Risk Methodology: An Introduction for Tokenised Assets in DeFi

Transcript

[music] Hi, my name's Alex. Um, my role and why I'm here is I uh was previously a risk quant at the investment bank Neora. Um I developed their Basel 3 methodology for um something called curvature or gamarisk and later moved uh to to basically do vaults in dradi um for a couple of years. So uh that's why I feel relatively qualified to give this talk. Um we're going to talk about risk methodology and effectively that lending protocols and defi are effectively

clearing houses automated clearing houses in fact. So we set something called initial margin in Tradfi and this is the same as setting the LTV parameters in DeFi. And I'm sure people have seen that this basically is the way people get into things like lever looping on chain. And what happens in Tradfire is when you miss your initial margin. So your collateral basically goes lower in value then you effectively end up with someone calls you up in Tradfire and they say, "Hey, can you top up more money?" You've probably seen this in like Hyperlquid and similar stuff to that or Binance had a famous thing about their ADL event if anybody can remember that and there were lots of complaints about the speed in which that happened but the same things happens in

DeFi but it's automated so you hit your uh level on Ave where effectively now you're at risk and then if your collateral keeps declining in value then you ultimately get liquidated. So all of this talk is effectively around about how these parameters get set, what people do to set them and what you should look for in terms of deploying into these assets. And this is one of the main things that traditional finance are looking for when it comes to deploying into D5 protocols because ultimately almost all the risk inherently within D5 protocols is on the off-chain component and who is setting the parameters and why they're setting the parameters and most importantly what are those things that are determining

why they they might be biased to do certain things instead of other things. So, I'm going to start with a thing from Lord Byron. Um, he has possibly the coolest headgear I've seen on the internet. So, um, the quote for me adds more weight. Um, he basically, and there's a number of other people that have similar quotes, but effectively what I'm trying to get across here is there's nothing new really in in DeFi other than a few things. There are a few kind of fundamental bases like shared ledger being the better than two SQL databases tying out, but ultimately there's a lot of things in DeFi that aren't actually that new and we have a propensity in this industry to try and

reinvent the wheel. Clearing houses have been around for a very very long time. For example, about 300 years or so. The first one was uh created in a Saka in Tokyo. I'm not going to dwell too much on the history, but I do want to highlight two things that are very very relevant for DeFi today and they are the BIS which is the bank and international settlements um creating a central uh counterparty model for um clearing in 2012 and secondly ISD's sim model or standardized initial margin model. Um, so I I might have pronounced the acronym slightly incorrect there because I always forget

what it stands for, but SIM is what everybody calls it. Now, the two of those, the BIS um clearing model effectively determined how clearing houses should operate and who should effectively foot the bill if there's a problem. And that's really interesting because if we look at what Ave did with DeFi United, it actually followed that exact model. We'll come to that later. Now, ISDA's SIM model is uh very very interesting because it's effectively all of the financial institutions globally sitting on what I can attest to possibly the most boring phone calls of my entire life. Um, but nevertheless, very very uh decentralized organization that comes together with parameters of how they determine you should set the haircuts for different assets. Now, importantly,

they don't set things like crypto assets. It's the only thing they don't set. In fact, I would suggest that they're very, very bad at setting the parameters of crypto assets. Notably, Basel 3 are terrible at setting the parameters of crypto assets, but when we're looking at the growth of RWAs or tokenized assets because I feel they are very all assets are real in in crypto. Um, they are very very relevant when you look at these institutions. So, one of the I I can't really go through clearing without talking about layman brothers in 2008. Um the big problem with this is that all of the risktakers were marking their own homework. So everybody involved in like say we've all seen the big short and the movies around it. So I don't have to dwell too much around this but it's

worth making a point that the reason this all collapsed is because people weren't having any transparency and why regulators came in they it came in to basically ensure that we could apply basically what are basic DeFi principles that we could verify what people are actually doing why they're taking certain risks and to make sure it doesn't happen again so that we can check things like why people are using um certain models when they shouldn't be for for certain uh qualifications. Um and two big examples of that are the ratings agencies which had big self- serving interests and um secondly AIG who effectively selling insurance with nothing to back it.

The g the the good news is that the clearing houses did indeed do their homework. So there was $9 trillion unwound um when lay layman came down in London clearing house and only 35% of the margin was used and the default funds were untouched. So that's actually quite impressive when you think about it and this leads into um similar examples in DeFi where we had like COVID related stress on Ave um particular I think the one of the biggest stress events was actually slightly after cover with um uh stake and wrap ETH during the Silicon Valley bank collapse. Ironically an offchain event led to the near collapse of some DeFi protocols. Um but in general clearing is is done quite well but it's important to understand like

why and and how these uh parameters need to be uh done correctly. Now in terms of um that's a slightly bad title here they who I mean they is is is regulators um regulators in tradies to DeFi. So I'm coming back to effectively what the uh BIS asked clearing houses to do which is on the left hand side here. So the defaulters margin is effectively this initial margin. Now on a D5 learning and borrowing platform that's your one minus loan to value limit. So it's effectively what you can sell to get the money back.

So think about this like a mortgage where if you default on your on your mortgage payments they can just take your house and then sell your house give you the cash back but you don't have a house at the end. Um that's the first thing in yellow. Um after that you have this default fund where there's additional stuff that they can they can put up as collateral and then you have the clearing house and then mutual uh parties who are affected in in in the recent event with um Ave. It's really interesting because they accidentally um followed exactly what the BAS suggested which is once you blow through all of these top things, you then just go out cap in hand to effectively all the people who want to use your protocol in the future or clearing house in the future and say

look if you've made a load of money with us, you need to come fundamentally come together and like help bail out the the situation that's arisen. So I thought it was really really interesting that effectively Ave fell back on the BIS model. And I mentioned this to the RV team and they were actually really surprised. Um this actually is a bit of a curveball and I felt that I can't go through a call um a chat about uh clearing and risk without actually referencing Nicholas Celeb. So this this I find is really interesting because clearing is not a robust process and I think a lot of people um in history thought of things like fragile and robust as being two opposites. Now, it's really interesting that about in 2006, a

new word was invented in the dictionary of anti-fragility. Now, what do we mean by fragile is effectively you drop uh a glass on the floor, it breaks. It was fine, you could drink water out of it until the point you dropped it, but when you drop it, it doesn't work anymore. Right? Whether it's robust, it means you can drop it and it doesn't break. You can keep drinking water out of it. Antifragile is where you drop a glass on the floor and it gets better. Suddenly you've got a super glass. You've got a glass that is going to give you water forever. Water for days, you know. And so clearing isn't the act of liquidation isn't something that's robust. It's actually something that benefits the party when the clearing actually happens. And I think this is really

important to understand. So when we come back to this this kind of parameter setting now I I don't know if anybody likes uh the wjack memes but I felt like I can't do a talk in a in a D5 conference without actually referencing some kind of like brainlet. Um so this is this is a this is a slight slight um slightly controversial topic because there are a lot of people in the industry that will say that they run um uh good risk methodology. I mean, who's going to put any money on a protocol if somebody says as a curator, we're not running good risk risk methodology? It would just be a stupid idea, right? The problem is you can't really verify who's correct. Now, what we do at the moment, if you say, okay, they've got a billion dollars in their protocol, so they must

be legit. They've not had any defaults in X years, they must be legit. These are obviously good indicators, but they don't really tell you what the hell they're doing. And as we saw with things like long-term capital management, LTCM being one of the biggest blowups of uh qualified um uh risk managers ever, um LTCM was, for those who don't know, was a hedge fund created by I think it was Mr. Black and Mr. Scholes. If it wasn't Mr. Black and Mr. Scholes, it was like Black and Merton, the guys who invented option maths, and they blew up spectacularly um and nearly brought down a lot of the financial system with them.

So that gives you a good example of like you can't trust people who have good credentials and Layman Brothers was brought down by one of the biggest banks in America uh managing some of the biggest capital in America in fact globally. Um so I would also push back at if they're running a lot of money they can also blow up but even more spectacularly. So when it comes to running internal proprietary risk models there's effectively nothing you can check here. Like there's there literally nothing. All you can do is you can say they're managing a lot of money. They look apart. They look like geeks. They have good credentials. I I don't know. I don't know what people generally check.

Um but really you need to have some way in which you can check through the modeling. And this is what regulators demanded past 2008. They effectively determined that the guys creating the Gaussian copulas on highly correlated mortgage portfolios were doing a really terrible job and then required them to then specify what they're going to do to actually uh model those portfolios and report to the regulators. So effectively the regulators became something similar to what we try and do in DeFi when we deploy our assets by saying this is a standard framework in which we need to be able to check your homework with and the the reason I chose this particular brain is another slight point which is to do with um kind of getting high on

your own supply. So when uh when people are doing uh risk methodology, one of the most common things you'll see people do is they'll say we've run a simulation on some kind of complicated Monte Carlo, which is where you model like a number of different future outcomes. And then you set parameters to make sure that you uh don't hit the worst like 1% or 10% of losses. Uh commonly known as like value at risk or expected shortfall in in finance. The problem with doing that is, let's put it this way, if I if I when when I was trading, if if if I um could set parameters based on insample data and make money on that, um I I would be

currently sitting on a Caribbean island wondering what all the noise at Jeffrey's Island was because you'd make a shitload of money. Um the issue with setting parameters on the actual insample data is they're always by definition going to be correct. And so when it comes to like an actual real risk event that isn't captured by your insample data, you're going to be sitting with your pants down and it's just not going to go well for you. Um so this this is the reason I spent so much time on this slide because it's it's a really really important point and at the moment I don't think anybody in the industry does it.

This is effectively a core DeFi DeFi tenant. You need to verify. You should never trust. And it's not even just a DeFi principle. It's just a it's just common sense principle. Um I think a lot of people aren't in a position to actually verify and you probably be thinking like, well, I can't verify. I just don't have the time. Why should I bother? Fortunately for you, there's a lot of autists in the world that will quite happily verify and they will spend a hell of a lot of time making sure that they can do so. Um, so really it's not about the fact that necessarily you individually have to go around and verify every single protocol you look at, but the key thing is that if there is a protocol that has core transparent risk methodology through through the parameter setting that they're doing, you can be pretty sure that somebody has

gone through with some kind of agenda and tried to dig some dirt on them. Now, interestingly, the SEC actually made some comments on this. I think about three days ago. Pester Pierce or Cryptom as people like to call her um made some comments on the fact that parameter setting needs to be transparent um for uh the future of DeFi lending and borrowing and moreover the ones who are taking discretionary uh decisions might be falling into uh regulatory purview under the US and I think that also is a particularly uh sobering thing for some of the people in the industry and um the activities um should in theory get better with with kind of oversight like that. So

where do I want to leave you? Before you deposit into any of these protocols, you really need to look at what is the publish published methodology. What can I actually verify here? And and with chat GPT, you can just go through and say I would ask the first question, are there any things that they're taking discretionary um uh decisions on and how wide are those discretions? Can I recomputee the parameters? And not necessarily can you individually recomputee the parameters, but could you in theory recomputee the parameters? Can you check that they have said if this is if they specify that a particular asset is a house, can you check it's a house not a boat? Is that something that you can verify? Is there an independent

parameter setter? Is there someone that sits in this process that is ultimately independently setting the certain parameters without the actual person in in the process who is doing the BD at the same time setting the parameters? This is an important one because what often can happen is you can set these certain parameters and and this is something that a lot of people won't understand so much in in unless they're really involved in this industry. If I am sitting managing the parameters of a D5 lending and borrowing platform, I can very easily have a very very big um asset manager or protocol or issuer that will come to me and say I want you to change my LTV to be higher because I want to I I have a counterparty that

wants to take a load of leverage because it's also me. And that happens a hell of a lot. So what you'll have is you'll have somebody lending a couple of hundred million dollars effectively to themselves and then levering up to hell because they've effectively taken no credit risk because it's themselves, right? If they blow up, they blow up and they were going to blow up anyway. The problem is is that when other people jump into that particular deal, what they're actually going to get hit by is the huge credit risk and leverage that is on that counterparty, that is both the borrower and the lender in the same market. And this is why sometimes on um a number of protocols you'll see LTVs approaching like 90 to 100%. Because ultimately the main anchor lender and borrower in that market is usually the

same counterparty and that's why stream finance blew up so spectacularly and took down a number of other entities associated with them. Um, so that one's really important. Um, because obviously if the getting back to the independent parameter setting, if you have no re if you're in that deal, the parameter set is making spreads on performance fees, right? So the higher the LTV, the higher the leverage, the higher the yield, the higher the interest fees they're getting paid. So this is why the independent parameter setting is important. Um, postage capital appears actually, I think, to be a GPT hallucination. So I don't know what the hell that's about.

>> [laughter] >> Um, but I think that might be the fact that if you have skin in the game, that one's an interesting one. I will I will spend a moment on that posted capital. I I think um there is an argument that you should have skin in the game as a as a parameter setter on these markets. However, this is a really interesting one that is commonly touted on X etc. I have mixed thoughts about this because if you are sitting in an investment bank trading floor, right, you will see that the market makers, the guys who set the bid and the ask on interest rate markets have a Chinese wall which is they're not allowed to talk to the asset management team. So it becomes an interesting conflict of interest if you've got the guys who are setting the parameters putting money in the protocol because it

biases them in the favor of the lenders. Similarly, if they're setting the parameters and you're borrowing, you're in bias to the borrowers. So, I don't really know how I feel about posting capital, but some people look at it as a good thing. Um, stress over back tests. Again, um, this is something that's actually quite common. I think it's a really good thing in the industry at the moment. People do tend to do scenario analysis and they tend to say like, okay, what happens if, uh, FTX happens? What happens if um, let's say uh, uh, Ferrer's capital collapses? What happens if Justin Sun gets hit by a bus and pull his family pull his cash out?

um that one's done quite well in general in in DeFi at the moment. Um I think what we do badly is looking at macro risks. So I think if anybody can look at anybody's got any knowledge of DeFi rates right now, I think um Justin um um uh might have made some comments to this. Um, so the rates are pretty bad in DeFi at the moment and the rates are pretty bad because Tradfy rates and so there there's a huge correlation between macro finance and crypto finance whether we like it or not. I think stress over things like looking at what happens if there's a global financial crisis, what happens if there's a big credit selloff in private credit. These are things I

think people don't do so well at at the moment or or publish well at the moment. So the summary is effectively everything's already done. you just need to copy it. Um, I think that's really it. And I think the the one thing I would I would I would leave you with is some some of the things we're doing on on our side um is is we're looking at uh creating a well we've created a risk methodology implementing um isim and implementing some of Basel 3 FRTB uh which are the regulatory models used across traditional finance so that you can look through all of these uh these models and see how everything's being set. Um, we're kind of thinking at the moment about maybe just open sourcing it

in the next couple of days. But it's one of those things where if you open source model and it's just you, then does anyone care? And so we're kind of thinking, should we look at other people to collaborate and co-author the paper and then open source it? And that's if anybody's interested or got any opinions on that, come and talk to us afterwards. >> Thank you very much. >> I'd just like to say thank you as well to Cartik for organizing this because it's it's one it's one of the best events in crypto. I appreciate that and it was one of the most entertaining talks I've seen. So, uh I'm sure those memes are going to go a long way from uh from this.

>> [music]