Quick Take (sorry couldn't help myself): Matteo Leibowitz’s article captures a number of issues with the current Synthetix design. Most of these issues have solutions that are being reviewed by the team and the community. We recently implemented a governance framework for change requests modelled on Ethereum's EIPs (Ethereum Improvement Proposal). We expect that most of the changes required to address the current design concerns will be implemented in the next few months. In addition to these genuine issues, the article also contains a few misconceptions about Synthetix, which is a reflection on the lack of up-to-date project documentation available. Synthetix is still a very early stage and experimental project and we are actively working on improving both the mechanism and the documentation.
Before responding to each point in the article, I want to say a huge thanks to Matteo for not only taking the time to review the system, but also actually testing it. It is hard to get a feel for a dApp without using it, so it speaks well of him that he took the time to do real-world testing. The walkthrough of the minting process is better than our current example on our site, so I would honestly like to steal it!
The second point worth noting is that while I spoke to Matteo briefly to help get some test SNX on Kovan we didn’t discuss the contents of his article before it was published. This of course meant he was relying on the poor documentation of the current system. This was actually invaluable to us as it has highlighted a number of misconceptions we need to address in our documentation.
The approach I have taken is to quote sections from the article and address them directly, it is probably worth reviewing the original article here in its entirety before reading my response, as it may come across disjointed otherwise.
So let’s get into it!
“Synthetic asset issuers solely looking to collect fees rather than going net long/short can hedge their exposure into a market neutral position. For example, if Alice mints sUSD and buys ETH, she is short sUSD – she wants the value of sUSD to decline, or, put differently, the price of ETH to appreciate, so that she can pay back her debt at a discount. However, she can neutralize this short exposure by shorting ETH, either on the Synthetix exchange itself through the inverse ETH product (iETH) or by borrowing ETH and selling it for USD via a secondary lending platform like Compound or dYdX.”
This is a critical point, and one I expected to be more controversial, the risk profile for participating in debt issuance using SNX is high. It requires an understanding of the underlying mechanisms and the need to not only stay collateralised but to have a strategy for hedging currency risk internally and/or externally to the system.
“This reflexive relationship flow is bi-directional. If trading volume on Synthetix drops then so should the value of SNX. In an efficient market, this leads to a cascading effect, with the collateralization ratio of outstanding assets in the Synthetix ecosystem falling.”
The system is bi-directional — scalability works both ways. But in an efficient market it’s not necessarily cascading, due to the incentive for SNX stakers to adjust their C-Ratios in alignment with a declining SNX price. In an efficient market if trading volume drops, then the SNX price drops, which means SNX stakers’ collateralisation ratio decreases, which means they’re incentivised to burn Synths to bring their collateralisation ratio back up to 500%, which brings the Synth supply back into equilibrium with demand. Matteo's point on a cascading effect appears to be predicated on the idea that if the Synth supply drops, this will result in a commensurate drop in volume on Synthetix.Exchange. It’s possible there could be a small correlation, but we are confident there is less correlation than the article implies.
“Synthetix chooses to employ a ‘carrot’ incentive system rather than the hybrid ‘carrot’ and ‘stick’ model used by competing synthetic asset platforms like MakerDAO and UMA. There is no liquidation process and/or penalty when collateral falls below a certain threshold – instead, synthetic issuers are supposed to be motivated to maintain their margin requirements through the cost of missing out on platform trading fees.”
This is of course a critical point, and we have a plan to implement a liquidation process in the future. But it is also worth pointing that like Maker the Synthetix system absorbed a 90% decline in the SNX price during 2018 and the system was never undercollateralized at any point due to SNX minters responding to the existing incentives. That said it is critical that a robust mechanism for liquidations is implemented. The current plan is to allow anyone holding synthetic assets to use them to unlock the collateral of positions that have fallen below a certain threshold. Part of the reason this has not been prioritised has been that we have observed the 500% collateralisation ratio has been a sufficient buffer, and we are currently planning to raise this further, to 750%.
“Because exchange fees are paid out on a weekly basis, can be claimed up to every six weeks, and incur no penalties under the condition that the collateralization ratio is maintained at the time of claiming fees, the system is incentive incompatible. The system could viably be severely undercollateralized for weeks at a time, with issuers simply topping up their collateral on a periodic basis.”
This is a very astute point, and there has been debate in the community to reduce the claim window to two weeks, which through our new community governance system is expected to be implemented in the next few weeks. Of course this would not resolve the underlying criticism. But it is expected that when liquidations are enabled that they will be instantaneous as soon as the liquidation threshold has been crossed. However, to describe the potential lag as ‘incentive incompatible’ is inaccurate, and it would be more accurate to describe it as a ‘lagging incentive.’
“Using SNX as collateral is a suboptimal design choice, although the Synthetix team did recently announce that they will be allowing SNX holders to stake up to 200% of their position in ETH.”
The criticism of SNX as collateral is where we disagree most with the analysis. Of course SNX is a worse form of collateral than ETH currently. But the reasoning behind using SNX as collateral is to ensure there is a direct relationship between utilisation of the network and the value of the collateral. An objection we typically see about SNX as collateral arises from the assumption that most cryptoassets have no underlying value, which leads to high reflexivity. While that is the case for SNX while the system is nascent, over time, provided exchange fees are sufficient to underpin the network value of SNX this will decrease. In the early phases the value of SNX is based on the potential future utilisation of the network, this is no different than any other startup or network. The reason we have users staking SNX and putting collateral into the network is that they can see the early traction and the utility of the network. Once the fee yield to stakers comprises a meaningful portion of the SNX value it will be usage metrics and yield that determine the price of SNX and therefore the carrying capacity of the network in terms of debt.
“A 500% collateralization ratio comes with significant opportunity cost in the context of high lending rates being offered across lending protocols as well as inventory risk: it remains questionable as to whether the value of exchange fees can outperform Dai’s risk-free rate, which currently sits at 7.86%.”
While this is true, there is strong evidence that stakers in high growth crypto networks discount the opportunity costs of staking and the cost of capital. There are numerous reasons for this including several powerful heuristics like consistency bias, loss aversion and sunk cost that while “irrational” from an economic perspective are powerful drivers of human behaviour. But the main reason a purely rational assessment of opportunity cost is not valid is that the potential network growth if the system must be considered. This is the reason using a dedicated asset, rather than Ether, is powerful in the early stages of the network, because if you are choosing between staking Ether in a synthetic asset platform versus loaning it at 7% with lower risk, the opportunity cost is clear. But if you are holding an asset that can only support one network you must make a decision to either to continue to participate or remove your capital entirely, this changes the decision because if you believe the network can scale then you will likely discount the cost of capital to continue to participate.
“The third shortcoming in using SNX as collateral is that the value of synthetics issued and the system’s ability to scale is inherently inhibited by SNX market cap. With a ~$30 million network value, the system can currently handle a maximum $6 million of synthetic issuance. As discussed, there is a reflexive relationship between SNX’s fundamental value and synthetic trading volume: in order for SNX to be valuable and continue to accrue value there has to be a continuously growing trading volume. This is difficult to achieve in such a capital inefficient system.”
This is actually a feature not a bug, as it ensures that the total carrying capacity of the network is constrained while the network is scaling. Something else which is often overlooked about the exchange is that there is no network effect around utility. A single user can present $100 to the exchange or $10,000 and the lack of counterparties means both will have an equally positive experience. Trades are only constrained by the size of the debt each user holds. This creates powerful alignment between the value of the network and the usage.
“Maintaining resilient price oracles is a must for any synthetic issuance platform – the system must be permanently aware of the value of both the issued assets and the supporting collateral. At present the Synthetix team controls all oracles, with a plan to migrate to a Chainlink product in the near future. The SNX price oracle seemingly does not account for liquidity, instead taking the last traded price on Uniswap. As illustrated by the previous table, this is problematic when considering the price slippage associated with sizable orders.”
This criticism is fair, but attempting to factor in liquidity of collateral is non-trivial, though our oracle does not simply look at the price on Uniswap, the fact that prices on Uniswap and the oracle are aligned speaks to the efficiency of the arb bots operating on Uniswap more than anything else. For some more detailed thoughts on this issue see my article cryptoasset collateral concerns.
“If Synthetix is to succeed in the coming years, the integration of a liquidation process, more liquid collateral types, and some kind of monetary policy tools are a must. This may prove to be unpalatable to existing SNX holders, who will be exposed to liquidation risk, but without these redemption guarantees demand for synthetic products, and trading volume on the Synthetix Exchange, will likely flounder.”
I completely agree with this, but I think Matteo has underestimated our community. They are very much long biased and understand the system must be more robust for it to gain significant adoption. While we are very excited about the current traction, with the exchange doing hundreds of thousands in volume daily, this is only the beginning and we have extensive plans to improve the mechanism and make it more robust.
As a final point, detailed and critical analyses of crypto projects is unbelievably valuable. The fact that we have analysts like Matteo taking time to intelligently dissect projects and call out issues they find is amazing. Contrast this to 2017 and as just one example "ICO Bench" and I think we can all agree that as a community we are much better placed to identify valuable projects building within the crypto ecosystem now than ever before.