Gyroscope is Different, Part 2: Algorithmic Stablecoins
Gyroscope has the best design and technology today for decentralized stability. Here’s how we reinvent the stablecoin.
At Gyroscope, we have a unique background on DeFi, having ourselves done the cutting edge research on every approach. We have designed Gyroscope for DeFi power users who want the best decentralized stability. In this sequence of posts, we illustrate how the Gyroscope design choices shape up against the wider space of stablecoins.
This is Part 2 on algorithmic stablecoins. Read Part 1 on meta-stablecoins and join us for Part 3 on custodial and leverage-based stablecoins coming up next.
Algorithmic stablecoins aim to maintain a stable price by automatically adapting the stablecoin supply to meet demand. There are three main types we’ll cover here: the Basis and reserve-backed models, as well as briefly touching on the seigniorage shares model (which will be further covered in Part 3).
Basis-type Stablecoins Don’t Work
In the Basis model, the stablecoin is uncollateralized. If the peg is functioning, some new user pays ~$1 for each new stablecoin that is minted. The protocol rewards all of these proceeds to stakeholders, who perform little function. In a crisis, stability relies on speculators who continue to bet on continual growth in the money supply, which must reach above-and-beyond the pre-crisis money supply level for the bets to be profitable. Speculators naturally abandon this bet when the required growth becomes uncredible, leading to the collapse of the stablecoin.
Many Basis-type stablecoins have been deployed in the past few months, including Basis Cash and Empty Set Dollar. All have suffered the same fate, which we predicted in our 2018 critique of the original Basecoin, and NuBits before that.
Reserve-backed Stablecoins and Bank Run Risk
In the reserve-backed model, the protocol builds up an asset reserve, whose value backs the stablecoin. When a new stablecoin is sold, all or a portion of the proceeds go to fill this reserve. Users can then redeem stablecoins for assets from the reserve. This is able to quell crises up until the reserve value is exhausted.
In the case that the reserve dollar-value is less than the number of stablecoins, the stablecoin faces bank run risk. In particular, if the system breaches an outflow threshold with stablecoin holders redeeming from the reserve, the health of the reserve will decrease, triggering remaining holders to try to redeem before it is too late. As a result, the reserve will be exhausted and some stablecoin holders will be left with empty bags. This can occur naturally or via a Soros attack (akin to George Soros’s famous attack on the British pound), in which speculators try to cause this to happen and profit from the currency collapse.
A few critical design questions arise here. What portion of new issuance proceeds goes to the reserve? The greater the portion, the stronger the reserve. Is the reserve designed to be all-weather, i.e., does the reserve contain assets that will retain value in a variety of crises? An all-weather reserve diversifies all risks, not just price risk, to maximize the likelihood that the reserve is valuable when it is needed in a crisis. Along these lines, the reserve ideally shouldn’t be composed of the system’s governance token, which can quickly lose value if the system is in crisis. And how does the redemption process work? The redemption process will factor deeply into stablecoin liquidity, reserve health, bank run risk, and whether people will even want to buy in to the stablecoin.
The NuBits stablecoin, dating back to 2014, was essentially a mix of the Basis and reserve-backed designs. It contained a small reserve in BTC, which was quickly exhausted and unable to quell crises. Then after reverting solely to the Basis-type mechanism, NuBits collapsed in 2018.
In the Frax stablecoin, a fraction of the value generated by selling newly minted stablecoins is stored in a USDC reserve. The remaining fraction is paid in Frax shares, the market cap of which also supports the stablecoin in a manner akin to seigniorage shares, which we will discuss next. In this sense, Frax is equivalent to owning some of USDC and some of a seigniorage shares stablecoin.
Brief Overview of Seigniorage Shares
In the seigniorage shares model, a new asset is created for the purpose of being a risky asset that backs the stablecoin. In the algorithmic version, the stablecoin supply is automatically adapted to match demand, with risk intended to be absorbed by the collateral asset holders. The seigniorage shares mechanism is similar to the leverage-based mechanism (e.g., Dai) we will cover further in Part 3 of this series. A primary difference is that the risky asset is much more risky because its value comes from the value of the same system in a self-fulfilling way. If the system is in crisis, the value of the risky asset itself can enter a death spiral, causing the system to collapse. This can also be precipitated like in a bank run or Soros attack: large redemption of stablecoins causes the risky asset price to deteriorate, which triggers more redemptions and the collapse of the system.
Seigniorage shares stablecoins have a turbulent history. Steem Dollars is an early example of a seigniorage shares coin on the Steem blockchain. Steem Dollars is stabilized on the downside when prices are below $1 (but not above $1) and has undergone many depegging crises.
Other examples of seigniorage shares, like Terra, have worked OK to date. However, they face the same fundamental risks. We will cover more of these in Part 3 of this series.
Lastly, note that Ampleforth is not a a seigniorage shares model and is not a stablecoin. Essentially it is only a UI trick on a fixed supply cryptocurrency. It transforms some price volatility into some volatility in the number of coins every user holds, but has no stabilization effect on the value that users hold, barring psychological effects. This has also proven true in practice.
Returning to Reserve-backed Stablecoins
Returning to Frax, the seigniorage shares part is fragile to bank runs, and the USDC reserve is not all-weather as it hedges only price risk and is not resistant to regulatory, censorship, and counterparty risks.
Risks in the Fei stablecoin are more extreme. The Fei reserve is composed of only highly volatile ETH and, by intent, starts under-reserved. Fei starts with a small fractional value in reserve for each stablecoin issued, giving the remainder to first stakeholders, which makes it inherently more fragile than it could be. Altogether, this is nearly the opposite of what we would describe as an all-weather portfolio. It results in Fei holders taking an inherently large bet on volatile ETH prices not declining, which amplifies bank run risks.
Fei will have high risk of liquidity crises, essentially by design. Its resistance to bank runs is questionable only in that it is highly costly to redeem from the Fei reserve. By design, redeeming incurs quadratic fees. This leads to a host of issues. The Fei system will have guaranteed low liquidity during any supply contractions. Fei holders will be essentially entrapped if secondary market liquidity is dry, even if the reserve is full. This begs the question of why users would want to participate (does anyone participate in equilibrium, after farming incentives subside?). Further, it doesn’t prevent the reserve from being depleted, especially when ETH prices fall substantially, which happens fairly regularly.
There are several other more subtle issues with the Fei design. For instance, the “direct incentives” (quadratic trading fee mechanism) won’t necessarily result in a net Fei burn for subtle AMM reasons; there will naturally be extra Fei minted simply due to ETH volatility. And the mechanism is also not incentive compatible as exchange volume may naturally migrate to separate markets without quadratic fees when Fei trades under peg, fracturing the already limited liquidity.
Celo Dollars is a stronger reserve-backed design that addresses some of these points. Celo Dollars aims for 100% reserves and effectively starts much higher, making it safer. A small part of reserves provides liquidity to a constant product market. In case reserves fall below 100%, Celo addresses bank run risks with an ad hoc governance process of periodically rebalancing the constant product market with custom amounts of the reserve, adapting liquidity on-the-fly. Further, the reserve receives a portion of new Celo block rewards.
The Celo reserve is mostly composed of Celo assets, which are realistically more risky today than ETH and effectively makes the reserve design closer to seigniorage shares, as the Celo system is not as widely adopted at this point. Celo addresses this by diversifying the reserve with BTC, ETH, and Dai, although these assets rely on bridges to the Celo blockchain, which introduces other vulnerabilities.
There are several problems that remain outstanding in the above designs. We’ve designed new mechanisms for Gyroscope to solve these.
- All-weather reserve. The Gyroscope reserve portfolio is designed to be all-weather by (a) diversifying, to the extent possible using cryptoassets, all DeFi risks and not just price risk, and (b) deploying assets to earn yield in DeFi while stratifying composability risk (introduced in Part 1) among different protocols. For instance, the reserve is composed of vaults that separate different types of risk, including different stablecoin, oracle, protocol governance, censorship, regulatory, and ETH price risks. The portfolio is designed so that the failure of individual vaults does not cascade into other vaults. Read more about how this works here.
- Liquidity and resistance to bank run risk. While the Gyroscope reserve aims for 100% reserves, with all revenue from new Gyro Dollar minting going to the reserve, the system still needs to control bank run risk in the event that the reserve value is shocked. Should this happen, the Gyroscope system will remain stabilized at $1 through the peg coordination mechanism (described fully here). Peg coordination is maintained through the specially designed primary-market AMM (P-AMM). Taking into account the state of the reserve, the P-AMM provides controlled liquidity in the primary market while balancing incentives against large outflows that lead to bank runs. Gyro Dollar liquidity is boosted by the additionally optimized secondary-market AMMs (S-AMM). Read more about the the P-AMM and S-AMM designs here.
- Adapting DeFi protocols securely has to be a slow process. The process of adapting protocols has to go through governance. An important step in securing this process from abuse is lengthy time delays. One of the contributions in Gyroscope is limiting the governance surface area around short-term adaptations. For instance, the P-AMM automates what would otherwise be fast ad hoc governance choices about how to handle reserve redemptions in a crisis (e.g., comparing to Celo Dollars). This means that the primary tasks of governors can be limited to tasks that can be approved slowly, and so more securely, like adapting the shape of the vaults and all-weather reserve as DeFi evolves and making longer-term tweaks to the system.
- Controlling governance extractable value (GEV). We next revisit GEV in the context of algorithmic stablecoins, and how new mechanisms introduced in Gyroscope solve the GEV problem.
Governance Extractable Value
We discussed governance extractable value (GEV) in Part 1. GEV describes when protocol governors are able to derive better payoff from actions that are counter to protocol health. For instance, by introducing questionable collateral types, seeking short-term profits at the expense of long-term health, or even outright malicious actions to steal user funds. GEV applies similarly algorithmic stablecoins. The heart of the issue is that (a) a governance time delay is only effective at lowering GEV if the protocol provides exit liquidity to all participants in the delay window, and (b) users who aren’t tracking very technical smart contract details remain unprotected.
In terms of algorithmic stablecoins, Fei is particularly worrisome in terms of GEV because it, by design, has almost no exit liquidity and governance is designed to be extremely flexible in upgrading the contracts. Because of this lack of exit liquidity, Fei holders are essentially entrapped by governance and can only redeem at increasingly bad prices, even if the reserve is full. Lastly, there is no current tokenomics design to Fei governance tokens, i.e., there is no proposal for how “honest” cashflows would accrue to governors. This means that, straight out of the gate, the only value in Fei governance comes from GEV.
Frax is a bit better than Fei in this regard because it provides the required exit liquidity and further limits governance ability to upgrade contracts. Similarly to many other DeFi systems, however, the level of GEV in these systems, and so the size of the problem, remains an open question. Further, GEV is regrettably not a focus of most existing protocols.
Gyroscope changes that. Gyroscope introduces pioneering new governance mechanisms that serve as checks and balances against GEV. Gyroscope uses Conditional Cashflows, which only accrue to governors in the long-term if the protocol remains healthy. And Gyroscope puts power into the hands of Gyro Dollar holders with Optimistic Approval, an option to veto during the time delay. The combination of the time delay and Optimistic Approval is quite powerful. Essentially, if bad governors commit malicious actions, Gyro Dollar holders have an opportunity to organize a YAM-like rush to exercise the veto. The process remains “optimistic” in that Gyro Dollar holders don’t ordinarily need to do anything directly in governance.
Stay tuned for Part 3 where we will compare with custodial and leverage-based stablecoins.
What to do next: become a Gyronaut!
- Follow us on Twitter @GyroStable
- Join our Discord!
- Try out the Gyroscope testnet! And watch for coming announcements
- To dive deeper on algorithmic stablecoins, read our Stablecoins 2.0 paper (see “endogenous collateral”, “implicit collateral”, and “algorithmic issuance” for further discussion)
Thanks to the participants of the Crypto AMA with Spencer Noon for great questions and discussion and many other DeFi enthusiasts and researchers for numerous discussions on algorithmic stablecoins and GEV.
*Image credit: NASA’s Goddard Space Flight Center/Jeremy Schnittman
