Quantitative Methodology for Morpho Markets, Gil Santos, RedStone | ETHSofia 2025

[music] Uh my name is Jill Santos and uh I am right now the lead quant at Redstone and um I've been working in traditional finance uh for for a while and joined crypto um two three years ago and before I really talk about anything I really need to contextualize you guys because I understand that you probably don't know the news uh that recently happened um which was that Redstone an Oracle provider acquired the Credora network and the reason why this is relevant is because I was um developer at the Credora network and was working on the risk models there. Credora um was and still is the lead risk ratings provider in the space and we rated assets, we rated Morpho, we rated Oiler and um the angle here about this partnership and this acquisition is that okay we have these amazing methodologies that I will talk about and now we have a distribution channel and now we have a team uh that is very talented and uh we are ready to scale this. So a little bit about Redstone. Redstone is the fastest growing oracle uh to today. Uh we have over a 100 clients.

We support over 60 chains and we secure over $4 billion in assets. You will probably know us from uh price feeds uh and data feeds in general. You'll see us on lending markets. You will see us on leverage protocols uh staking, restaking and yield products. And as I said when we talk about data feeds uh the feed that is most common is prices right but there can be other feed feeds and this is what we're trying to get at and there was a and there is a problem in this industry um that credora and both credor and rest really and you guys probably already realized this and uh which is that it's very difficult to choose the right assets the right lending and the right vaults for your personal risk appetite.

You can go on a vault page and you can sort it by yield and you will see the highest yield on top and just say let's go. I'm going all in on this one. But what's the risk behind it? How do you measure it? What's the risk adjusted yield?

Those are things that we are trying to to answer. And the problem is just getting worse and worse. We have D5 proliferation going on. When we started uh with stable coins, we had the three big players USDC, USDT, we had D. Now we have other Athena and others that are complexifying all this landscape and we have more than 200 stable coins at the moment.

Real world assets, the big buzzword. Again, we started with T bills. Now we have private credit on chain which is very very good but again more complex and more stocks coming in reinsurance refactoring all these different type of products that are available for you but how do you measure risk finally defy vaults we've seen a couple of keynotes about it uh we started with a compound or spark now we have permissionless vaults morpho oiler and you have risk curators that are trying to maximize these yield opportunities and just trying to create uh products that um eliminate uh these uh permissionless vaults. So the problem is clear. What is the solution?

The solution is risk ratings. Obviously the risk ratings is uh an approach where we basically take different risks. uh we look at network counterparty smart contracts depends on the what you're really assessing but what you're really doing is you are mixing this all in taking quantitative measurements and methodological methods to assess the risks individually and then put them all on one single scale uh that is used to compare risk across uh different products. You've probably have seen the scale on the left, right? This scale rings a bell.

It's what S&P, what Moody's uh and other rating agencies do. And uh what we've done is we've basically mapped our ratings to their ratings. And we've created a scale that is more uh easy to interpret and that is uh more cryptonative, let's say, because um it's tailored towards the risk that we see uh in in crypto. So our hope and goal is to bridge this gap a little bit about institutions and cryptonatives and doing that by creating ratings uh that are very simple to use uh that are consensus driven and very transparent. Right?

So about these ratings and how do we uh assess things so different like a moro market from a stable coin and how can we um organize these items. So we've basically organized them into three different verticals uh that have different methodologies around them but the really cool thing is that they are all under the same standardized scale. So we can compare the outputs. Assets uh is really anything that you can call like a stable coin, an RWA, liquid staking or restaking token. And in there what we want is to quantify the probability of default.

Probability of default is answering the question will this up? And um with that answer you get part of the answer because you say okay this has a likelihood of a default event but how much will I lose? That's the second variable loss given default and those familiar with traditional finance will um recognize these metrics. So assets we take this fundamental risk assessment we break it into sub methodologies and we assess each one individually. markets markets uh the approach needs to be slightly different because you need to have some sort of forwardlooking vision and you need to have a simulation framework.

So what we've done is we've uh created a simulation framework using Monte Carlo uh simulations to calculate the probability of a significant loss and a significant loss is a loss that you it's is the likelihood of having a loss over 1% of your deployed capital. Finally, vaults. Vaults are just like aggregated exposure to the underlying markets and assets. Um, so uh if if we do correctly part one and part two, part three is just aggregating these individual uh risks and also taking into account cons concentration, diversification and other government specific around the vaults. But you can see the value here because if you really do well the first part the vaults part is super uh distributable uh very easily.

And today uh the nerd part is about markets. So what I'm going to go is into a little bit more on how this market methodology and how the simulation really works. So if you think about this as like making a um a potion, uh these are the ingredients that go into the the market ratings. Uh so put yourself in the shoes of trying to analyze the risk of only one morpho market. On one morpho market for those of you who don't know is you have a collateral, you have a loan asset and you create a position with a certain loan to value.

um and uh you you want to assess the risk of as a lender um losing that that that capital right so the main ingredients are these I don't want to go into a lot of detail about each one of them but maybe we can talk a little bit about a few market characteristics are the most obvious one so if you go there you will see different uh characteristics on on different markets you have some markets that have an LLTV uh that is different from others. But that alone does not tell us a lot because you have stable to stable markets where you can have really really high LLTVs and you have volatile markets where you have a volatile to stable pair and that LTV uh can be a little bit uh lower. So you need to assess this as a whole. That's the key. Allocations distribution is also an interesting one.

We've noticed that uh different markets have different allocation profiles. So the positions uh that uh are inside the market impact the result. Uh if you have a certain market with all the positions really really close to that threshold limit at the little movement in the in the markets in the price you're going to have triggers you're going to have liquidations happening and that's no good. So the allocation distribution is important. And the final one I want to talk about is the collateral asset ratings.

uh collateral asset rating is what we credoro produced at the asset vertical. So you you imagine a market you have a collateral you have a uh that collateral will be rated by us and you have a probability of default if that collateral fails that's going to affect the market right so we take that into consideration as well in our uh simulation framework and because um we got acquired by redstone and because we are here uh with them um there's also uh a very interesting angle because during this work and during the the the times where we were just focused on on developing these methodologies we've noticed that the oracle characterization is very very important in terms of how risky the market are. Okay. So depends on on how the horra set up, you're going to have um uh different risks and you're probably if you dig into like uh reports on uh um like exploits, there will most of them will have this Oracle was exploited line. Uh and this is what we noticed as well for landing.

So we've categorized them into three different buckets depending on the nature and it doesn't necessarily mean that one is less risky than the other. Again, always depends on the whole context. dynamic oracles. You have a dynamic price feed between the loan and the collateral um pair and you're getting that price feed dynamically. This is typically used for most of the the markets.

Exchange rate oracles are kind of in the middle where you have some part of the price that is dynamic and some part that is fixed. U and you have hard-coded oracles that just rely on a fixed price feed. Um, so you can have a market that is just basically a stable tostable uh market and the price feed is just constant, just one. Um, what does that tell you? That tells you that if there's a DPEG event or a default event on the loan to collateral pair, you're not going to get liquidations.

And that doesn't mean that it's necessarily bad. It's mostly used for low liquidity markets and um smaller markets in general as well. Okay, so now the really nerd part. If we let let's dive into a little bit about the met markets methodology uh and how we solved this problem with a simulation framework here. Um we basically have a two-step simulation.

Okay, again let's put ourselves in the market shoes uh and look at a specific market in Morpho. You have you start with a collateral asset PDN rating. As I said before, you rate the collateral. Then you have a price simulation. And in this price simulation, the answer you're going going to try to get at is how likely it is that I'm going to hit the LLTV trigger.

So given a certain position, let's say you start at an LTV of 60 um and uh you have the LLTV at 86, how likely given the price movements in the future, it is that I'm going to get that trigger. This tells you how um the likelihood of hitting that trigger. But then you have to ask another question which is okay, I have hit that trigger. how likely it is that I end up with bad depth. So meaning that liquidations did not work properly and uh you end up losing money.

That's the answer we get from the second simulation and that gets us to a probability of a significant loss price simulations a little bit more into the weeds. Um but be before I I I jump into into this nerd part, I want to try to get a feel of uh uh how nerd you guys are. Um so who has heard about Monte Carlo simulations? Nice. Some.

Okay. So I can just briefly explain you. Monte Carlo is a simulation framework where you create a ton of random events and you measure uh the results. So imagine rolling a dice. If you roll your dice, your dice has six sizes.

You roll them uh 10,000 times, you're going to get eventually like one out of six in each. What we're trying to do here is creating uh the framework to create to have a simulation of price movements. uh that starts with a few um very specific conditions and then you launch uh the simulation framework and you see how likely it is that with those conditions you hit a certain um a certain target. What are those conditions? The most obvious one per volatility.

Per volatility tells you uh how likely it is that the price move uh historically or implied volatility. You can use both and how likely it is that the move happens on the central part of the normal distribution. So if you think about um ETH to USD prices um you're going to have like a volatility around two or three depending on on daily volatility and depending on on the time and that will lead to on the image on the left will lead to the center part of of of that. But we've al also noticed something that is very important which is that uh some the central part of the simulations are in the central movements are explained by normal movements normal distribution but in most of financial products products and crypto is no different the tail events so extreme events are under um understated by normal distribution because it's more likely to have if you look at historical prices you're going to see that those tail events happen more than what normal distribution suggests. So if they happen more what's what what means is that you're understating that the risk uh and we've we've built a a model specifically for the tails to capture these market extreme events and the final input is really the collateral default.

So again the likelihood of that collateral hitting a def a default event and what's the loss even that default with these ingredients you can create the image uh that you see which is basically a move a random movements of prices uh according to a starting LTV position and how likely it is that those scenarios uh reach the trigger and if they reach the trigger what happens Um let's think about again our little market at Morpho. Uh our market you have a certain position opened and you pass this loan to value that is the limit loan to value. What's going to happen is you're going to get liquidated. But if if you think about it not from a small position alone but from all the exposure that is in that market there are markets with huge exposures, right? So and there are markets with huge exposures with uh very close to that limit.

So it means that if there's a price movement and you are getting wrecked the others are going getting wrecked as well unless they act. That's why liquidity matters and this is the question that we're trying to answer is okay at the moment of um passing that trigger you're going to have uh two things that factor in uh if the liquidation is successful or not. One is the price velocity. So how fast are we moving uh in terms of price move and the second one is how much liquidity is available elsewhere. So how much uh liquidity is available for liquidators to act.

So if you have like $1 million that passed this limit uh that 1 million will have to have to be liquidated and liquidators want to use liquidity elsewhere. So what we do we basically look at the slipage curves and we define a point where that's what we our assumption for liquidity uh at the moment of liquidation and we discount it a bit uh based on volatility based on competing sell pressure and based on how much is actually being liquidated. Right? So with that ingredient we have our final bit of the simulation. uh and what we here do is just simply we break the simulation into steps and uh we emulate this liquid liquidation process.

So if you have 1 million um you know outstanding that cats is getting pushed uh over the limit and it's getting liquidated. If you have a 100,000 liquidity available, what's going to happen? The price will continue to move and you're going to try to liquidate partial uh amounts. So 100,000 price continues to start to to to move, liquidate again, liquidate again. Price continues to evolve.

You see where this is going? This is going either one or two directions or the liquidation is complete or you have actually bad depth. And this is um something that uh in our opinion with the growth of TVL um will just get bigger and bigger because you have more amounts uh that are um at risk um and less liquidity or the liquidities is not evolving uh the same way. So uh we need to to consider that that that factor as well. Okay, nerd part is over.

Let's go back to the general part. Why do we care? Why is the market methodology a thing? We care because this really serves as the foundation for vault level risk assessment. Okay, again if you look at a certain vault, you can have a rating for a vault.

And that vault is the composition of each individual market and each individual asset inside that vault. Together with correlation, track record, governance uh we can get uh this modular approach to risk. Uh and this modular ports works for multiple lending markets. And finally um why why us why do we why are we positioned to be uh leading this space uh which is basically a space that is at the moment growing because you have on one side institutions that are ready to deploy capital and on the other uh you have uh all these modular components um and all these curators that need an external validation of their um strategies. is today.

Uh we are present on Morpho. Uh so our methodology rates around $4 billion in TVL. Uh we are present uh in um with the biggest curators in in the space stakehouse, RE7, MAV and more. And the trend that we've seen was that the moment we started rating those vaults, the TVL evolution of those vaults was much bigger than the TVL evolution of the unrated vaults. So it's outperforming uh the other vaults in in in in uh TVL and uh it um it basically means that the the the ratings work as a stamp of approval for for morpho faults.

We are uh cooking oiler as well. Oiler has a slightly different modular approach for those who who are familiar with and uh we plan to be live with those soon and from here the plan is really to continue to grow on the asset side as well continue to rate more and more assets and um rate more and more vaults because at the moment that you have all the your fundamentals ready volts is just aggregating those individual pieces. Thank you very much. I think we still have some time for questions. [applause] Yes, I think so as well.

I was just checking if there are any questions. Uh yes, there's a gentleman over there. If not, I will just walk there with the mic. Let's have a quick look. We have one or two questions.

Here you go, please.

It's a wonderful presentation. I thank you very much for it. Uh you started you started your talk with the buzzword of the season the RWA. Uh would you say that quantifying risk that is linked to collaterized um assets to RWA in crypto is going to be harder in terms of auditing the assets because uh nowadays when you do collateral let's say exposition ranking uh you can clearly track where the assets are going but when you talk when you tokenize the assets the real world assets it's going to be much harder to audit actually whether whether the assets are there or not.

Yeah.

Well, this is this is going to be let's say

quite hard to to quantify and to measure and to put it on a on a higher perspective to just put a risk rating on the on the collaterization ratio.

Yeah, that's an interesting question. So, our our experience with RWA was really taking the same methodology we've been taking with stable coins. Um and the risks that we see there are um layered right. So first of all you start with asset quality itself. So uh you can start with um tokenizing a T bill is different from tokenizing a stock in I don't know a very niche stock um and the quality of that underlying makes a difference.

Then um in terms of how the tokenization process actually works, proof of reserves, uh what factors go in, we we we think that is very similar to uh stable coins actually. Um so we are not very worried about that that uh and our methodologies work with with RWAs as well. Uh the the main problem we have with asset ratings is that um the automation really because each asset is different. You need to have risk teams do proper uh due diligence and have like proper analysis together with um with the team kind of like S&P does it like traditional ratings, right? But with vaults, vaults are automated like markets are simulations.

They're running fast. We can do this daily. We can do this hourly. Um it's a much more dynamic product and easy to sell. So the the key thing about RWAS is will they be willing to go through that process and will they be open to have this longer uh time to get a rating.

Uh because our approach is we can have a simplified version that is very fast and based on our internal knowledge and then if they want they can do the the proper rating process. Heat. [music] Heat.