Keynote by VT - AMM Research, Orbswap

So what we're going to be looking at is stable coin DPEGs in the largest stable coins. So think USDC, USDT and die. We're going to look quickly at the stable coin trailemma, examine some of the DPEGs, what we could do about DPEGs. Um yeah, that's about it. So a quick recap of the stable coin trilma.

So what you have is you want to have something that is both stable, decentralized, and capital efficient. For example, if you were to look at something like cash, $1 is equal to $1. you're in control. Not really very capital efficient. Um a US bank account pretty stable.

You can borrow beyond a dollar on it. Completely not decentralized. And so you have these new um stable coins like LUSD or CRV USD or die. And even though these mechanisms are very interesting, they're not the largest stable coins. The largest stable coins happen to be something like USDC and USDT.

And the question is, are these things really stable? Um and how stable are they? Let's take a look. So one of the things that we can do to examine uh the stability of something is to look at the face base of it. So for example, this is a case of dye traded on the centralized exchange.

So if D is exactly $1, it will always be in the same spot right here. But it is not. As you can see, it kind of like differs all over the place. So as you may know, D usually has like a plus minus 1% mechanism of minting and burning. And you can actually see some of these mechanisms around it.

Now, if you were to examine something like USDC or USDT, as you can see, it's a very very sharp face base. The price stays exactly where it is as time goes on. But we can't really tell that much um just from a phase space. For that, we have to look at the returns distribution. So, just a quick recap, uh just like a regular distribution, if you look at the returns on the x-axis, red and blue, but there are different tails.

And what do the tails t tell? Well, you can have something like a thail distribution which is dashed like a gshian uh and a heavy tailed one. This is the called the Koshi distribution. So, usually it's very difficult to tell them apart. So, what people usually do is they look at it first in uh log scale on the y-axis.

This can now you can clearly see how the tail is different. For example, this looks something like a parabola a thin tail distribution. You can tell a lot about the behavior of an asset that would behave like this. This one is very difficult. And so what sometimes people do is when they examine tails or DPEGs or catastrophe events, they look at not just the logarithmic axis on the y- axis, but the logarithmic on the x-axis as well.

So for example, if you have a stable coin that depicts by.1%, 1%, 10% or like 100%. And here we can very clearly tell that there will be a difference. And what we're going to be looking at is the differences in these tails. So let's take a look at um a couple of these.

So we have on the left one we have USDC. Here we have USDT and there we have D all traded on the centralized exchange. Now uh unfortunately my animation doesn't work but when it goes down you can see how the slope actually increases which is like not very good for USDC. For USDT the slope largely stays the same and it's largely fixed. But for D what you see is that it starts going down.

Now you have this buildup and it continues. So this is the phase space that we saw previously. So when we look at these log lock log plots, these are the same log log plots that we saw before, but they're just bunched up in histograms. But the problem with histograms uh and when you look at the tail, you can see they kind of like bunch up over here. So it's very difficult to tell the tale with histograms.

So what you have to do is you have to do this thing called the rank frequency plot. What you do is you rank the frequency of all of these DPEGs and you get a kind of a plot that looks like this. And as you can see, they're all very unique. So USDC is very different from USDT and very different from D. See you see like in D you see this little mechanism where it builds up and creates a cenosidal wave doesn't happen with USDC USDT or USDT.

So for example USDC is allowed to actually slightly deep USDT on the other hand is a pretty straight line and what we want to do is we want to find out what is the slope of this line because if we know the slope of this line it will tell you how risky the asset is. So let's look at a full picture of DIUSD. You have the price moving in A. In B, we have our little histogram. In C, you do the histogram and log log.

Doesn't really tell you much. So, we do the rank frequency plot. Now, in the rank frequency plot, what we want to do is when we fit the line and if we fit the line through the entire data, we capture too much of the body. It's largely useless. And so, what we do is we cut off some of the data.

And as we cut it off, we find that we lose more data and there are few infer events. And so, our error rate increases. So there has to be a sweet spot between an increase in the error rate and how much you cut out. And when you do that, what you find is the alpha and the slope of this distribution. In this case, it happens to be around three.

What does this mean? It means the following. If your alpha of your tail is, let's say, below two, then the means nothing. What does that mean? It means when you take the average of it, you see $1, $1, $1.

You're thinking to yourself, the next prediction is going to be a dollar. No, it's not going to be a dollar. It can be infinity or zero. Similarly, if it's like around three or below three, you you lose the standard deviation. You lose the variance.

You you lose the volatility. When you sample the volatility, you think you're seeing a pattern. You're not. And so, a lot of people build these technical indicators or whatever it is. They don't work.

And the reason why they don't work is some of the data is actually below three. Below four, um, you uh lose skew and five is curtosis. Don't worry about those. So, let's now look at those stable coins that are traded on the centralized exchange and let's look at that tail. What does that tail give us?

USDC, strangely enough, gives us a tail of two. Uh, DUSD, strangely enough, three, slightly a bit better. And this is really surprising to me for um, Tether is actually a more stable one. So, this is actually a result that I found was kind of unusual. So uh conclusion on a centralized exchange tails are slightly less fat than on the centralized exchange as we're about to see when we look at the tails on a decentralized exchange.

What I look at is the stable coin pools on unis swap the oldest ones that have been ever in existence uh USC USDT USDC die and USDT die. And what we find is a lot of them have an alpha that is below two. And this is very strange that you may have heard um people recently say this word like lowrisk DeFi. Well, it's kind of strange to say that because when you when when we analyze um pools, we usually have two assets and it's usually priced in a stable coin. The denominator and the denominator itself is not stable.

Now, that's that's some really disturbing implications there. So, I wouldn't really call it lowrisk DeFi. Um so, what could we do? What could we do to fix the situation? Well, one thing that we could do is maybe we could create a portfolio of two stable coins for the sake of simplicity.

They're not on an AMM just a just like this. So, for example, if we look at USDC and USDT, they have a slight anti-correlation here. So, um what we could do is we could do a 50/50 portfolio. But if we were to do a 50-50 portfolio, uh what will happen is the tails will still be there, but the fluctuation of these tails will not be as large. So you still have the fat tails.

You don't get rid of them. The only way to get rid of them is if the correlation is a perfect negative one. Um so here's like another suggestion uh that we came up with, which is what you could do is what if you have some sort of a prediction market or like an insurance for stable coins. Here's how you could potentially do this. This may get a bit confusing.

I fortunately enough have 10 minutes so I can explain to you everything here. Um what if we allow people to lend out a stable coin LP position consisting of a stable coin pair. So for example, let's say we have USDC USDT and the first LP provides liquidity in a particular range like a uniform range which ranges from let's say 0 degrees to 90° where 0 degrees is 0, 45° is $1, 90° is infinity. And so if you're if you've heard of like orbital AMM uh they use polar coordinates something that uh kind of how it works. So what now another person can come along and provide liquidity in a slightly different price range between let's say I can't read I'm too close to the screen sorry like let's say between a dollar and 80 cents or something like that but now let's say you were to borrow this particular position and if you have your regular LP position concentrated around a particular range it would look like this but if you were to also borrow the one right right next to it and then short it your position would look like But what if you were to combine these two and your prediction it starts to look like this.

And in this case, this kind of replicates what uh binary prediction market PF looks like because but the problem with a binary prediction market um is that the liquidity has to be concentrated in that market and so your liquidity is fragmented. On the other hand, maybe it is possible to just have it directly in the AMM. And actually an interesting question pops up which is how much such insurance cost like how much would you pay for um deep insurance? Well, the answer there is from a purely theoretical perspective if you have let's say your risk-free rate is zero um but you have volatility then a theoretical perpetual put option would approach the strike price which would be like a dollar. uh practically we don't know.

So I don't think anybody would pay a dollar to ensure a dollar. It doesn't make sense. And so somewhere in between there there would have to be a price and I think it would be nice if the market figured it out. So as a final message I will say um remember that the organic yield in DeFi that you see is largely from being exposed to a concave payoff uh with usually fat tails. And so one of the things that you could do and mitigate to mitigate this undefined risk is to hedge it with a counter party.

Hence the insurance. So thank you [Applause]