A non-Ponzi, algorithmic stable coin should essentially be a distributed mechanism for gathering and maintaining appropriate deposits for the stablecoin supply issued.
In traditional banking, there are centralized circuit-breaker mechanisms to prevent the system from collapsing in times of extreme stress. For instance there can be a limit placed on withdrawals. This is a necessary part of the system. Without that withdrawals can spiral. $LUNA / Terra didn't have those mechanisms. In fact, under stress they increased the available rate of withdrawals, making the problem worse not better. This is a kindergarten level mistake. That's why it failed so badly. They shot themselves in a foot and then kept on shooting.
In a distributed system, that kind of limiting cannot be imposed centrally, but rather has to be maintained algorithmically. This is a huge challenge as often the community will have to deal with unknown unknowns. To me it seems that the best course of action would be for the algorithm to behave overly-conservatively, overreacting to stress, rather than underreacting. Then the community would be able to reverse or adjust some of those reaction via voting process. However the tough thing about it is that what is best for everyone will not be best for every single person individually. Essentially a large-scale Prisoner's Dilemma.
For now at least, USDN seems to be doing better or "less worse" than Terra. My understanding is that it's mainly due to limits imposed on withdrawals and due to rate limiting of the algorithm (NSBT is prohibitively expensive at the moment). Not sure if they will survive, but at least they seem to be collapsing much more slowly.
In the article you mentioned placing limits on withdrawals and always assuming worst-case scenario. This to my understanding will be inefficient and will effectively lead to over-collateralization. No central bank would ever conduct their policy this way, it's wasteful. But they don't have to - if everyone indeed decided tomorrow to withdraw all their money, withdrawal limits would be swiftly enacted, until the pressure lessened and the underlying reasons for the panic went away. It is much more efficient to design monetary policy for a reasonably probable situation and then have extreme measures for extremely rare events. Conducting monetary policy like every day is a Black Monday is moving back thousands of years in fiscal policy, towards a much less efficient economy.
Thank you for your thoughtful opinion. I mostly agree with you. However, I would like to add some comments on your reply.
First of all, my article is not about ultimate solution to make stablecoin protocols sustainable but about additional measures that could be taken care to reinforce its sustainability. With respect to protocol-level solutions, I could list tons of safety measures that could protect the protocol.
Regarding your comment in last paragraph that states monetary policy, I want to stress that stablecoin protocols are standing on more fragile ground where any stablecoin holders could turn into bank runner who try to exit their stablecoin position. Furthermore, ecosystem related to stablecoin protocols play various roles these days: creating main demands for the coin, first-line price stabilizer. In that aspect, stablecoin protocols should monitor risk factors in its ecosystem.
"What if only 10% of UST total supply was in Anchor protocol?", "What if Anchor protocol's UST deposit has lockup period?" "What if there were more than 10B reserves in UST3CRV pool?". In the above scenario, there wouldn't be that fierce UST sell-off pressure or Terra protocol could successfully absorb the sell pressure.
Protocol-level limit on withdrawl
Protocol-level measures to provide circuit-breaking mechanism is necessary. In fact, that's what early-day Terra protocol was trying to implement into its protocol. TFL team tried to make daily mint/burn cap based on the total supply. However, after deciding the implementation to be like CPMM-like style, they somewhat overlooked that CPMM mechanism could not operate as hard cap (circuit break) mechanism during death spirals as you have seen last week.
Flaw of CPMM algorithm
The severe problem last week was that swap spread fee generated from on-chain market's CPMM algorithm could not take LUNA minting speed into account during death spiral. It's because their fee could not outpace the LUNA price fall speed during plummet. TFL should have put some circuit breaking mechanism based on LUNA total supply change due to mint/burn mechanism.
Comment on USDN mechanism
I think USDN mechanism is far worse than Terra protocol's mechanism. In fact, USDN should be regarded as a collateral-backed stablecoin where price-volatile asset (WAVES) is used as collaterals. NSBT is just a measure to complement the reserve when WAVES price falls. However, after NSBT price becomes more than $1, there isn't any reason for people to buy NSBT. Furthere, arbitrage privileges provided when staking NSBT is not profitable enough for arbitrageurs to invest around $20k to buy NSBT and stake it. In this regard, USDN couldn't restore their collateral backing ratio and it is somewhere between 30-40%. It just relies on WAVES price changes. I think USDN mechanism is worthless to comment or reference. It is just based on wrong understaning on seigniorage-share and collateral-backed stablecoins.
Placing limit on withdrawls
Even though I wrote some protocol-level mechanism to limit mint/burn speed is needed in this reply, I didn't argue that statement in this article. Rather, I argue that stablecoin protocols should be aware of the plausible worst-case sell pressure. Especially for algostables, no protocol could endure massive sell-off pressure where more than 70% stablecoin holders are trying to exit their position.
You said no central banking protocols is estimating the worst-case sell off pressure. However, Frax finance is doing that in a similar manner. They always estimate and compare worst-case sell-off pressure and their protocol-owned liquidity. Based on that, they calculate the protocol's ability to defend FRAX price during blackswan event. Furthermore, it is not that difficult in decentralized economy as all I have to do is just monitoring the smart contract storage.
May 20, 2022·edited May 20, 2022Liked by Lano Technology
Hey, thank you for the insightful information, I'll definitely look into the other examples you list in your comment.
About the USDN / WAVEs. To be clear, I'm not in any way supporting USDN / WAVES. My current prediction is that it will suffer a "slow spiral" event where roughly the same thing will happen to it as LUNA, except in slow motion due to withdrawal limits placed. That doesn't of course address the core issue, and is probably beyond saving at this point. Unless they find an investor stupid enough to take the risk and try to recapitalize the ecosystem (doubtful, sounds like catching a falling knife to me).
A non-Ponzi, algorithmic stable coin should essentially be a distributed mechanism for gathering and maintaining appropriate deposits for the stablecoin supply issued.
In traditional banking, there are centralized circuit-breaker mechanisms to prevent the system from collapsing in times of extreme stress. For instance there can be a limit placed on withdrawals. This is a necessary part of the system. Without that withdrawals can spiral. $LUNA / Terra didn't have those mechanisms. In fact, under stress they increased the available rate of withdrawals, making the problem worse not better. This is a kindergarten level mistake. That's why it failed so badly. They shot themselves in a foot and then kept on shooting.
In a distributed system, that kind of limiting cannot be imposed centrally, but rather has to be maintained algorithmically. This is a huge challenge as often the community will have to deal with unknown unknowns. To me it seems that the best course of action would be for the algorithm to behave overly-conservatively, overreacting to stress, rather than underreacting. Then the community would be able to reverse or adjust some of those reaction via voting process. However the tough thing about it is that what is best for everyone will not be best for every single person individually. Essentially a large-scale Prisoner's Dilemma.
For now at least, USDN seems to be doing better or "less worse" than Terra. My understanding is that it's mainly due to limits imposed on withdrawals and due to rate limiting of the algorithm (NSBT is prohibitively expensive at the moment). Not sure if they will survive, but at least they seem to be collapsing much more slowly.
In the article you mentioned placing limits on withdrawals and always assuming worst-case scenario. This to my understanding will be inefficient and will effectively lead to over-collateralization. No central bank would ever conduct their policy this way, it's wasteful. But they don't have to - if everyone indeed decided tomorrow to withdraw all their money, withdrawal limits would be swiftly enacted, until the pressure lessened and the underlying reasons for the panic went away. It is much more efficient to design monetary policy for a reasonably probable situation and then have extreme measures for extremely rare events. Conducting monetary policy like every day is a Black Monday is moving back thousands of years in fiscal policy, towards a much less efficient economy.
Thank you for your thoughtful opinion. I mostly agree with you. However, I would like to add some comments on your reply.
First of all, my article is not about ultimate solution to make stablecoin protocols sustainable but about additional measures that could be taken care to reinforce its sustainability. With respect to protocol-level solutions, I could list tons of safety measures that could protect the protocol.
Regarding your comment in last paragraph that states monetary policy, I want to stress that stablecoin protocols are standing on more fragile ground where any stablecoin holders could turn into bank runner who try to exit their stablecoin position. Furthermore, ecosystem related to stablecoin protocols play various roles these days: creating main demands for the coin, first-line price stabilizer. In that aspect, stablecoin protocols should monitor risk factors in its ecosystem.
"What if only 10% of UST total supply was in Anchor protocol?", "What if Anchor protocol's UST deposit has lockup period?" "What if there were more than 10B reserves in UST3CRV pool?". In the above scenario, there wouldn't be that fierce UST sell-off pressure or Terra protocol could successfully absorb the sell pressure.
Protocol-level limit on withdrawl
Protocol-level measures to provide circuit-breaking mechanism is necessary. In fact, that's what early-day Terra protocol was trying to implement into its protocol. TFL team tried to make daily mint/burn cap based on the total supply. However, after deciding the implementation to be like CPMM-like style, they somewhat overlooked that CPMM mechanism could not operate as hard cap (circuit break) mechanism during death spirals as you have seen last week.
Flaw of CPMM algorithm
The severe problem last week was that swap spread fee generated from on-chain market's CPMM algorithm could not take LUNA minting speed into account during death spiral. It's because their fee could not outpace the LUNA price fall speed during plummet. TFL should have put some circuit breaking mechanism based on LUNA total supply change due to mint/burn mechanism.
Comment on USDN mechanism
I think USDN mechanism is far worse than Terra protocol's mechanism. In fact, USDN should be regarded as a collateral-backed stablecoin where price-volatile asset (WAVES) is used as collaterals. NSBT is just a measure to complement the reserve when WAVES price falls. However, after NSBT price becomes more than $1, there isn't any reason for people to buy NSBT. Furthere, arbitrage privileges provided when staking NSBT is not profitable enough for arbitrageurs to invest around $20k to buy NSBT and stake it. In this regard, USDN couldn't restore their collateral backing ratio and it is somewhere between 30-40%. It just relies on WAVES price changes. I think USDN mechanism is worthless to comment or reference. It is just based on wrong understaning on seigniorage-share and collateral-backed stablecoins.
Placing limit on withdrawls
Even though I wrote some protocol-level mechanism to limit mint/burn speed is needed in this reply, I didn't argue that statement in this article. Rather, I argue that stablecoin protocols should be aware of the plausible worst-case sell pressure. Especially for algostables, no protocol could endure massive sell-off pressure where more than 70% stablecoin holders are trying to exit their position.
You said no central banking protocols is estimating the worst-case sell off pressure. However, Frax finance is doing that in a similar manner. They always estimate and compare worst-case sell-off pressure and their protocol-owned liquidity. Based on that, they calculate the protocol's ability to defend FRAX price during blackswan event. Furthermore, it is not that difficult in decentralized economy as all I have to do is just monitoring the smart contract storage.
Hey, thank you for the insightful information, I'll definitely look into the other examples you list in your comment.
About the USDN / WAVEs. To be clear, I'm not in any way supporting USDN / WAVES. My current prediction is that it will suffer a "slow spiral" event where roughly the same thing will happen to it as LUNA, except in slow motion due to withdrawal limits placed. That doesn't of course address the core issue, and is probably beyond saving at this point. Unless they find an investor stupid enough to take the risk and try to recapitalize the ecosystem (doubtful, sounds like catching a falling knife to me).