Hello, my name is Alex. Um, my role and the reason I'm here: I used to be a risk assessment quant at Neora investment bank. Um, I developed their Basel III methodology for something called curvature or gamma curvature or gamma risk, and later moved on to...actually working with assets in traditional finance for a few years. So I feel quite qualified to give this speech.
We'll talk about risk methodology, and how lending protocols and DeFi are essentially clearinghouses, even automated clearinghouses. In traditional finance, we set the initial margin, and this is the same as setting LTV parameters in DeFi. I'm sure you've seen this is how people do things like leverage online. And what happens in traditional finance traditional finance when you don't meet your initial margin? When your collateral depreciates, they call you and say, "Hey, can you put in more money?"
You've probably seen this on Hyperliquid or similar platforms, or remember the famous case with ADL on Binance, where there were many complaints about the speed of what happened. The same thing happens in DeFi, but automatically. You reach a level on Aave where you become vulnerable, and if your collateral continues to fall in price, you will eventually be liquidated. So this conversation is about how these parameters are set , what people do to set them, and what to look for when investing in these assets. This is one of the main things that traditional finance looks for when considering working with DeFi protocols.
After all, almost all of the risk inherent in DeFi protocols lies in the off-chain lies in the off-chain components: who sets the parameters sets the parameters , why they do it , why they do it , and most importantly—what influences their possible biases when choosing certain solutions. I'll start with a quote from Lord Byron. He has probably the coolest hat I've seen on the internet. This quote, in my opinion , adds weight to my words. He is essentially—and there are many other people with similar statements— statements— saying that there is nothing fundamentally new in DeFi, except for a few details.
There are some fundamentals, like a shared ledger, which is better than reconciling two SQL databases, but ultimately a lot of things in DeFi aren’t that new, and there’s a tendency in our industry to try to reinvent the wheel. Clearing houses have been around for a very, very long time. For example, about 300 years or so years or so . The first of these was created in Osaka, Tokyo. I won't go too deep into the history, but I want to highlight two things that are very, very relevant to DeFi today: it's BIS, which is the Bank for International Settlements, which created the central counterparty model for clearing in 2012, and secondly, the SIM model from ISDA, or the Standardized Initial Margin Model.
Um, maybe I mispronounced the acronym a little because I always forget how it stands for it stands for , but everyone calls it SIM. So these two things, the BIS clearing model, essentially defined how clearing houses should operate and who should actually pay the bills if there was a problem. And it's very interesting because if we look at what Ave did with DeFi United, they actually followed this very model. We will come back to this later. Now, the ISDA SIM model is very, very interesting because it's virtually every financial institution in the world sitting on what I can attest to as possibly the most boring phone calls of my life.
Um, but despite that, it's a very, very decentralized organization that comes together with parameters for how they determine how to set discounts for different assets. Importantly, they don't set them for things like crypto assets. That's the only reason they don't install them. In fact, I would say they do a very, very poor job of setting parameters for crypto assets. In particular, Basel 3 sets terrible parameters for crypto assets, but when we look at the growth of RWA or tokenized assets, because I believe all assets in crypto are real.
Um, they are very, very relevant when you look at these institutions. So, I can't talk about clearing without mentioning Lehman Brothers in 2008. Um, the big problem is that all the risky players have been checking their own homework. So everyone involved, let's just say, we've all seen "The Downside Game" and the movies around it. So I don't need to dwell on this too much, but it's worth noting that the reason noting that the reason why this all fell apart is because people didn't have any transparency.
And regulators stepped in, essentially, so that we could apply the basic principles of DeFi, so that we could check what people are actually doing, why they are taking certain risks taking certain risks , and so that this doesn't happen again; so that we can examine why people use certain models when they shouldn't for certain purposes. And two big examples of this are the rating agencies, which had big vested interests, and secondly, AIG, which was actually selling insurance without having anything to back it up . The good news is that the clearing houses have really done their homework. So, during the collapse of Lehman Brothers, $9 trillion worth of transactions were liquidated in the London clearing house, with only 35% of the margin being used, and default reserves remaining untouched. This is actually quite impressive when you think about it, and it brings us to similar examples in DeFi similar examples in DeFi , like the COVID-related stress on Aave ; I think one of the biggest stressful events was the period just after COVID with staking and wrapped ETH during the Silicon Valley Bank collapse.
Paradoxically, an off-network event led to the near- total collapse of some DeFi protocols. But overall, clearing is performed quite well, but it is important to understand why and how these parameters should be set correctly. Now, regarding the somewhat unfortunate headline here: by " headline here: by " them" I mean the regulators. Regulators in traditional finance have found a specific answer, and this also applies to DeFi. So I go back to what the BIS asked clearing houses to do, which is pictured on the left.
So, the default margin is actually the initial margin. On a DeFi lending and borrowing platform, this is your “unity minus the loan- to-value ratio limit” value. That is, it is actually something you can sell to get your money back. Think of it like a mortgage: if you default on your mortgage, they can just take your house, sell it, give you your cash back, but you end up without a home. This is the first one highlighted in yellow .
So, after that, you have this default fund where you can deposit additional assets as collateral, and then there's the collateral, and then there's the clearinghouse and the parties that were affected by the recent Aave incident. It's really interesting because they accidentally did exactly what the BIS suggested: when you exhaust all these upper levels, you just go out and reach out to anyone who wants to use your protocol or clearinghouse in the future and say, "Listen, if you've made a ton of money with us, you should come together and help solve this situation situation ." So I find it very interesting that Aave actually resorted to the BIS model. I mentioned this to the Aave team and they were genuinely surprised. This is a bit unexpected, and I felt like I couldn't talk about clearing and risk without mentioning Nicolas Taleb.
This seems extremely interesting to me, because clearing is not a reliable process, and I think many people throughout history have considered "fragile" and " have considered "fragile" and " reliable" to be opposites . Interestingly, around 2006, a new word appeared in the dictionary— “ “ antifragility.” What we mean by " mean by " fragile": If you drop a glass on the floor, it will break. Everything was fine. You could drink water from it until you dropped it, but after that it no longer worked.
Right? And reliability means you can drop it and it won't break. You can continue to drink water from it. Anti-fragility is when you drop a glass on the floor and it gets better. Suddenly you have a super glass.
A glass that will give you water forever. Water for long days, you know? So, clearing, the liquidation process itself , is not something reliable. This is actually what benefits the party when clearing actually occurs. And I think this is extremely important to understand.
Going back to the settings, I don't know if anyone likes Wojak memes, but I felt like I couldn't speak at a DeFi couldn't speak at a DeFi conference without mentioning some meme about "brains." This is a somewhat controversial topic, as many people in the industry claim to use good risk assessment methodology. I mean, who is going to invest money in a protocol if someone like a curator says we're not using proper risk assessment methodology? That would just be ridiculous, wouldn't it? The problem is that you can't actually verify who is right.
Now we do this: if you say, "Okay, they have a billion dollars in the protocol, so they are reliable." They have not had any defaults for X years, so for X years, so they are reliable. These are obviously good metrics, but they don't really tell you what the hell you what the hell they're doing. And, as we saw with Long-Term Capital Management, LTCM is one of the greatest failures of skilled risk managers; For those who don't know, LTCM was a hedge fund created, it seems, by Mr. Black and Mr.
Scholes. If not Black and Scholes, then Black and Merton—the guys who invented the math of options, and they burned out spectacularly, nearly taking down a significant portion of the financial system. This is a great example of not trusting people just because they have good recommendations; Lehman Brothers was one of America's largest banks, managing vast amounts of capital in the US and around the world, but it too collapsed. Therefore, I would also object: if they operate with big money, they can also burn out, and even more spectacularly. So when it comes to using internal proprietary risk models, there is actually nothing you can test.
That is, literally nothing. All literally nothing. All you can do you can do is say they are managing big money. They look solid. They look like geeks.
They have good recommendations. I, I don't know. I don't know what people usually check. But what you really need is some way to test the simulation. And this is exactly what regulators demanded after 2008.
They actually found that the developers of Gaussian copulas for highly correlated mortgage portfolios were doing a terrible job, and then required them to clearly state how they would model these portfolios and report to regulators. So, in effect, regulators have become similar to what we're trying to do in DeFi when we're deploying our assets, saying, "Here's a standard framework that standard framework that we should be able to verify your work with," and the reason I chose this particular example is because of another nuance—that you shouldn't rely too much on your own development. So when people are talking about risk methodology risk methodology , one of the most common things most common things you'll see is claims of doing complex Monte Carlo simulations, where you're modeling a number of different future outcomes. And then you set parameters to make sure you don't end up in the worst 1% or 10% of worst 1% or 10% of losses. In finance, this is commonly referred to as "value at risk" or " at risk" or " expected shortfall."
The problem with this approach is this: I'll put it this way, if I could set parameters based on sample data when I was trading and make money from it , I'd be sitting on a Caribbean island right now wondering what the fuss was about on Jeffrey's Island, because you would have made a ton of money. The problem with setting parameters based on actual sample data is that they will, by definition, always be correct. So when a real risk event occurs that your sample data doesn't capture, you're left with nothing, and nothing good will come of it. That's why I spent so much time on this slide, because it's an extremely important point, and at this point I don't think anyone in the industry is doing it. This is actually the basic principle of DeFi.
You should check. You should never trust. And this isn't even just a DeFi principle. It's just a common sense principle . I think a lot of people are unable to actually check.
And you're probably thinking, "Well, I can't check I can't check ." I just don't have time. Why do I need this? Luckily for you, there are plenty of enthusiasts in the world who will happily check everything out. And they will spend a lot of time making sure they can do it.
So, it's not really a matter of you personally having to go and check every protocol you look at. But the main thing is that if there is a protocol with a transparent risk methodology through setting parameters, you can be sure that someone with some interest has already tried to find incriminating evidence on them. Interestingly evidence on them. Interestingly , the SEC recently made some comments on this. It seems like about three days ago.
Hester Pierce, or “Cryptom” as people call her, has expressed some thoughts on how setting parameters should be transparent for the future of DeFi the future of DeFi lending and borrowing. Moreover borrowing. Moreover , those who make discretionary decisions may come under regulatory scrutiny in the US. And I think it's also a particularly sobering moment for some people in the industry, and, theoretically, the business should be better with that kind of oversight. So, what do I want to emphasize?
Before you invest in any of these protocols, you really should take a look at the published methodology. What can I actually check here? With ChatGPT, you can just walk in and, say, ask the first question: are there things they have discretionary decisions about, and how broad is that authority? Can I list the parameters? And not necessarily, can you individually list the parameters, but could you theoretically list them?
Can you verify what they said? If they indicate that a certain asset is a house, can you verify that it is a house and not a boat? Is this something you can check? Is there an independent parameter setter? Is there someone in this process someone in this process who ultimately sets certain parameters independently, without the business development person setting those parameters at the same time?
This is an important point because it can often be the case that you can set these certain parameters, and this is something that a lot of people won't understand unless they are involved in this industry. If I manage the parameters of a DeFi DeFi lending platform, I can very easily get a call from a very large asset manager, protocol, or issuer who says, “I want you to increase my LTV because I have a counterparty that wants to take a lot of loan funds, because that’s me too.” And this happens very often. So you'd have a situation where someone is lending a few hundred million dollars essentially to themselves and then over- leveraging because they're not actually taking on the credit risk because it's themselves, right? If they " right?
If they " explode"—they explode, and they would have exploded anyway . The problem is that when other people get involved in this particular deal, they are actually exposed to the enormous credit risk and leverage of this party, who is both the borrower and the lender in the same market. That's why on a number of protocols you will see LTV figures will see LTV figures approaching 90-100%. After all, the main key lender and borrower in this market are usually the same party, which is why Stream Finance exploded so spectacularly and dragged a number of other organizations related to it with it. So, this is really important.
Because, going back to independent parameter setting, if you're involved in such an agreement, parameter settings generate revenue from performance fees, right? So the higher the LTV, the higher the leverage, the higher the yield and the higher the interest fees they get. This is why independent parameter setting is important. It seems that " important. It seems that " postage capital" is actually a hallucination of GPT.
So I have no idea what it's all about . But I think it might be because equity risk is an interesting point. I'll take a moment to talk about this " capital invested." I think there is an argument for you to risk your own money if you are setting parameters in these markets. However, this is a very interesting question that is often discussed in X and so on.
I have mixed feelings because if you work on the trading floor of an investment bank investment bank , you will see that the market makers who set the buy and sell prices in the interest rate markets have a " rate markets have a " Chinese wall"—they are not allowed to communicate with the asset management team. So there is an interesting conflict of interest if the people who set the parameters invest money in the protocol, as it tilts them in favor of the lenders. Likewise, if they set the parameters and you borrow, you will be biased towards the borrowers' interests. So I'm not entirely sure about the capital allocation, although some see it as a positive. Stress positive.
Stress tests on historical data. Again, this is a fairly common practice. In my opinion, this is very good for the industry at the moment. People tend to do scenario analysis and ask, " analysis and ask, " Okay, what happens if FTX crashes again?" What happens if, say, Three Arrows Capital collapses?
What happens if Justin Sang gets hit by a bus and his family withdraws all their money? Hmm, things are pretty good in DeFi right now overall. Hmm, I think we're doing a poor job of assessing macro risks assessing macro risks . So, I think if anyone can look, if anyone has an understanding of the stakes in DeFi right now, I think Justin could give some comments on that. Um, so DeFi rates are pretty bad right now, and they're bad because of TradFi rates, and so there's a huge correlation between macro finance and crypto finance, whether we like it or not.
I think it's worth thinking about things like what would happen in the event of a global financial crisis or a big sell-off in private lending. These are the things that I think people aren't doing very well with or covering very well right now . So, a quick summary: basically, everything is already done. You just need to copy it. Um, I think that's really all.
And I think the one thing I would like to leave you with is what we are doing on our side, um, we are looking at creating, well, we have created a risk methodology, implementing ISTIM and some of the Basel 3 FRTB provisions, which are regulatory models used in traditional finance traditional finance , so you can look at all these models and see how everything is set up. Um, we're currently thinking about maybe just opening up the source code in the coming days. But it's one of those things: if you open a model and it's just you, does anyone care? And so we're wondering if we should look for other people to collaborate and co-author the article co-author the article , and then open source it. So if anyone is interested or has any thoughts on this , come and chat with us after the talk.
Thank you very much. I would also like to thank you for organizing this event, because it is one of the best events in the crypto industry. I appreciate it, and it was one of the most entertaining talks I've seen. So, I'm sure these memes will have a long life after this.
Automatic transcript — names and jargon may be misspelled.