In this episode of the DeFi Download, Piers Ridyard explores risk management solutions in DeFi with Potion.Finance contributors Alfablok and Buendia.
Potion is a risk management layer for DeFi, allowing people to manage the risk of their crypto in DeFi positions as well as the volatility that is inherent in the DeFi space.
Alfablok and Buendia take Piers on an expedition through the intricacies of risk-taking and risk management, proposed alternatives to the Modern Portfolio Theory, Nassim Taleb’s risk management theories, and applications of the Kelly criterion.
[00:00:46] Where did the inspiration for Potion.Finance come from and what is the purpose of its creation?
[00:01:52] An overview of Modern Portfolio Theory and its current place in traditional finance
[00:04:30] Limitations of the Modern Portfolio Theory
[00:08:38] The first steps in active risk management on DeFi
[00:10:32] Explanation of the terms “hedged positions” and “put options”
[00:15:44] Potion’s solution for making hedging accessible to people who are unfamiliar with the complexities of put options
[00:18:51] Addressing the two sides of put options: the buyer and the seller
[00:25:58] Potion’s simulation laboratory
[00:33:00] The Kelly criterion and how the Potion team implemented it into its protocol
[00:45:23] Creating an emergent pricing model based on the utilisation
[00:47:06] Conclusion: Does Potion provide an easy way for people to hedge, and does it generate option prices that people are willing to pay?
Hello, and welcome. I am Piers Ridyard, CEO of the decentralized finance protocol Radix, a public layer entirely focused on bringing DeFi to the mainstream. This is our podcast, the DeFi Download, a show about decentralized finance and all things crypto, where we dive into the details of the projects, assets and services that are powering the DeFi revolution. Today, I'm joined by Alfablok and Buendia, contributors to the Potion project at Potion.Finance, a risk management layer for DeFi, a way for people to manage the risk of their crypto in DeFi positions and manage the volatility inherent to the space. Guys, thank you so much for coming on the show.
It's a pleasure. Thank you for having us.
Thanks a lot, Piers.
So, let's just start off with where did the idea for Potion.Finance come and what's happened since you first— what was that initial insight that you guys got to the space that meant that you wanted to build this so much?
So, it all started, I believe, with the attendance to a seminar by Nassim Taleb, in New York [laughs]. He ran this thing called RWRI, Real World Risk Institute, I believe it stands for. It basically is a seminar on how people should think about risk taking, and I was really interested in how to manage the risk in my portfolio, like a traditional portfolio. In the past, you'd do risk management through basically diversification, like Modern Portfolio Theory is about if I combine different assets with different volatilities, then my overall portfolio will be risk managed like this. And he has a different point of view; he thinks that you should take a more active risk management—
So, let's dive into that. Because there's going to be some listeners who are going to be like, “I don't know what modern portfolio theory is,” right? So, Modern Portfolio Theory, if you can just give a quick summary on what's the current thinking around that in traditional finance, and why diversification is such an important concept?
So, this is something that was crafted by this guy, Markowitz, and he ended up winning a Nobel Prize for it. I think it's been the bedrock of risk management for a long time. But this is now dating… Oh, man, I'm not sure if the 60s or something like that, some time ago. And the idea there is quite simple, which is that there is this risk-reward space, and—
There you go. So even before that. So, the idea is that each asset, you can think of it as having some volatility and producing some return. So, if you think of it like a map, it has two coordinates: the risk and the return. And his big insight was that if instead of taking one asset, you take a basket of assets, something really special can happen, which is that you can kind of go left on the map, you can have less risk without sacrificing reward. And that's the holy grail in risk management: that you want to have as little risk as possible with as much return as possible.
And the key insight for Markowitz was this idea of covariance, right, and essentially minimizing the covariance of the portfolio. Because if you have covariance, which means that two assets move at the same time, if they're highly volatile, but covariant, you actually don't reduce the risk of your entire portfolio, right? It's this idea of finding assets that have the similar risk profile that you want, but fundamentally, that have a different variance in the market. So that when you add them both together in your portfolio, ideally, if one is going up and the other is going down, they cancel each other out, and you [crosstalk] you reduce the overall variance of your portfolio.
That's correct. You want uncorrelated assets and if you can get uncorrelated assets, which obviously you have no guarantee that they will be moving forward, but if you make the assumption that some assets are uncorrelated, then you can cancel out the variance without sacrificing the return. So that's the standard, the gold standard in risk management for a long time. And if you buy books on risk management, you're going to learn about Modern Portfolio Theory. But so, let's say that this is obsolete and it has flaws in many dimensions. To begin with, you don't know whether assets will be uncorrelated moving forward. So, a lot of the assumptions that are used within Modern Portfolio Theory may not be found, and he basically teaches that there is this massive risk, one to be aware of, and that's the risk of the Black Swan or the tail risk. There are some events that have very, very low frequency but that when they happen, they're massive in nature, like very, very negative returns, for example.
Right. So, it's like saying basically look, it's nice that these assets in the normal economic conditions are not correlated, but come a massive external event or massive market crash, I'm sorry, but everything's correlated and your whole system of risk management is messed up. Because what actually doesn't matter is the small variance. What actually really matters in terms of a portfolio's return is these large variance events that happen less frequently than the small variance, which means that the Markowitz portfolio management covariance model, by the law of averages, doesn't work it out, right? Because you have one massive event, and then over four years, you have little events. So, your averages moves that massive event to a tiny proportion of the total results, but actually, it has this outsized effect on the portfolio, right?
That's completely right. And the idea is that you may be managing your risk really well for the four years, but in that one single event, you may lose 70% of your assets, which is basically not doing any risk management. So, you want to manage your assets in such a way that you're resilient to those events. And actually, he speaks about antifragility. You almost want to benefit from those types of events. The way you achieve that is basically by getting into active risk management, which is basically having active coverage of your position through options, for example, put options.
Right. And I just want to say an extra thing here, which I think is important, because a lot of people in crypto at least talk about this thing of weak hands and strong hands and like, “Oh, you know, if there's a massive risk event and the market goes down, just hold the position.” The thing is that often what happens is these massive economic events are correlated with need, like with personal need, especially for companies if they're treasury management, especially for individuals, if suddenly there's an economic downturn, usually you lose your house, or for funds that actually have to sell out of a position. So, it's not a factor of these stocks and shares and whatever will recover. Yes, they probably will. But actually, these massive outsized events can have this issue that you're now massively underwater. And because of the economic situation that is external to this, you're actually in a situation where you have to sell underwater. You have to realize the loss if you're fund manager, people are doing withdrawals or redemption claims so you have to be selling these assets underwater. If you're an individual, you've lost your job, you need to actually sell out so that you can live. All these massive events are often correlated with upsets in the economy as well, which means that cash on hand suddenly becomes important at the same time that assets have gone underwater. So, holding the asset isn't often an option.
Correct. And also, it just takes a lot of effort to recover from a severe loss. So, if you lose 70%, it doesn't take a 70% gain to go back to where you were. You have to actually do many multiples of that. And so, you basically want to avoid, as much as you can, this type of scenarios where you have very severe losses.
I never thought about that. But you're right. If I've got 70% loss then I have to 2.1 times - 2.2 times my original position that I've reduced—
Yes, something like that, almost three times. Yes. It's asymmetric. It goes one way on the downside. For you to neutralize your loss, you need to go three times up. So, you basically don't want to be in that situation. And so that's the starting point. I'm thinking, oh, gosh, oh, how do I manage my own assets, so that I caught my exposure to these black swans, or these tail events. And this is, I think, somewhere around the tail end of 2019 and we see Opyn coming to the stage, and we see Hegic. And that gets super, super exciting because these guys are offering option platforms on DeFi. And that's like, okay, so that kind of starts opening the door to active risk management on DeFi. So, I can protect my crypto holdings against these pretty massive events, which by the way, are a lot more frequent in crypto. Like a 70% drawdown on the S&P 500—
A casual Monday [laughs].
Yes, exactly. That's your regular week in crypto, especially for some altcoins. And so, we just talk like these will be incredibly interesting and exciting for the crypto community and then the lockdown came along. This is now March and we see this hackathon taking place from ETHGlobal and Aureliano and I joined it with a simple idea which was to create a permanent hedger, the idea being that I will basically protect myself against these types of downfalls. And the idea was, we wanted to simplify it so that a regular person without very sophisticated financial knowledge could just basically put their collateral in some smart contract and then the smart contract would create the risk protection that would basically insulate that capital against these losses.
So, can you give me an example? Hedging is one of those things that a lot of people throw around the term of and I don't think is always completely well explained. So, what does “hedging a position” actually mean? Can you give me a worked example of a hedged position?
So, a hedged position is you have, for example, one ETH and you have a put option on one ETH. And so, it means that… Maybe we should describe what a put option is. A put option has several parameters, the more important one is the strike and the duration. Okay, so right now Ethereum is at about $1,800. So, I could buy a put option with strike $1,500. What that does is it puts a floor to how much money I will lose on one ETH. So, if one ETH goes to 1,500, then that option is still worthless, I haven't made any gains. But the moment that I go below the strike, the moment I go below 1,500, then I'm getting rewarded, that option starts getting value, and I'm able to exercise it. It just creates a financial contract that means that the lowest price effective to me for Ethereum now becomes 1,500. Even if it goes down to zero, I'm always able to sell 1,500.
So, I always go back to the CFA definition of calls and puts, which is a right but no obligation to buy or sell an asset at a pre-arranged price within a pre-arranged date range. So, with this, what you're saying a put option is the right to sell, and your strike is the price at which you have the right to sell it, and your duration is how long that lives for. So, let's say it's a one-month put option at 1,500. So, the price now is at 1,800. Spot price of Ethereum is $1,800 per ETH. What you're doing is you're buying the right to sell one ETH at $1,500 for a one-month period. So, you're paying someone who is promising to buy from you. So, you have a price of the option. So, let's say it's going to cost me 100 bucks. I have to pay someone, I'm saying right now, at $1,800, you will buy from me over the next month ETH Ethereum for $1,500, if I choose to sell it to you, and what you're saying the reason it's worthless is because it's worthless, all the way down to $1,500. Because if the market price for my ETH is $1,600, why would I sell it to the person that I bought the right to sell it to for $1,500? Right?
And actually, it's still worthless at $1,500. It only starts to be what is often referred to as “in-the-money” once you've also taken into account the cost of the option as well. Right? So, it actually only starts to be worth anything at $1,400. Where as in I'm starting to make a profit at that point—
Correct. That's exactly correct.
And more generally, regarding hedged positions, I would say that what one is looking for when hedging a position is basically to make some profit, or just minimize the losses, no matter the direction of the underlying price. You have some security and you don't care what— well, you don't want to expose yourself to a directional risk. So basically, you sell part of your potential profit to reduce the directional risk. So that sometimes includes more general— Well, it has a whole bunch of possibilities, like options or long positions, short selling, things like that.
Right. Because you also have this whole thing of like, well, an option is in many ways a realized cost. Not going to say a realized loss, but a realized cost on a costless position. Like I'm holding ETH but I've now had to pay $100 to cover the downside risk of this ETH for the next month. So that's now making my costless position into— there's a cost to carry. And so, this is where you start to do things like sell a long option to offset the price of the short option and stuff like that, right? So, I can sell some of my profit, as you're saying, some of my upside. Let's say I sell my long option at $1,200 for 100 bucks. And I use that 100 dollars to buy the put option to save myself 1,000 from 1,500 downwards, which means that if it moons, I get up to 2,200 and then everything else I have had to give to someone for the next month. And if it falls like a stone, I'm at least covered up past 1500 bucks.
I think some of the listeners may be really well versed in all that stuff and some of them may not and this may sound like an insurmountable amount of complexity to get to really absorb that. That's exactly the problem that we wanted to solve. We wanted to create a smart contract that will take away all the complexities from having to understand how to select the strike, what is a fair price for that premium and so on, and that it kind of happened magically, in a way that it was trustless, in a way that you could inspect the contract and be confident that what was happening was fair for everybody involved.
So how do you do that as a big— That's a big challenge.
Yes, definitely. And, but anyway, we were in lockdown, we didn't have much to lose, and it just felt to me— I mean, those days felt kind of magical. There were still very few people in the space and you could just jump into any Discord and be talking to the founders. And I think the crazier the idea, the more people got excited. So, we wanted to build this “permanent hedger,” we called it, this idea that I put my holdings, and then I have like a permit. I don't need to be babysitting my position. It's like the smart contract is doing that for me. And so, the idea was, we were going to build this with either Opyn or Hegic, one of the two. And we actually were talking to both of them, which were a huge inspiration for everyone in the team. And what it turned out was that it just was not going to work. Because if you think about the permanent hedger, you want to have almost an open-ended expectation that there will be somebody making a market for you. Whatever option you need, in the process of having to hedge the position, somebody is going to be there selling that option to you. I hope that makes sense.
Yes. And you always need a maker and a taker to be able to because of the duration, whatever the duration is, it has to close out at some point. And there has to be a settlement as well.
Yes. And so, the idea was, if you think about this like, I'm permanently hedged, and I'm buying 30-day durations, that means that every 30 days, I need to rebuy some new options. And I have no idea in two years’ time, what the strike price will be. So, I just need to have a counterparty that I can rely on that I know will sell me the option I need, and neither Opyn nor Hegic have this model. It just felt it wasn't scalable and it was not very well suited for this type of robo-trading algorithms that we had in mind. That's when the idea of Potion was born. We decided, okay, so we're going to try and build one protocol that is general purpose, that it's basically able to use whichever liquidity is inside to sell as many different option configurations as possible such that, a permanent hedger, like the one we had in mind was going to be able to rely on the counterparty.
So, if I read that back to you, in how I understand what you've described, what you're basically saying is there are people who are willing to sell options, because there's a profit in doing so and there are people who need options. And the traditional way of doing it was essentially to say, well, before I can provide my capital to buying or selling an option, I need to know what the asset is that I am going to be actually taking the other side of, so I have to actively decide as an option seller, an option— someone who's providing the [crosstalk] giving options, I have to constantly be in market and constantly making decisions on whether or not I want to be providing this duration, this strike price or whatever, based on my own view of risk. And I would like to talk about the Black-Scholes-Merton model and the Kelly criterion, later on in this call, but, fundamentally, I have some models for working out what I think is the right price for an option. And so that's how the traditional model works. But what you're saying is look, actually, what there is, is there's a bunch of people who want to make money selling options, that's their thing, and they have a criterion for doing that. But they're generally happy as long as it falls within that criteria to sell those options. And then there's another side of people who actually want to hedge positions, but they can have radically different portfolios, and creating a discrete market for every single one of those assets is really difficult, and really difficult to get a threshold of liquidity for being able to do that effectively and guarantee that there's always a position to take that option. So, what we need to do is fundamentally think about how we create these two sides of the market differently.
That's completely right. We just wanted to do something that was more sustainable, long term, for both parties, so that they could have this kind of permanent hedging or permanent selling, like this idea of, I commit my capital to some smart contract and then that capital is risk managed on my behalf on a consistent basis, almost like trying to transition these markets from very active, very professional type markets to very passive, very nonprofessional, but equally sophisticated, and I think that's one of the powers that smart contracts can deliver.
Yes, and it's worth highlighting that in this last minute, what we have introduced is the need for risk management for these sellers of options. So, at the beginning, we were chatting about how the buyers of options were doing that, basically, to risk manage their positions. And now we're reaching the conclusion that to get these other actors, the sellers of options, to sell options passively they need a risk management protocol embedded in the way they sell these options—
Right. Because if you think about - which you have, but if you think of options, if I'm the person who is selling options, or even buying options as my main thing that I'm doing to create a return on capital, I'm just taking bets, right? I'm taking bets, discrete bets on what the price of an asset is going to be at some discrete point in the future. And for me to be able to make money long term, the value of the bets has to be more than the losses I sustain against those bets. And I have to be able to stay in the market for a long period of time. And if I take too many big bets, even if my betting process is right, I may end up losing all of my capital and lose the ability to take any further bets and, actually, push me out of the market. So, as I get closer and closer to my capital limits, I need some way of managing my risk to make sure that, essentially, a black swan doesn't wipe me out in the same way that it doesn't wipe the people who are holding the collateral that they want to hedge against in the first place aren't wiped out, right?
Correct. And this is a tough problem, this is a very tough problem to tackle. But that's exactly what we decided to work on. And so, continuing with the story, we decided that okay, if we want to build a protocol that is reliable for the LP side, for the liquidity providers in this scenario, the sellers of the options, they are the ones that are collateralizing the protocol so that options can be bought, we need to find some way to price those options such that it generates some safety net for the LPs, or some kind of expectation that they will not go bankrupt. Because if we want to do this passively, we can't rely on the LP fiddling with the prices. We want the smart contract to set the prices for them. And I think that the initial candidate, and what we shipped for HackMoney, was what we call the “perfectly priced Black-Scholes machine.” And this basically was a— we were really excited about it. And it was like a new way to price protocols that was back-ended. You didn't know what the price of the option was going to be until expiry. So, the idea was the users can put in a deposit that should cover very large premiums. And then once volatility takes place at expiry, you measure what volatility actually took place, you plug that into the Black-Scholes model, and then you know the fair volatility and that's the price of the option.
Right. So, what you're actually doing is you're [laughs] you're actually [laughter] pricing based on what actually happened at the end. Oh, this actually happened and the fair risk price for this was this.
Okay. Yes, clever.
Yes, and there's a nice question here like, why would a seller agree on doing that, right? Or why would a buyer agree on doing that? Yes.
I mean, the answer is the same thing as what was wrong with Markowitz, is that the Black-Scholes-Merton model uses the standard deviation across a period of time. So, if there is a single large event, and the option is a long-standing option, you're going to actually get an incorrectly priced option based on a long tail event that happened in that period. So even what the perfectly priced - I'm guessing here - is that the perfectly priced Black-Scholes-Merton model came out with as in large, long tail event cycles for a sufficient duration, you actually got a put position where the perfectly priced options still meant that the option seller lost money.
I mean, you're explaining the protocol much better than we are, but that's exactly right.
I'm glad that's right [laughter].
Yes, you got it, you got it 100% [laughter]. So, we're really conscious about, if we're designing a risk management protocol, we need to be sure that it works. If people are going to trust the protocol to manage the risk, it better be doing its job correctly. And so, we build this, the simulation laboratory, where we basically tested a ton of scenarios for what would the experience be for an LP that puts their capital in the platform and leave it for five years, 10 years, 20 years. We wanted to get a feel for what was the risk they themselves will be facing, when using perfectly priced. So, this is like if you have full visibility over what volatility will take place and you price every single time with zero error using Black-Scholes, what would happen? And that's what's on screen now. We tested different assets. So, you see Bitcoin, Ethereum, Chainlink, Maker, and then the S&P 500, gold, and the Turkish lira. Some people from the Turkish community came to our Discord, because they're really keen on having something to hedge their currency in DeFi, you know [laughter], so we decided to put it on our benchmarking desk. And what you can see is that the picture is really nasty. You can see the drawdowns of 95%, 100%. So, this is like you go bankrupt within five years, if you're consistently selling options— once again, with zero estimation error, you're following the Black-Scholes model to the tee, and you're still going bankrupt. And you can see that this phenomenon is very different in traditional finance assets relative to crypto assets. You were already alluding to this. This is because the Black-Scholes formula is based on this assumption that the distribution of returns is normal, or it's thin-tailed, or Gaussian. And the example I normally give is, this happens a lot in nature, like a lot of phenomena in nature are normal. For example, the height of people. You have a lot of people that are around 1.7 meters, that's the mass, and then you have less and less people as you go away from the 1.7. And maybe you have a few people in the world that are 220, maybe some people, I don't know how that is, seven feet, I never know the American— but you don't have three-meter people. You don't have four me— they just don't happen in nature. And similarly, you don't have five-centimetre people. And this is a big problem. Because in crypto, you do have seven-meter-tall people, and you do have 10-centimeter people that happen with worrying frequency, and these are these black swan events we've been talking about. And that's why you see that, for assets that behave more in keeping with a normal distribution, the risk is better managed, but the moment you step in crypto land, then you're in trouble. And so that was a problem we basically saw that even though Black-Scholes is the gold standard in option pricing and even though we had come up with a way to do perfect pricing on Black-Scholes, we just have to look for something else. And that's where the Kelly criterion comes into the picture. I don't know if I—
So, I do have a couple of questions about this, which is interesting, right? If this is true, and the constant selling of options on the S&P, gold, and the Turkish lira, all of which do have liquid options markets, then how does the traditional financial system currently deal with this? Because obviously, people who are selling options are not making a constant loss on the S&P 500, gold, and Turkish lira.
First, I should clarify that this is not talking about the loss. This is talking about what's the worst drawdown seen. So, the maximum drawdown is over— In this, what we're looking at is five years’ worth of trading. So, this is saying, what's the worst, the lowest balance?
Let me start over. In the simulations we assume that the LPs were starting with $100. And then we just run 1000s and 1000s of simulations and for every single simulation, we looked at what's the lowest balance that was ever observed during this five-year period. We could also look at 10 years, whatever.
Got it. So, this is not actually the last position, unless you get to 100% and then you're out.
Or unless you get out of the business of selling options when you have these 95% max [unintelligible], and you make it permanent, like an impermanent loss of a [unintelligible].
Right. A new type of impairment loss [laughter].
Okay. All right. Cool. Understood.
But to add to what you were saying before, I guess an [unintelligible] market, like the option sellers, they get this active approach, they have their own internal strategies, and they decide when they're selling and when they're not. And this filtering is basically what allows them to have profitable strategies. In these simulations, we’re assuming that the seller is always selling, is always open and there's always demand for that. And that's what causes these numbers.
Right. Okay. Got it. So where did you go from here?
So, we went into a place of pain [laughter], because we didn't know what was going to happen and by that time, some excitement had gathered around the project and there were some people that wanted to help us bring this to life. But we had evidence that this was not going to work, and we didn't want to launch a protocol that had this kind of figures behind it. But then—
It'd be great for the option buyers.
Yes, yes, correct [laughter]. Exactly. Exactly right, but not great for the sellers. And that's one of the things in crypto, I think, the protocols that are long term sustainable, are those that create the selling point where both sides of the market have a reason to be there, and have some chance of making some money. Maybe not the crazy amount of money, but you want the other side of the party to not go bankrupt. If they do that's also bad for you, you're not going to have anybody selling anything to you.
Right. It's not a zero-sum game.
No, no. Ideally not. Yes. And then Taleb also gives the answer for how should you actually do risk management and he talked about the Kelly criterion. Which to introduce briefly as well, here is this guy, Kelly. He was a researcher from MIT, and was working at Bell Labs with this gentleman, Claude Shannon.
Claude Shannon is amazing, amazing.
Just for the listeners, if you haven't actually read up about Claude Shannon, he's probably one of the most spectacularly clever people in information theory in computer science. He's just unbelievable. I have a little bit of a crush on him, he's amazing. So yes, just—
Claude Shannon, yay.
Yes, he's the man. He's one of those geniuses, and a genius for a long generation. He's the father of information theory.
Right. You can argue he's the father of modern cryptography, in many ways, and his work on randomness still forms the basis of how our cryptography algorithms work. So yes—
Right. So, these guys were collaborating, and this is a nice person to be working with, if you can, Claude Shannon. And he basically tried to answer a very easy question. Which is, if I'm offered a positive-odds bet, if I'm offered a bet, where I know that I have some edge, I know that statistically speaking, I have an advantage, is there an optimal amount that I should put at risk from my available capital? And it turns out the risk you know that— It makes a difference. It's not the same thing if you bet 10% of your holdings than if you bet 90% of your holdings than if you bet 100% of your holdings, and this is the shape that it has. If you bet— unless you're on a dynamic basis, this is not one single bet, but a bet that you're expected to repeat over and over and over and over. Right? So, the idea is that, if you are taking risk, and you put all your capital at risk, it doesn't matter if you have a positive edge. You can say, you have a nine-to-one chance of winning and if you win, you double, and if you lose, you basically go to zero. If I do this bet enough times, eventually I'll go bankrupt. Maybe I'll keep doubling three or four times, but then at some point, some bet will go against me and then I'll lose everything.
All right, so would it be right to put it in this way? So, I have 100 coin tosses, and each time heads comes up, then I win 1.25 times my money, whatever I betted. And every time tails comes up, I lose whatever I betted. Not my whole money, but whatever I betted. And the question is, how much capital of the pot I have should I allocate to each one of these coin tosses over the 400? Right?
Correct. Correct. With a big caveat. This is talking about an unfair coin toss.
Yes, so heads, 1.25 times the money, I bet, and tails, I lose my money. So, if I've got $100, and I put $1 on it, at the end of a bet, I'll have $1.25 if I won, and I'll lose $1 if I lost. So over 100, I should be able to be up, right? I should be able to be up 17.5% over, statistically speaking, right? I think that's right. But it's unfair advantage to me. But if I bet all of my money on just one coin toss, I could lose it all. And so, it's not the right amount to put $100 on one coin toss. I need to put some subcategory amount of my capital, right?
I think there is like a nuanced distinction to make here. One is what are the odds of the process? When I flip a coin do I get 50% times tails and 50% of times heads? And he basically is only dealing in the world where you have favourable odds. So, forget about the payoff. The inherent process is sided towards you like—
So, 60% chance of heads, 40% chance of tails, how much should I put on each bet? Okay, got it.
Correct, correct. Correct, right. And he basically says, you should not put all your money ever— the expected growth rate— and growth rate is almost like the compounding rate or the alpha, so how much your money is compounding over time. This varies as a function of the percentage that you're going to put out of your stash. So, if you had 100 coins, and you put 100% of the coins every single time on that debt, over time, your expected growth rate is basically going bankrupt. And so, it divides your allocation space into two areas. One area, which is you're over-allocating, and then one area in which you are under-allocating. And then one very special area, which is the optimal fraction. And this optimal fraction he made some really intense claims about. It says there is no other strategy that delivers a higher growth rate for the amount of risk you're taking, which is an incredibly powerful statement to make. So, this is what Kelly does and poker players and people that are in the gambling space know this very well. In fact, this has been proven to work in practice. I think there is this guy called Thorp that famously, I think, won something like $800 million betting against the casinos using this strategy. He was waiting until a situation arrived where he had positive odds and then he was using Kelly to decide what allocation to use. And he basically made money. Warren Buffett uses it, James Simons [unintelligible]. That is not something we've invented at all. It's a well-established risk management discipline. But it doesn't say anything about option prices. Basically, it just talks about optimal allocation of assets. So, the big idea here was, can we actually use the mathematics of Kelly to produce a bonding curve that we can use to price options? And this basically made us do something— and by the way, if this is not making entire sense, by all means, jump in. But this is how the Kelly criterion works normally: You put the premium, you put the pay-out rules and you supply the odds, you put it into the mathematics, and then it spits out that's how much you should bet. And so, our challenge was, can we actually use this backwards? Can we start by saying, this is the order size that I'm taking? So, I'm an LP, and somebody comes to me and says, I want a put option that comes with strike $100. So, it's almost like to begin with somebody saying, I want 10%— Let's say you have $1,000 position as an LP. Somebody comes along and says, I want to take $100 of your capital. So, I know already, this is 10%. Right? So, can I do something so that I make the 10% f*? Can I reverse-engineer Kelly to make that the answer the order sides that are being asked, right? And can I isolate the premium as the only variable that I leave available, so that I can work Kelly backwards, and that from an order size, I get a resulting premium?
Just to make sure I'm following, the reason that this creates a bonding curve is because the more of your capital that you have allocated to bets that have not yet actually come out as whether or not they've won or lost, the more you should be charging, more of a premium you should be charging to make sure that you don't end up in a situation where you're bankrupt on your bets?
Yes. And that's exactly the right intuition, which to us makes very intuitive sense from the start. If you have more capital at risk, like as you start going higher and higher in utilization, then your risk of a black swan manifesting and going against you becomes higher. So, the higher your risk, it becomes intuitively sensical to increase the premium so that by increasing the premium, you're automatically reducing the risk because the premium is just money in the bank, you're not going to lose the premium, the premium is safe money. And to the point where you almost want to go really high as you go to very high utils, to make sure that you're going to be able to survive, even in the event of high losses.
And the reason that you'd never go to 100% utilization is because the Kelly criterion basically says, never do that. You have to always be on the active bets that you have. You should never have 100% of your money allocated to the active bet. So that you once the bets either pay-out or they don't, you then have a renewed pot of money for which you can then allocate based on this bonding curve. Is that the right intuition?
Yes, this is definitely the right intuition. Kelly wants you to bet in such a way that you will always have some capital remaining, so that you can grow back up and survive. Kelly wants you to survive, and all the mathematics around Kelly are about maximizing this growth rate. If you have a positive growth rate expectation, then almost by definition, you're not going to go bankrupt. It prevents bankruptcy because you're never going to go to zero because you're never fully exposed. You always have the premium. And so even though you may suffer some losses, over time, you will recover. And actually, we saw this in the lab, we tested all of this and in a few moments, I think you'll see this visually, how this works in real-life. Also, just for people who are not familiar with bonding curves, for me, this is a very simple intuition that it almost gives me a way to price things that is autonomous. So, I don't need to go and ask some oracle for how much should I be charging. Much like Uniswap, or Balancer when I'm doing swaps between ETH and Dai, Uniswap is not going out and asking Coinbase or asking Coingecko. Uniswap actually makes a price emerge directly from internal variables. So, it has these two pairs and it looks at what's the ratio between the two pairs. Then according to that ratio, there is this mathematical formula that spits out a price. So here is the same idea. I'm just looking at some internal variable, which is utilization. And just looking at utilization and utilization alone without having to go outside the smart contract, I'm able to produce a price, which is really interesting, right?
And then the question for us was, okay, so what is the optimal bonding curve? I'd love to do this, but I'd like to find what is the bonding curve that makes the LP so right. That's what the Kelly criterion gives us. It gives us a way to work out what's this bonding curve.
Yes. So, what you do with what [Alfablok] described before this reverse Kelly thing, and you would go— you see these pink dots here in this curve. So, you would go for each of these dots. Okay? What's the utilization? That means what is the bet that the buyer would be asking for. I'm going to plug this in this Kelly reverse calculator and it's going to tell me what is the premium I should be charging. And by doing that I can basically draw this bonding curve.
Right. So obviously, every single person is going to be a different point in their utilization. Right? Some people are going to have been in there for longer, so, they've sold more options, other people have only just joined. So how do you create an emergent pricing model from that?
Yes, you're asking all the right questions, man.
Almost like I did my research before we had the call [laughs].
Almost, almost, right yes. So basically, there is this router that sits between all these pools, so each LP has its own independent pool with its own custom bonding curve. We will get some pre-made curves, but people can define their own if they want. And then the users will be on the other side of this router. The router looks at all these parameters and from these parameters, they're able to calculate what's the cheapest possible route. And this creates, we call it an emerging bonding curve, which is like this aggregate bonding curve that comes up from the entire universe of LPs in the network.
Yep. And what their stage of utilization is, as well, right?
Correct. And what you can see here on screen is an animated version in dynamics, where you have LPs coming into the system, exiting the system, and orders taking place. And you can see that at any point in time, there is this blue bonding curve, which makes the users see the system as if it was one single, large LP. They don't need to understand the complexities or do anything like this. They can trade with it as if it was a single system.
So, we're coming to the end of the time that we have available. I suppose there's two things that come out of this for me, which is: the original problem was, how do you make this easy for users to hedge themselves. What we've spent a long time discussing is how do we make it profitable for people to provide autonomous risk pricing to people who might want to buy options, which is a related and important problem, but not quite the same problem. And then the other thing is, at the end of this, does this create option pricing that people would actually want to pay?
So, these are really good points. I think on the first one, this is a little bit like a chicken and egg, but our view was you have to solve liquidity provision first, because without liquidity provision, there is no service. Right?
Once you've sold that, then you have a market that the users can come to, and then hopefully, the system operates in such a way that the two sides of the market can organically find the pricing that works for both sides. And so that's why we've been super concentrated now on liquidity provision. But our long-term vision remains to close the loop and provide this permanent hedging capability, which is not there in our development, but that's where we're going, eventually. And the second question was whether the prices produced are market-compatible.
Right. Is that the euphemism? Market-compatible [laughter]? Okay [laughs].
That's how we talk about it. And, basically, we've looked at a number of assets, heavily traded like the S&P 500 and whatnot and they are in the bonding curve, they're at some point in the bonding curve. So, if you looked at the S&P 500, I think, in our calculations, it's equated to, I'm talking off the top of my head, but something between 40 and 50% util, and that's where the market is pricing it. And so, you can imagine that if people take that as the reference, and you say that that's how it's being traded in traditional finance, you would expect the system to become 40% utilized organically.
Right. I suppose that a nice feature of the bonding curve as well is that it will, at the start, under-price Black-Scholes and then over time, depending on utilization, you'll get increasingly expensive and you'll just find the market equilibrium price for things, which I think actually, just coming to some sort of conclusion here, is the genius of DeFi. The genius of Decentralized Finance is this ability to create a market-aggregated risk price that is at a constant equilibrium between the people who want to buy one side and the other people who want to sell the other side that is dynamic. And it's so completely different from these discrete financial products that we create in the financial sector, in the traditional financial sector, and I think that that's what makes these things so exciting and so powerful, because, you know, market— what did you call it market, market?
Market compatible pricing, that's such a good euphemism [laughter]. Market compatible pricing. Like, it doesn't matter. The way that this works is that it makes sure that it always starts off as market compatible and finds an equilibrium point that the market compatible pricing is really whether or not the starting point is too expensive. But if it's not, then yes, this has got a really good chance of being a fantastic new way of pricing options.
Yeah, and I totally agree. And just to build on that and in closing, and we actually used this slide in our last community call to close, this is from Dan Elitzer, who worked at IDEO, and he is a prolific investor and a really strong member in the DeFi community. And he said, "The most successful defi protocols will be the ones that realize the truth: they are not lending or exchange protocols, but LIQUIDITY PROTOCOLS. Lending, exchange, futures, options... just starting points. What matters is sitting on the efficient frontier of risk and return. " And that's exactly how we think about it. We want to create a protocol that exposes this risk frontier and helps participants find it in a way that's simple and passive as much as possible. And the Kelly criterion just had to get there.
And I think that's fundamentally what the whole DeFi space is. I always say, the function of the global financial markets is efficient capital allocation, and what you're able to do here is create dynamic, efficient capital allocation in a passive manner, which is just unbelievably exciting.
It's super cool. Yes, it's necessary. It really feels to us like we're trying to build something that's necessary. And whether it's us or somebody else, that's where we're going, and the protocols that will be successful are those that get quicker and faster to these selling points of market equilibrium.
Guys, it's been such a pleasure talking with you both. Thank you so much for your time. If people want to find out more about Potion.Finance, where should they go and what should they be reading?
So, our website is a great place to start. We have a Medium post that describes these in some detail. But above all, we have a Discord community that is quite vibrant and young. And we're just trying to create an environment where people that are excited about risk management can contribute and build together with us.
Thank you very much, guys. Have a wonderful—
Thank you very much, Piers. Pleasure.
Thanks a lot, Piers.