Algorithmic Stablecoins

Transcript

for your transaction fees in in that currency since sell us a project that focuses on emerging market payments of frontier market payments this might be a very useful feature there and block had us especially the size of the block Harris is optimized for fast mobile phone synchronization I promised to talk about stable coins and probably everyone here in the room is aware what why there's a need for stable coin and why a stable coin has has value for for user to use their transactions here is a special year in hope to price history

the prices of different cryptocurrencies in the year of 2018 and another decentralized stable coin on the etherion platform died and as you can see died was yeah more less stable in that time a very low volatility especially very low downside volatility while other currencies even those currencies that pack die lost a lot of value for example eath in in in 2018 I think almost lost 80% or maybe even more than 80% of its value so there's there's a need and there's a use for for stable coins for transactions there are

different types of stable coins and I want to briefly talk about those different types there's the Fiat and or commodity collateralized stable coins which would be io u--'s and have as a collateral for example the the fiat currency they trying to track in terms of values in in terms of value there are crypto collateralized stable coins like for example die where the collateral is a cryptocurrency and there are algorithmic Orson Yard style stable coins that in their functioning or in the economics around that

Sakhi mimicking a systematic central bank basically or an algorithmic central bank and probably all of you also know examples for the different types of stable coins out there I think for fiat collateralized stable coins and there's one popular example is tether tether is basically based on the promise that there is or might be some collateral somewhere in a bank and you can either believe that or not however the short history of prices we have on tether shows that is relatively

stable and a lot of the stability is derived from this speculation around tether is there actually a collateral in place could a bank run be happening and people all want that has the tether back at the same time and if this fear comes up then tether can be Peck maybe briefly and then there's again a lot of speculation maybe it won't it won't come to a bank run or maybe the the collateral is actually in place and then in reap X and around this speculation around the one tether trading at a US dollar there's a lot of trading volume and and so far the peg holes there are other other examples for

fiat or commodity backed stable coins they're also stable coins and that have different types of commodities in as a collateral so there are examples that hold a gold in collateral so physical gold in a wall somewhere which then also has to be basically audited that the gold is actually in place and they're examples for stable coins that hold diamonds in a collateral since they are easy basically to store and relatively small and they also stable coins that hold different types of other precious metals in collateral and yeah so these are

typical Fiat or commodity collateralized stable coins I think the advantage is that there's a very direct link to the value they are trying to track so if a stable coin holds US dollars in the collateral there's a very direct link if you are always able to exchange your stable coin for that dollar in collateral and therefore this this link to the value they're trying to track us is very direct this is a big advantage the big disadvantage is also as you know around for example tether this is a more centralized way and you actually have to believe that the collateral is is sitting somewhere in the example of

tethered I think there's are stories that this can or this might be sitting with banks somewhere in Costa Rica this Bank then has a certain counterparty risk this counterparty risk can actually be priced so when you go to a bank and take out a loan or credit you can actually look at the rating of that bank and this rating then should also reflect basically the riskiness of that bank to default so if example that bank in Costa Rica might default with a certain probability then credit at this Bank should have a certain price that reflects this probability and therefore

in the long run you would expect the tether trades at a discount to a dollar that reflects basically the probability of default and the probability that collateral is actually in place second example or second type of stable coin I included here are crypto collateralized stable coins probably everyone here knows the most prominent example for this with which is die there are different types of of die single collateral die that holds eath in collateral and multi collateral die that holds different other liquid cryptocurrencies in

in collateral and and you basically can create or mint died into existence by closing a CDP which is a collateralized that position and against the CDP you can basically issue die stable coin to the market by locking up your your collateral and then the last example of stable coins are seniority or algorithmic stable coins one first example in the market of of this type of stable coins is for example the base coin base coin was a type of stable coin

that also we basically tried to adjust the supply of stable coin in the market to demand of people that want to do transactions with this stable coin and to adjust supply and match it to demand there are certain types of expansions and contractions big problem with base coin was that it's actually not easy to predict how much you have to expand or contract if the price deviates from one so if you're tracking for example US dollar and the stable coin trades at one dollar and ten cents you would need to expand your supply of stable coin to

match the increased demand that led to a rising price in stable coin and this prediction how much you actually have to expand or contract is not easy to predict since you have to make an assumption around the demand function and and I think this and several other problems around the base coin design then led to the shuttle on of off the project I think some time last year where the basis team decided to give back the funding they received for for developing that stable coin I think another problem around the base coin

design was the day four contractions in supply of stable coin out there I wanted to do a certain type of security they want to issue a a bond which lets you allows you to buy back stable coin on the market when they designed it and said this is basically a bond they asked the SEC if it can certify or underwrite that this bond is not actually a bond and not actually a security something that s SEC typically not not does and and therefore I think another another problem around the design of base coin there are other examples out there around algorithmic

stable coins I I mentioned the project I'm working on hello it's an another type of algorithmic stable going I'll talk a bit about the design on sell Oh in in a minute but in in general a bit later in in my talk today I want to basically point you to some I would say empirical facts you could have a look at when judging on the stability features of a certain algorithmic stable coin design so that the talk is a bit more general than just about seller which which obviously like in terms of design I talked about these different features

already so you have again the different types of stable coins that are in the market so there are the fiat collateralized stable coins to cook to collateralized stable coins and there's in your niche style or algorithmic stable coins I also mentioned the example of of tether which is a Fiat collateral stable coin in in the first row and the governance feature of tether is basically that this is a more centralized type of stress a stable coin which has a regulated and regulated collateral in place which is off chain

I mentioned the problems around this off chain collateral that basically is built on trust that this collateral is actually sitting somewhere you also have a counterparty risk of the entity that holds this collateral another maybe item I could maybe point out is that this off chain collateral you also have the problem that you're you're stable coin could be traded 24/7 as any other cryptocurrency out there your collateral basically cannot be traded 24/7 so you have certain opening hours of the market

in the design for example Libra currently envisions you have a similar feature if Libra actually holds a portfolio of different risky asset classes like government bonds and in corporate bonds and equities then there needs to be a manager that basically manages all the risks associated with these types of asset classes so for example for the the government bonds they want to hold in the portfolio area you have a duration risk you have an interest rate risk you need to roll over the security at the time of their maturity and you need to do that when the market market is basically open

which which means then that your your collateral cannot basically match your your underlying in over the weekend for example or at night I also mentioned sell Oh so basically sell oh is a hybrid between a crypto collateralized stable coin and this seniority style or algorithmic type of stable coin so there is a a feature or a mechanism on sell oh that always adjusts the supply of stable coin to the current demand in the market

and therefore there are expansions and contractions happening and typically with these types of stable coins expansions are relatively easy you can just mint stable coin and sell it against something to the market where it gets tricky are contractions contractions mean that you want to reduce the supply of stable coin in the market to again bring back the price two to one and when you do that you need to somehow buy back basically your outstanding stable coin and therefore you need some collateral out there that is able to buy back your stable coin one important part of designing the sello stability protocol was that the

collateral should only be on chain so it should be tradable 24/7 it should be liquid and it should be basically for everyone who looks at the blockchain everyone should be able to verify that the collateral is actually in place something we cannot do and when the collateral sits with a bank for example then verification of the collateral for example is more more difficult and when you think about this algorithmic stable coins it's always the story basically that you want to match with the supply of your stable coin the current demand

in in the market and that current demand is fluctuating and here in that example in in the perfect pet case which is where the green and the blue line intercept you can see that there's a supply of 20 million stable coin in the market and there's the blue line as your demand and blue and Queen line actually meet at a price of $1 so in this case there's a perfect Peck for the for the stable coin everything is I would say going good for that stable coins it since it it holds its price back if for example the supply in this example here would only

be 19 million but the demand is still 20 million would have more demand and supply in the market and the price would then Deepak the amount of Deepak depends basically on the demand functions or how for example steep this blue line in this example here is then determines how much the price would put Deepak of of the stable coin maybe I start with this with this slide big problem is that the demand in in the market is basically random so the demand is fluctuating all

the time it depends on how many people want to do transactions with that stable coin how many people want to hold a stable coin and and therefore the demand fluctuates all the time and therefore basically matching supply and demand at every point in time is a tricky thing and when when we talk about how you could simulate and think about how demand behaves in in future when when you want to issue a saleable coin it's even more tricky because you have to simulate somehow this this random demand that that you're stable coin basically hits and and this is tricky so what you could

do is basically you could drive and assigning a stable coin you could try to mimic a a fiat backed stable coin because in the end as I said the big advantage of a fiat packed stable coin is that it's very directly linked to the underlying that it's trying to track so if you for example issuing a stable coin that is tracking the US dollar and you're offering a facility that always allows you to exchange one stable coin for one US dollar then in times where this stable coin at a centralized market like coin base or by Nance or something else would deepak then somebody could either buy it or sell it at that different price and then actually go to

the stable coin issuer and and exchange it for US dollar so for example if you could buy a stable coin at coinbase for ninety-five cents you can buy up every every tether at coinbase for 95 cents and then go to tether and say hey you actually have collateral in place give me a dollar back and then you earn basically five five cents on every transaction until the prices back at one and you actually buying up every stable coin that way it's below the dollar then drives up the price when we design sell oh we wanted to design something very similar something which offers you basically that facility that you can always approach a protocol on chain and say here I want to exchange the stable

coin for the value of the underlying however with a crypto only backing off of that stable coin and therefore there's basically a unchain mechanism which is a bit similar like a Yunus whap exchange which is an Unchained market-maker that always allows you to transact with your blockchain and when you approach basically the sello on chain market-maker you can always say I want to for example sell oh I said I want to sell one SLO dollar for one US dollar worth of collateral or I want to buy one seller dollar for one US dollar worth of collateral and as collateral as I said is as on shine collateral so in

that reserve there's a governance token which is called sell o gold and this governance token is basically similar to two eath on the etherium blockchain so it's basically your util utility token and you always have that standing facility that you are able to trade one seller dollar for one US dollar worth of seller gold and therefore when when at centralized exchanges to sell o dollar is trading below a dollar you can basically buy it at coin base and then sell it for for this on train exchange at this on train exchange and then you get sell o gold in the value of one US dollar and then you go back to two

coinbase and and get your dollar back and the whole time where there's seller dolla is basically trading below the dollar keep on doing that rate and therefore driving up the price so it's I would say a similar design you try to match supply and demand with an on chain exchange without having a non cooked or collateral and about algorithmic stable coins and how you could actually analyze the stability of of a stable coin but this design is as I said the demand is random so if you want to simulate or maybe empirically look at how could the

demand be for your stable coin you could look at historical data for example since you're you're stable coin is not life and not traded yet it's very difficult to get historical data you could then look at historical data of other stable coins like for example tether but the the big problem with that is you only have a very limited amount of data on demand for tether and maybe the first like five or eight years in in that industry or in a type of asset class is a very limited data set to actually make some inferences or some some statements around behavior of demand for stable coins since yeah I mean it's it's a new type of asset class

and maybe in the first years there's some some like bubbles people want a lot of it people want to dump a lot of it so it's maybe not the typical market behavior you would expect going forward therefore another thing you could do to analyze a stable coin and analyze the stability of a stable coins basically simulate these these path lists here and this is what we did in our stability analysis and the old version of it you can find out on our website the new and the seller website the new version of it you can find here I hope this week on the seller website and and when when we simulated different passes of demand we used a stochastic process as I said we don't know how

demand will evolve in future so we use some statistical process that can simulate as us this this demand what we did in the first version is basically we used a geometric brownian motion which is a certain type of stochastic process that you could use for for doing that for simulating demand the geometric Brownian motion is basically a process we used here which gave us a demand in amount of stable coin at the price of one US dollar at every point in time at every maybe hour or day or a second however the frequency is you want to simulate over the next

thirty years and then you can when you use this process you should specify two parameters to simulate M the stable coin one parameter is basically your expectation how it will behave in future and this is the MU here and you could for example simulate that the mu or the expectation is basically on average the demand will grow by zero and then you simulate some what or you specify some volatility around that mu which is U which you could take from a his distribution of other asset classes or from a distribution from from stable coins out there how the volatility could look like and what this process then

gives you is basically around the zero a lot of different set of paths that stand for future demand for stable coin and what you can already see here is a feature of the geometric Brownian motion which is I think more known in research where around other asset classes where this process is used more often so for example when you do research on equities you also use this process and and since for example equity is similar to cryptocurrencies could behave different then this process might suggest you can even at jumps you could add market bike jumps that

something like a coronavirus is hitting the market and and everyone goes crazy and then the stock of hand sanitiser goes up and all the other stocks go down so you could simulate that and build it in and what then half is in finance literature called the Merton model which is basically the geometric Brownian motion and you added some jumps to that geometric Brownian motion that could be a set specific or market right and what you can see here is a very specific problem of that geometric Brownian motion because when you simulate over 30 years in the end you want to have a lot of different random paths and you want to have an expectation value a mean of 0 and when you then have two paths in your

set of simulations that completely go crazy and say ok the demand in in in 30 years is 1 trillion stable coin then all the other paths should make up for it should make up for basically for it since you want to have an expectation of value of 0 and therefore all the other paths go to 0 since two paths go go crazy and this is a very specific feature which brings a lot of I would say difficulties when you actually want to simulate a future future demand I also added the geometric Brownian motion with different parameters so here you can see a set of of scenarios with

mu of 0.5 and Sigma of 0.75 and then what you can see over the 30 years is that all these paths at some point in time go to infinity because yeah you you simulate for a very long time you have an expectation that this is growing and and therefore at some point in time one of these paths starts to go go crazy and when we go back to that example one one observation which which leads to this behavior as basically the on the downside your demand for stable coin is capped because the lowest demand you can have for stable coins basically 0 nobody

wants that that stable coin on the upside is not kept and therefore you have this feature that some basically go crazy and the rest has to go to to 0 then because you can't make up for it that like some of these paths can go to minus crazy which is not not possible when you want to simulate demand and therefore in the end what we did and and I just want to point you to these different types of processes to simulate the demand for seller dollars is we actually use a process that has a mean reverting feature therefore even if some paths go too crazy at some points in time they they come back to normal in in other

points in time and therefore this process which is called geometric online uhlenbeck process is a different stochastic process you could use for for simulating demand for stable coin which has nicer features than the geometric Brownian motion for these types of empirical analysis what we then also did is we added jumps to the simulation so it could be done at a certain day someone wants to attack your stable coin and tries to sell every every stable coin in short there's George Soros in the market who wants to launch a soros type of attack to your stable coin in other points in time they could be positive jumps so it could be that you

have a worldwide organization that wants to use that stable coin for remittances for example and you have an upward jump in demand for stable coin since they use it maybe for charity payments and they send a lot of of payments with that stable coin and then your demand jumps up on a certain date and and this is something you cannot capture with this geometric engine and wound back process and therefore we added again these these jumps and these jumps when you do a simulation of different stable coins for example can then be either as it's specific so if you syncretic jumps off that certain stable coin or they can be basically market-wide jumps so basically the whole crypto market jumps to like a

lower value because I don't know a country bans crypto or or everyone now wants use crypto and therefore jumps and and what you can see here in the picture is basically the difference between the geometric origin back process pass and you can here only see the percentiles which are the dotted lines and then these solid lines are basically the differences to that percentiles that come out of the debt geometric on shine or in back process when you add these jumps to to your analysis and and as you can see there's yeah a difference that that is actually of value basically and therefore you should maybe consider that

when you do a analysis of of stable coins actually need to change my coffee so to like sum up a bit more general the points I wanted to to make around analysis of different types of stable coins is basically when you look at an analysis of the stability of a stable coin you should first make the difference between is it a scenario analysis do you want to basically learn something around about the behavior of stable coin in different for example

market environments or is it a forecast and that forecast can can be also done in a different way you can just say okay I'm a marketing person for for a tether and in one year you have 10 billion teather in the market and this this would rather be a forecast if you want to do a scenario an analysis you can do it either by doing an empirical or an theoretical study when you do a theoretical study it's more around you try to model in closed form solutions basically a a model that matches the supply of stable coin to demand of that algorithmic stable coin if you want to do an empirical study you need some type

of scenarios some type of demand behavior in in future you can basically get these different demand paths in future by doing either a historical or a simulated study I briefly talked about that so if you do basically a historical study you can basically bootstrap the behavior of demand in the market for other stable coins out there the demand for tethered the demand for for terror and so on so you can basically use different datasets from different cooped occurrences out there and can maybe even bootstrap them and generate more scenarios by for example we are looking at them in in different ways but

you can also simulate the demand and when you do that you need to have a certain type of scenario generator at when working on CLO we used these two the geometric Brownian motion and the on Steinbach process we found that the geometric Brownian motion has very difficult feature to be used in the analysis of stable coins and also maybe in the analysis of other assets out there and and therefore in the end we decided for the on Steinbach process and as I said if you expect a very normal behavior of your market in future you don't have any big outliers heavy tails

or anything then this geometric on genuine back process is fully fine to do your analysis if you actually expect that there can be like crazy outliers and there can be heavy tails in the distribution of demand for for your stay at a stable corn you should maybe look at features how you can basically bring these heavy tails into your simulation and then adding jumps for example could be one way you you you could do because basically jumps then are outliers which are basically extreme demand shifts on a certain day which wouldn't be predicted by for example a normal distribution so this is how you could think about analyzing a stable coin if you're very

interested in in more details this is work we've done while while working on cell oh and we we published a paper on this on the cell of a page if you want to have a look at it I think this week we will update the paper version which then also includes dawn shine Owen back process and so the heavy tails and and so on so if we want to download it maybe wait three days or something I hope I can can upload it by then a few more votes around around cell OH when I have this this crowd here if you are interested in cell oh we we actually just finished our

incentivized test net competition so as assets hello as a proof of stake network and you could basically try out that network by staking and validating in what we call the great celos takeoff this basically staking competition now now ended and we launching the the salame net including the governance token cell Oh goat and the first stable coin on cell oh the salad Allah around April so if you have an idea on a a project together with cell oh you couldn't basically apply for a fellowship you could run a note you

could run a validator on on main net you actually can contribute to the code base everything is basically open source and you can find a github link here you can also look at the many open positions at core team working on cell oh and also there's a medium link down here which you can also find when going on the seller webpage or the sell o clock we launched sell o camp which is basically a boot camp that shows you on how you can contribute to the sell o code base so this is a corporation of C lapse the

team working on sell o and up right so it's basically sell o camp not ypres camp so the the cooler and permissionless version of it and then there's also an ecosystem fund so if you have a great idea for a project that basically contributes to two the decentralized finance system that we hope emerges also on Unseld oh then you can pitch that idea to the ecosystem fund the ecosystem fund is run by poly chain capital and you can basically pitch your idea and then one last thing as I said we are launching the main net

soon before our main net launch we want to distribute the governance token more I for a proof of stake network it's important to have a very diversified holder base of governance token holders the first distribution will be done via coin list so this is targeted at non-us purchases only so this is if you are if you're from the US or or live in the US this is nothing for you but for everyone else can basically participate in that auction and and buy a share of the governance token if there are any questions around that or around stable coins or around stability I think I have

ten more minutes or five more minutes so please go ahead otherwise I'm also around at the conference for for the next two days so please approach me always happy to talk about economics around stable coins so yeah thank you [Applause] I don't know too much about seller but I assume that it works in a similar way that's assume that it's public permissionless network so so is the increasing or shrinking supply mechanism is like a smart contract or something so

it's kind of deterministic right and how does this fact that it's deterministic and possible game able by an attacker so how does it impact its stability and also like we had some ideas to create something similar on a term called meta coin I don't know if you heard about it and they used PID a regulator for this I mean this was the idea so what do you think about this problem so yeah you have your fully right this is basically all based on smart contracts so sello as a blockchain is a fork of the etherium blockchain so the stability protocol is implemented we

are a smart contract so everything is deterministic therefore basically the only random part in this analysis is basically the demand and therefore we use this stochastic process to to simulate the demand because then all the rest that follows how the protocol reacts to changing demand is deterministic and in the end we we try to solve the problem of basically a deterministic and and random quantities hitting each other we are a marketplace and this is basically this unchain exchange the sunshine exchange is basically this facility that always allows you to exchange stable coin for governance talking back and forth what

then can actually be happening is that market participants would try to basically trick this mechanism and trade a lot with this mechanism and this is very borrowed from from this uni swap idea of a decentralized market maker that basically on Shane adjusts prices for varying basically trading volumes trading with with this mechanism and you could imagine this Yunus work type of exchange like two sets of Tanks so basically if if the tank of for example governance talking or utility to open on chain gets depleted because a lot of people want to buy this this utility token on chain then there's this

mechanism this unit swap market maker just surprised and therefore is basically we try to mimic a a market behavior with with this this mechanism so I'm curious why why you want to build a new blockchain instead of Israel just build it on its wing so I think there are different reasons for that one reason is basically the focus on the ultralight client the focus on basically this mobile first blockchain in terms of

how you can actually interact with the blockchain so everything we built is basically for frontier markets where people don't have access to to stable currencies or a traditional banking system and have in many of the times access to ultralight clients like mobile phones phones like very very cheap Android phones for example and therefore a blockchain needs to be very I would say efficient in in terms of data in terms for each transaction it also needs to be very efficient in terms of in terms of energy it uses for each transaction and therefore basically we

used used on this blockchain certain technology like CK snarks to to reduce block headers for transactions to to have like very efficient transactions that basically cater the need of these ultra light clients and this is one of the reasons it's also a one or another reason then it's also in in terms of basically the stability protocol especially in these markets it's very important that you are able to pay for your transactions in the same currency as you do the transaction so if you for example pay and sell o dollar you should also have the option to pay for your

transaction fee in salad Allah and and therefore this is also a feature of this blockchain there's very geared towards that that use case so there's basically the PKI is basically there's a mapping of on the chain out of hashes of phone numbers to to your public key therefore you can basically find other users in that system by knowing their phone number so it's basically the the hash mapping is public so the phone number itself is not public well thank you thank you very

much and you have feel free to to come by the next day's always happy to talk about that thank you [Applause]