#DappCon22 : Day 1 - Credit Scoring In Web3, Kerman Kohli and Kyle Quintal

Transcript

okay cool there we go sweet hey everyone today we're going to be talking about credit scoring in work three and uh yeah I'm Kerman Kyle yeah we got introduced exactly and how do you do this cool so the current state of D5 there's about 4.5 million uh well active wallets right now of which there's 60 billion dollars in TVO and about 100 million dollars in Daily liquidation so these sound like kind of all right numbers but relative to 2019 where maker had about 10 million in uh tvl like the industry is growing like a tremendous amount and when we started out it was how do we just get a basic Financial system that's predicated on Smart contracts and the idea was just to get

something that works so fast Road from 2018 2019 were at a much more mature place and what we've realized is kind of going through the data that we've kind of made some fundamental assumptions about how all this technology works that we haven't really Revisited and what we've spent a lot of the past two years doing now is really understanding this data and seeing going back to First principles of analyzing okay how do we actually make a much better system that what than what exists today so the current state of defy liquidations typically when we see a number go down it kind of gets bad in two ways because number goes down and then you have a bunch of liquidations

that happen and the liquidations cause number to go down more and the cycle repeats and when you have any of these D5 platforms such as compound maker or Ave all the collateral types that they on board they have to do like pretty in-depth liquidity modeling with the key assumption that every single person gets liquidated and it's a somewhat Fair assumption because in crypto we always assume adversarial thinking and everyone's here to do the absolute worst thing possible so uh that's the state right now Kyle do you want to add anything more to that no not really it's it's a very common theme right every year markets go up markets go down when Market goes down everyone's you know Whalers going off on Twitter these mini

wallets got liquidated this much of tvl is gone right and and everyone's treated the same um and I want to talk about a little bit uh uh an event that happened that you guys are all aware of in May um let's talk about let's go back let's talk about the May 2022 market crash right so we're talking about liquidations here um internally in our risk team we did some analysis on you know many different Market crashes this one particularly was interesting so we pulled data from three places um from Ave um on both ethereum and polygon and also compound so we took all of these um events there was around 18 000 liquidations and do I yeah I can sort of move the mouse here so as you guys see on this

everyone sees this even in the back cool so in this circle here um of everyone that got liquidated 19 of the liquidated address had a previous history of liquidations so what does that mean well that means that a fifth of all the users you know had a negative history and of those 20 almost half of them got liquidated again right this that wasn't their first time and so and then if you look at the opposite so the remaining 80 or 81 percent only you know 12 of those it was the first time being liquidated right but everyone is treated the same right everyone gets the same rates everyone gets the same LTV that's just the way

the market works right now right and so the conclusion here is well currently the menorah the minority is sending the rules for the majority there's a lot of behavior that exists on chain you know it's an open Ledger we can see it but we're not using it right there's a lot of value that's still kind of left on chain on many chains right and because of evm you have the you can use the same wallet and so you can you know even pass that information between chains so there's a lot that's left on the table and even a little caveat um I don't know about you guys but personally when I've been liquidated in the past I just creating your wallet right whatever new history start over so these numbers are probably a little

worse um but yeah so let's talk about why we're here the defect credit score right so there's many ways of building a default credit score um we've looked through you know using a big complicated ml model that ingests all this data and does all this you know heavy computation to spit out um a quantitative value of is this person credit worthy or not that works we tried that out for a bit then we realize well how do you explain to someone hey sorry you're getting these rates and they say well why and you say because the model sets up it's not you're not going to get people to embark on you know that journey of let's use the D5 credit score so we

wanted to think about that a little bit more and we started to move towards a more algorithmic mathematical approach so hey I have a score of 500 why because of this because of this because of this because of this that way we can communicate with the community and say hey what do we think should be included in a credit in a D5 credit score right let's talk about it let's make it um let's let's make it interactive let's change the values as we see fit you know maybe the market goes up the market goes down we can adapt it right it's not just a black box that we can play with so I talked about the D5 credit score there's five big components that we are looking at currently um I'm summarizing it them into five big

components let's put it that way okay so the first one is defy experience so pretty self-explanatory how much have you used Define the past um you know which protocols for how long so on history and survival that's an interesting one so when the markets go up and the markets go sideways everyone's a good borrower no problem right it's when the markets go down that that's when you see hey did you manage your position hey did you pay off your debts did you top up did you completely ignore it right that kind of thing so we essentially reward people who manage their position through a downturn in the market more than the people that don't have a position at all right so history and survival is essentially you get a better

score if you survive during downturns in the market liquidations while self-explanatory don't get liquidated simple um if the Market's going sideways and you get liquidated that's not good you shouldn't but hey some people would like to hit the max button when they borrow I don't know what to tell you okay um uh last will not last but data sources so lots of new chains popping up lots of new protocols popping up lots of tvls you know everywhere so we're trying to aggregate all the popular ones for now um and then build take all those transactions compute them and then create a score from that so data sources were you know pulling in more and more every month um so that's the fourth piece and the final piece is identity composition I don't know about you but me personally I

have multiple wallets for different things right some that are docked together some that are not um so we you want to be able to sign messages and say hey these are all part of the same identity they should all have the same credit score these are not blah blah blah that type of stuff so five big components to the D5 credit score great we have D5 credit score awesome what can we do with it right so the first thing that we decided to do with the D5 credit score is provide better Capital efficiency on our platform so what do I mean by Capital efficiency well better use of your Capital right if you currently go on compound you go on Ave I think you get 80 LTV the interest rates change depending on their their algorithms different things like that um so better borrowing on arcx that's

what we're doing now so we want to provide based on the credit score right a mapping of favorable borrow rates right hey you just starting that's fine you get Market rates right everyone you know the Market's been pretty good at you know saying if you're a new borrower or you're just coming into the space for the first time you get Market rates right everyone gets Market rates but if you've been in the market for a while maybe you get a little bit more favorable because we know we can trust that wallet address if this wasn't obvious it's D5 there's no kyc there's no AML there's no none of that it's crypto native second thing well um so better of borrowing rates is you know uh interest rates and that kind of stuff on your debt loan to value ratio

well you get more bang for your buck as you deposit collateral and just to emphasize we offer 100 LTV loans if you have a really good credit score so these are near basically one-to-one loans something that's never really been done in crypto before so it's not like we're trying to push this just a little we're pushing us to the point of just collateralized loans and once we get an even better understanding of like how to quantify the financial value of an identity we'll probably be able to do truly under collateralized loans scale via algorithms and no human intervention or risk assessment so but yes yeah exactly yeah exactly and and especially you know if you value your credit worthiness well there's that's inherent value right that you can leverage on top

of the collateral that you deposit so you know for me personally as a personal anecdote if I have a you know 999 credit score like I'm gonna do everything I can to protect that I get liquidated like I can't forge time right I can't go back and you know undo the last two years I can't redo the last two years of perfect borrowing right to get these better rates so yeah there's more stuff too you know boring against exotic assets and all that stuff if you guys get the idea so better borrowing is great but this inherently comes with lots of risk so let's talk about that yeah risk it's something that we get asked a lot about especially when creating a credit score because as we've seen with this entire bear Market uh a lot of people haven't managed their risk correctly so a lot of

what we do is first of all understanding how much leverage a user takes and looking at like the reasons why they've taken it the Great whenever I talk about credit scores people they say what if you gain the system but you can never gain the future prices which is the best fit about this so understanding a user's relationship to Leverage is something we spend a lot of time on the second really big bit is profit modeling so when we're offering one-to-one loans uh there's still a liquidation penalty of anywhere between three to five percent so we do make a loss on those liquidations but we've done our modeling in a way where we mathematically always make money so making sure that profit modeling uh bit is really important for us the third bit is around decentralization and

censorship resistance how do we make sure that the computation we do sticks to the like values and ideals of crypto and then real interest rates there's no ponzenomic Behavior there's no token that you earn there's no like back doors none of that stuff um and last of all it's just like always balancing between the incentives of over collateralized and under collateralization so there's a lot within risk that we've taken a lot of time to understand we spent about one and a half two years understanding this I've personally been in D5 since the birth of it in 2018 I've seen every platform every situation play out so it's we're not going into this with a lot of naivety um yeah call anything went out to that

no no no no no not done really I mean risk is risk isn't an Ever ongoing challenge right just when you think you know it all something new happens a new you know a new stable coin gets depegged who knows right like things are constantly evolving so this isn't a once and done your understand you're good it's a constant constant thing that we always want to want to work on improve yeah so just to summarize we've got liquidation parameters are set by the minority in D5 today we want to create a world in which everything is set via your credit score or personalized to you and your behavior and eventually this unlocks a lot more capital and credit for the entire ecosystem which allows us to better understand risk and of course uh there's a lot of technical challenges with enabling this on-chain identity

slash reputation layer and that's about it we want to call it here because we'd love to hear questions from you guys foreign [Applause] so in some protocols what you can do is you can um do a like um you can ask for money and you can you can deposit and then ask for zero right so for instance liquidy allows you to put the collateral and then ask for

as little as you want even zero but that means that the score will be high because now that will be impossible to liquid oh yeah and that means like if you have like you create 100 wallets you create 100 zero positions that will be never liquidated they will accumulate good race uh credit score over over the time is is it something that you consider when absolutely how the amount of money that that a user borrowed yeah yes all right so there's two uh solutions to this problem the first is there's an active component of the score which

looks at the debt utilization of the collateral leave deposited so if you're borrowing if you deposit a hundred dollars and then you're borrowing one dollar of debt your credit score is going to grow by a very small amount per day then to the second leg of your question of like okay what if I get a dollar and put it across 100 wallets your uh credit limit is only going to be set to about like three or four dollars because the system has learned that hey you're a small player you steal with small amounts of money what's a credit limit Kerman oh I don't know what this is so credit limit basically algorithmically look at the average ball amount that your wallet uses and we're only going to give you that much credit that you've actually been accustomed to handling it's just like a bank in the

real world if you're borrowing ten dollars from the bank and then you say hey give me a million dollars it's gonna be like no off like it's the same thing here yeah exactly like in in your example of you know if I take a hundred dollars right if if you don't borrow at all your score's not going to move period right like you're not using your credit so why would your credit score won't um and to kerman's point with the credit limits like that's how we alleviate with risk right if this in the same example you could have one person take 100 in one wallet manage it at like the perfect optimal borrow rate get to 999 and then put in millions of dollars in leverage like crazy well no we won't let them do that because you've only proven that you can you know

properly handle 100 so maybe we'll give you we'll give you more than a hundred dollars obviously but you won't be able to like Leverage out and take Millions so all right yes and one one simple question why are liquidations bad in your in your model where you penalize liquidations so I think it's actually the opposite with our model we try to encourage people not to get liquidated our business model relies on people not being liquidated because we believe in long-term games with long-term people a lot of if you look at Mac maker Ave compound a lot of their revenues are driven by their liquidation set by higher liquidation penalties so we try to minimize the liquidation penalties because we believe if we all the game of lending is I give

you something valuable and then you give me cash in return and the more we can play that game cooperatively together the more credit and value we can create for people over the long run so a lot of our ethos and philosophy is built around that way um and we only penalize liquidations because we want to reinforce people don't get liquidated manage your debt so that we can create more credit collectively okay thank you no problem cool any other questions there's no dumb questions we've heard all the all the questions possible like this you know on the on the point of of uh liquidations like there's there's when the when will the liquidation occur right like someone

wins right most of the time it's the Liquidator right they get a fee for maintaining the system and all that kind of stuff and we wanna like Herman said it's part of our ethos does not require hey it's me against you it's borrower versus lender it's who can extract the most value out of hue out of who right so that's why we're trying to say no instead of everyone just hating each other let's all kind of like cooperate a little bit figure out hey you're a trustworthy person you know so now the lenders can decide hey well I'd rather lend to a higher risk person and get you know take on more risk for more more value or no you know I don't want to lend to high risk people I want to lend to lower risk people and get a more stable lower risk

return on on my on my capital and yeah so just an extra little point on liquidations first encourage people to put more money system s it is I don't think I quite understand so you're saying Liquidators bring Liquidators bring liquidity [Music] well so

think I personally liquidations mostly are all flash loans so they don't actually bring in any liquid liquid liquidity is that what you mean because like like we have our own Liquidator and like different things like that that we use we don't actually hold any capital for the liquidator foreign of course yes and that's the incentive for him to maintain the system yeah [Music] they don't take Chris though because this is executed atomically via flash loan so in a single transaction they're repaying the debt selling the collateral

four dollars and are completely hedged so they don't actually take on risks these are typically very sophisticated uh Bots that are running and have like perfected this down to a science and raking millions of dollars when the market tanks and they're very good at front running too correct because we've learned um I think we had a question here yeah sure uh we take a cut on the interest rates primarily yeah sorry on the interest rate from the debt that you're effectively borrowing exactly and usually like we're still discussing with the community on you know and internally about the exact like numbers and the spreads and all that kind of stuff

um but you know the better the credit score then the better the rates right it's usually that's kind of how this works right like if if someone's building up their score and you know they've proven that they're very credit worthy then they're going to pay next to nothing right all right there's a question here uh are you looking and I see you guys I think they're in a similar sort of space um are you looking at like partnering with like Morpho and like compound and you know other Landing boring portfols protocols to to provide like your scoring mechanism to give users a better experience or basically how are you thinking about like your go to market like or are you kind of trying to build your own and then have your own kind of ecosystem because I just know you guys from the scoring aspect but it'd be cool

to see hear about how you're trying to like make it more accessible yes we've been built our own lending borrowing Market itself um and we're integrating it directly into that but we've also built this like very sophisticated data capability so if there's any other protocols that are down to collaborate um that's like definitely avenues that were Keen to explore as well but our primary go to market is through our own Landing borrowing protocol yeah hey so thank you for the talk uh I have one question you mentioned that newcomers to to the space get Market rates how do you consider uh bridging their like real world credit scores to offer them an advantage when training

it's kind of something that like a lot of other people are trying to do but in my view like anytime philosophically I believe the on-chain world is very different to the real world and as soon as you try and bridge between the two it creates a very sub-optimal experience because you're relying on trust guarantees of like some lawyer and like some random like us Delaware estate so uh that whole thing like it's just not that appealing there's a lot of companies trying to do stuff like that but ultimately I think it's more of as more things happen on chain that history will be quicker and better to build so uh yeah it's there's no perfect answer right now but like compromising to like saying Hey give us your iris scans and we'll give you better rates it's

something we're about sorry yeah yeah like you're gonna have to pick a side right it's like either you go the hey we're going to kyc we're going to get your you know real world stuff and try to bring it on chain we're going the route of no you don't need to tell us who you are we just need an evm address or an address an identifier of some kind right um which one's better time will tell thank you hello are you going to Cluster addresses because maybe at some point a an address imagine that I have an illness and at some point people will die or

in the sense that at some point I have a high risk Corp a high credit score and maybe the best way for me to get a big party in the end of my life will be to screw the system by asking a big loan sure and don't reimburse it because it's the end of my journey on this life yeah if you've done cluster addresses let's say if this is not link it somehow to other things maybe a scammer will be able in the end to make profit

of it I mean it's yeah that's the way because the long con essentially right how do you like like basically do a long con and then screw the system at the end so our solution to that is I think a lot of crypto native people like this idea of like civil resistance of like how do you prevent multiple wallets from like gaming a system or one wallet from gaming the system to the last point in time and our answer is like very rational we just care about the math we look at the profit we look at the potential losses and as long as the math Stacks up correctly sure like we know the delinquency rate we can model everything out and as long as our systems stay positive with a margin of error it doesn't really matter but we don't have any under collateralized loans at this point in time so that's

not really a risk that we run okay thanks everyone all right thank you [Applause]