem impenetrably complex. Almost every white paper is mired in equations and newly invented jargon, as though their authors are trying to convince you: trust me, you’re not smart enough to understand this.

I don’t agree. At bottom, all stablecoin design is pretty simple. I’m going to show you a simple visual language to understand how all stablecoins work.

Think of each stablecoin protocol as a bank. They each hold assets and owe liabilities. Each of them captures value somehow and distributes that value to “equity” holders.

Consider a normal full-reserve bank.

On the left side are its real assets — the actual physical dollars it holds in reserve. On the right side are its liabilities—call them “digital dollars”—which are claims on the assets in reserve.

In a full-reserve bank, each liability is matched 1:1 with assets in reserve. If someone with a digital dollar asks for the cash back, the holder is given the physical dollar and the corresponding digital liability is destroyed. This is how Tether, USDC, and every other fiat-backed stablecoin works.

The equity of the bank belongs to shareholders — investors in the bank — and they make money from the fees the bank charges. In Tether’s case, the owners of Tether Ltd. are the shareholders, and their profits come from the Tether minting and redemption fees.

Every liability of a full-reserve bank should maintain a close peg to a dollar, since it’s always redeemable for $1 in reserve. So long as the bank maintains cheap convertibility, arbitrageurs will ensure this effortlessly maintains its peg.

So that’s a vanilla full-reserve bank. It’s an obvious model, but it will help illustrate how crypto banks are different.

Full-reserve Crypto Stablecoins

How would you create a crypto full-reserve bank whose liabilities are stable dollars?

Given crypto just reinvented money, the first thing you’d want to do is swap out the USD assets for crypto assets. But crypto is volatile, so 1:1 backing won’t work if your liabilities are in dollars. If the value of crypto goes down, the bank will be left undercollateralized.

So just do the obvious thing: put down an extra cushion of crypto to give you a buffer in case the crypto goes down.

This is basically how MakerDAO works.

Dai’s peg is currently stable.

Notice that the assets in reserve are significantly larger than the total liabilities (Dai). This keeps the whole system secure.

(I’m glossing over a bunch of details here. But for the purposes of comparing MakerDAO to other models, this is a good start.)

Now let’s look at Synthetix.

Synthetix takes a different approach: instead of holding a diversified basket of cryptoassets, Synthetix issues its sUSD stablecoin against a pile of its own SNX token. This SNX is also the “equity token” — in other words, the only asset Synthetix allows as a deposit is its own equity. Because SNX is so volatile, Synthetix demands 600% overcollateralization for each sUSD in circulation.

sUSD’s peg is currently stable.

Both MakerDAO and Synthetix are analogous to traditional full reserve banks, except they are overcollateralized because their assets are in crypto. On some level, their pegs are secure because there is some mechanism to redeem the stablecoins into their underlying assets. (In both, there’s also a system of interest rates that targets a desired price.)

However, there’s another kind of stablecoin commonly known as “algorithmic central banks.”

Algorithmic central bank stablecoins aren’t redeemable at all, and don’t have depositors in the traditional sense. This makes them less like traditional banks and more like central banks. (Central banks tend to use methods other than redeemability to keep prices stable.)

Each algorithmic central bank works in a slightly different way. To analyze an algorithmic central bank, we’ll try to understand what it does in the two important scenarios: when the stablecoin is above the peg, and when the stablecoin is below the peg.

Algorithmic Central Banks

Structurally speaking, perhaps the simplest algorithmic central bank is Fei.

Fei launched recently to much notoriety, although it almost immediately broke its peg. Here’s how Fei works in a nutshell:

FEI’s peg is currently broken.

Fei functions much like a real central bank, defending its peg directly in the market. Note that Fei is not meaningfully overcollateralized and most of its assets are in crypto. This means that in a black swan event, Fei’s assets could significantly drop below its liabilities and leave it unable to defend its peg.

While the above animation gives you the high-level intuition, Fei’s real mechanics are quite involved. Fei uses Uniswap for all of its trading activity, and employs a technique called “reweighting” to perform its actual trades. It also uses “direct incentives” (effectively a type of capital control).

But the net effect is the same: the protocol participates in the open market to push the price toward the peg.

A similar algorithmic central bank is Celo protocol, which produces a stablecoin called Celo Dollar (cUSD). Celo Dollar uses CELO as its reserve collateral (the native asset of the Celo blockchain), along with a diversified portfolio of other cryptocurrencies.

Like FEI, the Celo protocol is continually willing to buy and sell Celo Dollars in the market, using a Uniswap-style market. The Celo reserve was initialized with significant assets in reserve, and the reserve aims to always stay overcollateralized. If Celo’s assets ever dip below 200% of its liabilities, the system attempts to re-capitalize by collecting transaction fees on CELO transfers.

Thus, the main difference between Celo and Fei (besides its trading mechanics) are the assets it holds and its rules around collateralization.

Celo Dollar’s peg is currently stable.

A third stablecoin in the same family is Terra’s UST. It is collateralized by Luna, the native token of the Terra blockchain. Like FEI and Celo, the Terra protocol acts as a market maker for the stablecoin. If the stablecoin system runs out of assets, it restocks by inflating the native LUNA supply.

UST’s peg is currently stable.

FEI, Celo, and Terra do not allow redemptions. Instead, they market make their own currency in the open market (that is, they are willing to buy or sell across a spread).

On the face of it, this seems quite different from redeemability! But it’s actually a closer continuum than it seems. This is because a credible commitment to market making is economically identical to allowing mints and redemptions.

Imagine a stablecoin, collateralized by ETH. Call them STBL tokens. The protocol is always willing to market make the ETH/STBL pair. This means the protocol will be willing to sell 1 STBL for $1.01 ETH and buy 1 STBL for $0.99 ETH. If STBL is below the peg, it will keep swapping STBLs until its ETH runs out.

If STBL instead uses mints and redemptions, it might let anyone mint 1 STBL for $1.01 of ETH and redeem 1 STBL for $0.99 of ETH. If STBL is below the peg, it will keep redeeming STBLs for ETH until its ETH runs out.

This has the same net result!

In traditional central banking, being a market maker rather than allowing redemptions allows the central bank more discretion. But algorithmic market making is different, because smart contracts can make ironclad, self-enforcing commitments. In this light, market making and redeemability are two paths to accomplishing the same goal: providing liquidity and ensuring a tight peg.

We’ve now looked at “central bank” style stablecoins. But there is another type of algorithmic stablecoin that is significantly more exotic: Seigniorage Shares.

Seigniorage Shares Stablecoins

The classic “Seigniorage Shares” stablecoin is Basis Cash, based on an unlaunched predecessor called Basis. It is perhaps the quintessential algorithmic stablecoin, from which many other designs were later derived.

Here’s how Basis Cash works (this one’s long, so it’s a video instead):

Basis Cash’s peg is currently broken.

You can think of Basis Cash as working in two phases: when there are outstanding bonds, Basis Cash is in a contraction cycle. The money supply is not growing fast enough to pay off all the system debts. But if demand continues to increase, eventually all bonds will get paid off and the system will enter into an expansion cycle, wherein shareholders are once again getting rewarded with newly minted Basis Cash.

Newly minted Basis Cash are the “seigniorage”—the profit the central bank gets to make from issuing new currency.

Normal central banks keep the seigniorage on their own balance sheet for a rainy day. Basis Cash, on the other hand, pays all seigniorage to its shareholders the moment it receives any.

You can intuitively see that this makes Basis very “collateral-efficient.” Basis literally has no assets on the balance sheet! This allows it to support a very large high stablecoin supply on 0 assets. But this also makes it susceptible to “death spirals” or confidence crises, as in fact happened to Basis Cash.

It’s important to understand how Basis Cash works. Most later algorithmic stablecoins are descendants of the Basis design, including the final stablecoin we’ll examine.

Empty Set Dollar (ESD) is a fair-launched stablecoin with a pseudonymous founding team. The original version of ESD, now known as ESD v1, was based closely on the Basis Cash design.

ESD v1’s peg was broken and they have since pivoted to a new design.

ESD’s innovation was to fuse the “share” token with the “stablecoin” token. This means the stablecoin, if staked, produced more stablecoins. As you might guess, this resulted in the stablecoin becoming highly volatile and drifting away from the peg, sometimes as high $2.00, until it finally collapsed to below $0.20.

So far, pure Seigniorage Shares coins have universally failed. The many Basis and ESD knock-offs like DSD all met the same fate. This tells us, at the very least, that stablecoin design really matters. These illustrations should help you reason through why Seigniorage Shares is so vulnerable to confidence crises.

In the early days of DeFi, many believed that decentralized stablecoins were fundamentally impossible. For now, it seems that these claims were premature. There is a large design space, and some designs are genuinely more robust than others.

But one thing is for sure: you shouldn’t assume a decentralized stablecoin will be robust simply because a white paper insists it is. Think for yourself what it takes for that stablecoin to be stable. (And if you’re confused, try drawing a picture. At least for me, it helps.)

Disclosure: Dragonfly holds positions in many of the assets discussed in this piece.



Haseeb Qureshi
Dragonfly Research

Investor at Dragonfly Capital. Formerly Metastable, @Airbnb, @earndotcom. Writer. Effective Altruist. Former poker pro. One always finds one's burden again.