Gauntlet

Transcript

foreign [Music] super good to meet you all I'm really excited to talk today at ECC about long-term growth incentives uh on uniswap and some of the work that we've been doing for the uniswap foundation over the past four months as well as how that might apply to your protocol so let's dive into it got a lot of slides to get through so just to introduce myself my name is Gabe I work at Gauntlet networks I'm currently leading up the applied research effort which is basically a goal to take the data science platform that Gauntlet has built originally to service risk protocols and expand the reach in the scope to be able to service all of defy and so my background as I previously worked in high frequency

trading at companies like Jane Street global trading systems optimizing traditional markets and my entire career in crypto for the past three years has been trying to take the learnings and the modeling skills from the traditional Finance market and apply that to the decentralized finance market and so the goal of my team is to help protocols be more efficient grow faster as well as save money and so that's what the talk today is going to be about is the work that we've done with uniswap and how that might apply to your protocol as well and so for the uninitiated what is unit swap I know we just had to talk about this but ultimately like uniswap is an exchange it enables people to buy and sell Assets in a very decentralized way using an automated Market maker uh mechanism which essentially allows liquidity providers to supply liquidity

on chain which then allows traders to use that liquidity in order to do transactions and so the real question from new Subs perspective is what does success look like if you're trying to run unit swap as a protocol what should you care about how do you judge the success of a protocol from first principles now the kpis that people mostly reference when we're talking about these types of issues like often come to be fairly simple people talk about trying to increase TBL increasing volume increasing having a ton of users or just making the token price go up and so when I talk to Founders about what they want their protocol to look like these are often the type of metrics that people bring up but in reality it's a little bit more complicated than that and I think we can take a lot of learnings from the traditional market

and think about what is a business really for what are we actually trying to achieve with the protocols that we build and so the two kpis that we've identified at Gauntlet and through working with uniswap is really there's two points here is one is high extractable Revenue the goal is how much money can we extract from their protocol and then redistribute to token holders or simply like enable us to accrue and then the second question is high governance value how can we make the protocol really important so that the tokens which control the protocol have a ton of value in incentivizing uh the rest of the community and making big decisions that impact the crypto community at Large and so given that we're from Gauntlet how do we quantify this uh we can make a little bit more mathy but it's really fairly simple on the extractable value

point it's a question of what is the present value of future cash flows that we can drive from a protocol if we're going to turn on a fee at some points how big is that fee going to be how much can we extract from this what kind of a flywheel can we build today that in the future is going to enable us to pull money out of the protocol and then redistribute that to token holders and the other side is the value of governance and so even if your protocol doesn't allow you to extract Revenue even if the goal of the protocol is simply to control something important the tokens still have value because it enables you to make big important decisions about your protocol which might impact the rest of the community and so what type of things go into governance well it's hard to perfectly

quantify but it's kind of more heuristic based so it's focused around things like Brands Azure protocol grows as it becomes larger how well known is it in the rider D5 community market share how important is it how much control do you have over all the other competing protocols that are doing the same thing and lastly just absolute size if you have a protocol that has a huge huge amount of money that processes a huge amount of transactions it's obviously going to give value to your token holder simply because they control something that's at the Baseline important and so when we're considering a new program this is the equation that we always think about and so to kind of simplify it it's this idea of how do you calculate net benefit for any protocol program that you might consider running if you're going to change your protocol

if you're going to run an incentives program if you're going to change your fee structure how do you think about that from a holistic view in terms of how it's going to affect your protocol and so it's really this idea of change in net benefit is going to be equal to the change in net present value of future Revenue plus the change in governance value minus the cost and what this really means and it's fairly simple is that if you're going to spend money if you're going to put effort into doing something whether that's through token incentives whether that's through hiring new developers it better be the case that that one-time effort is going to lead to either short-term or long-term gains for your protocol essentially Roi needs to be positive to justify any effort that you're going to do and so what is cost uh the real idea

here is that token spend is not free if you're going to be spending a dollar of uni tokens in order to try to run a new program if you're going to hire a new developer one dollar of spend is equal to one dollar of spend even if it's denominated in your protocol's native token and so from a dow perspective why why would this be so I can't just treat printing tokens infinitely well no like one dollar I've spent today means one dollar less spend tomorrow and that can be through two things either you have a Max cap on your token and you can no longer print infinitely or also it could mean inflation as you print One More Dollar of a token it means that there's more circulating Supply and the next token that you spend is going to be worth exactly one dollar less relative

to the rest of the tokens that currently are out there and so what we see is that you need to think about spend especially even when it's denominated in Native tokens as spend any dollar of spend that you're going to do it needs to have at least one dollar of impact to be justified and so to bring it back to uniswap we can sort of talk about this liquidity volume flywheel how do we build in long-term Network effects and long-term growth for uniswab and how do we think about that and so the network effect on unit swap although it's of course different for other protocols really focuses on this flywheel between liquidity providers and Traders as the main users that exist on the uniswap protocol and so how does this work well LPS deposit capital and so the reason they do that is through additional LP

returns uh through that increased Capital there's like higher liquidity as a result of the higher liquidity you have decreased slippage for Traders decreased lipids for Traders result in increased volume increased volume means increased fees increased fees I mean more LP returns more LP returns mean more liquidity and so through this flywheel what we're seeing creating a long-term growth in creating a baseline of activity on uniswap today will perpetuate to long-term growth in a baseline of activity on uniswap tomorrow and so it's really important not just for uniswap but for all protocols to ensure that you're creating this long-term Network effects because given the open source nature of the majority of the work that we do it's important that Network effect is what keeps the activity in the community on your protocol rather than potentially

allowing other protocols to steal it and so diving a little bit deeper how do we think about unit swap users what do they actually care about and so there's two users that are primarily important for the uniswap ecosystem and liquidity providers and there we can kind of measure that indirectly through tvl how much liquidity is being provided on uniswap pools and so they care about three things really first of all is LP returns if you're going to be depositing Capital into a uni pool what is the returns that you're earning in in return for that capital and so you can think of that as fees as well as liquidity mining incentives so how much money are you being given in order to take on the risk of taking on this LP position but you can also think of other things there's also risk certain Auto pools are created

equal certain pools enable more risk for LPS than others things that new swap can do which reduce risk for LPS is naturally going to increase the overall expected return for them and the lastly is operational cost what can you to swap do to make it cheaper and easier for LPS to access the protocol less overhead less Staffing all of this makes it easier for LPS to add liquidity at a very low price and secondly is Traders and the kpi here really is a volume and here the idea is thinking about price execution how do we lower slippage how do we lower fees that Traders are paying and in doing so increase the amount of overall volume that we experience on the protocol and then again we have to think about operational cost how expensive is it for Traders to use this how can we provide

Traders better UI make it easier for them to access the protocol and in doing so increase long-term volume and so in terms of how do we increase long-term growth uh it's really just using these kpis in order to drive future growth to uniswap as a protocol uh but getting better price execution for Traders is going to increase overall long-term volume lowering operational costs will do the same and for liquidity providers better returns lower risk lower operational costs all of these things together are going to unable to increase this flywheel where we increase volume as in doing so increase LP returns and in doing so increase volume yet again this long-term flywheel drives long-term growth through the unit swap protocol and so any program that we want to run necessarily needs to be adding to this flywheel and adding to the

long-term momentum that we have and so how do we think about running a program well there's sort of two sides at which we can attack this one is on the trader side and one is on the liquidity provider's side and so what we're going to talk about today is how Gauntlet considered both sides of this uh this program whether we want to incentivize on the lp side or whether we prefer to LP on the trader side and talk about some of the research that we performed in order to make that decision and hopefully through that it can inform you about the type of rigor that's necessary to make a strong decision as well as help understand that sometimes things that seem obvious on their face and seem like good programs on their face in reality you're just going to be a waste of money and not drive the

long-term impact that we expect and so to kick it off let's start out talking about trade Mining and so first of all what is trade mining the idea of trade mining is essentially like let's subsidize traders to use uniswap and so in doing so what are we targeting well we're increasing price execution uh and in doing so we expect to see an increase in volume and so what would this type of program look like well what you could imagine is topex Traders on uniswap earn some amount of uni token as a reward you could imagine uh referral programs if you sign up for unit swap for your first time with a new address you're going to earn uni token as a reward for a limited time only get all your LP fees uh rewarded so that you're not paying fees to use uniswap help onboard new Traders

and make it easier for people to use the protocol and this is actually something that has a ton of parallels in the traditional Tech Market you can really equate this to things like uber providing uh sign up bonuses for their new users you can imagine this as credit card companies providing reward points for using their credit card all of these things Drive user growth and enable users to be more active on the protocol and drive long-term value and network effects and so there's really two things that are necessary to think about when thinking about a Trader trading Rewards program on uniswap or really on any decks or protocol and that's civil resistance as well as sort of the existence of sticky Traders and sticky users on the protocol to begin with and so let's talk about that so why is civil resistance so important and what even is civil resistance well

what it really means is that you're creating new wallets you shouldn't be able to earn a profit and so we can look at the example of looks rare which in January of 2022 provided a trade mining program using looks token on top of their nft trading platform and so what happens well for every trade that you did on luxury you paid a certain amount in fees you would earn a certain amount of looks but what happened is you would earn more in looks than you would actually earn than you actually pay in fees for doing an individual Swap and so that's interesting so what happened well as you might expect a bunch of Mev Bots set up new wallets increase their trading activity massively drained all of the pools of looks that were being provided and so yes you do get a massive spike in volume back when this program

was first launched but as soon as the program was turned off all of that trading turned off none of the traders that existed were actually wanting to use the protocol they were simply there to soak up the incentives this in no way drove long-term flywheel effect or long-term value for looks where it was simply a waste of money to generate some vanity metrics on a very short time scale and so we need to have a civil resistant program if we're going to run this we need to come up with a mechanism that ensures that when we're providing incentives to various Traders we're not potentially opening up the opportunity for people to build new wallets and simply soak up those Traders without adding long-term value and so what would this potentially look like well we could onboard new Traders using a civil resistant trade mining program and so what what does that sort

of imply what does that necessarily need to include well what that means is that volume today needs to be get volume tomorrow if you onboard a new Trader to the uniswap ecosystem uh that means that that Trader is long-term going to prefer to use uniswap over other exchanges even without considering other effects uh such as things like slippage um it also means we have to do something civil resistance whether that's like kycing wallets to allow referral programs whether that means providing less incentives in the cost and simply subsidizing flow it requires very very careful mechanism design when we're actually going to be building out that program if that's what we want to go with and so the question is then obviously are Traders sticky if we provide incentives to Traders and onboard new traders to the unit swap ecosystem do we expect this to long-term Drive revenue

and value to unispot because those Traders are going to stick around and so that's the question that we wanted to answer if you're going to swap to decide whether or not a trading a trade mining program made sense is something to pursue and so in order to answer this we built out a simulation which is sort of the bread and butter of Gauntlet so we have a simulated representation of uniswap V3 and then we simulate all of the orders that transact on major dexes on mainnet so we had about 70 billion dollars of order flow that existed across all of the exchanges on mainnet uh over six months across about 10 pairs and so we looked at exchanges of unit spot V3 Sushi swap balancer and curve and the question was basically if we try to Route the volume that was going to go to sushi swap or curve or balancer and

see what would happen if we instead routed that volume to uniswap would that volume been price improved and so why is this an interesting question well the hypothesis here is is there a is there a network effect for Traders preferring to use other protocols even if their price execution is worse if you find that Traders are preferring to trade on Sushi swap balancer curve despite the fact that they could achieve a better return and better trade price by executing in uniswap that provide really strong evidence that Traders can be enticed to continue using a single platform even uh just even if there is no long-term incentive program to continue enticing them to do this so it's a question of like can we onboard traders to drive long-term flywheel effects and so the first step here after building the simulation was calibration

and so through this what we did is we took all of the volume that actually was routed to uniswot V3 pretended rerouted it to unit swap E3 and the goal was basically to check whether or not the volume was in fact accurate if we had a perfectly calibrated simulation you would expect that you would get Perfect Price execution every every uh amount of slippage and fees that were paid on uniswap V3 and the simulation should be one to one with what happened in reality and in general we found that the simulation is fairly well calibrated with about 90 of Trades being within one bit of the actual price execution and the reason for this light discrepancy is due to complicated trades and swaps that are mostly related to things like Mev so you can imagine that having very complex order trades where liquidity moves around is going to cause price execution

to change and so what we found is that the simulation is quite accurate at predicting this but there is some error tolerance as well so when we rerouted the orders from the competing exchanges Sushi swap curve balancer we found that there's some but not that much volume that actually could be profitably rerouted to uniswap E3 about 703 million dollars across the 70 billion dollars of trading volume that we observed across the trading period so this isn't that much volume if you really think about it well it seems like a big number it's really about one percent of the overall volume that we observed and so if we assume that this is all traders who prefer Sushi swap curve or balancers UI just prefer the brand of these other dexes we could potentially onboard those Traders onto

unit Swap and increase long-term flows by about one percent assuming that these Traders are actually sticky in the first place and so that's interesting but the second thing we did is then look at this volume by cohort and so what we did is we grouped together the different Traders by their total trading volume across that entire trading period of the six months uh with anyone trading more than a million dollars being classified as a whale anyone over a thousand dollars in a million dollars as heavy Traders anyone less than a thousand dollars is retail and then we also use an Anson datasets to tag any addresses that were tagged previously as using Mev as Mev Bots and this is particularly important because me v-bots are often doing very complicated transaction orders that may on their face look like they were misrouted but are actually taking

advantage of some complicated Arbitrage that otherwise might be missed and what we found is that while the vast majority of Traders and active users on uniswap are in fact retail and heavy users so we see that by far around 90 plus percent of users on uniswap are these smaller market cap Traders the vast majority of trading volume actually comes from like whale Traders as well as Mev Traders and so this is interesting of itself right if we're if the idea here is that retail Traders are preferring to use uniswap's interface and that's why either going to be staying on the protocol it's unlikely that whale Traders and Mev Traders are making that mistake they're likely routing accurately but just in a complicated way that appears as poor order execution and so if we look at do the same

analysis but group it by cohort we actually find that about 92 percent of the misrouted volume on uniswap is actually coming from Wales or Mev Traders and so what that means is that there's actually not that much volume uh which is coming from smaller traders that we believe we can onboard to uniswap through things like a trading rewards program and so the real conclusion here is that Traders just really aren't that sticky there's really like less than two percent of Trades that seem misrouted in the first place and then of that volume that's misrouted less than 0.1 percent of overall volume on uniswap appears to be misrouted when it comes from non-mv or very large cap Traders and so what this means is that the programs that are onboarding Traders onto a deck so better referral programs having a ton of money into better uis really isn't going to

move the needle that much when it comes to trying to create long-term Network effects in flywheel it might be pretty intuitive but Traders are going to route where their price execution is the best and they actually seem very very efficient at making that happen for themselves and so if we're not going to do a trade mining program the other question is liquidity mining if we can't uh if we can't attack the trader side of the equation can we maybe attack the lp side of the equation and so brief history of liquidity mining we can talk about sort of the sushi swap vampire attack on uniswa back in 2020. uh here's a market share graph of wbtc eth volume across those exchanges and what we see here is that Sushi swap was actually quite effective at using liquidity mining in order to steal market share uh back when it happened

they were able to get to about 40 market share as compared to unit swap V2 although then of course you have unit swap B3 come in which has an order of magnitude more efficiency in terms of capital efficiency and then goes back down to where uniswap again dominates the volume as we see today and so why does this happen why did this work why when Sushi swap adds a bunch of decks incentives are they able to pull in the liquidity and generate this new flywheel for themselves that enables them to at least for a short period of time compete with uniswap and so the question then is like is liquidity mining sticky as you onboard liquidity miners are they going to stick around and continue to use the protocol after the incentive program has ended so this is the question that we wanted to answer as well

and so the the question is like the mechanism that people often like to talk about with sticky liquidity mining is this idea of like forgetful LPS uh when I talked to a lot of protocols about liquidity mining the idea is really okay we're gonna have this big incentives program a bunch of LPS are going to move Capital onto my decks or onto my protocol and then after I end the protocol they're effectively just going to forget that they're no longer earning these incentives they're going to forget that they're no longer achieving High returns but reality is LPS aren't as stupid as we think they are we did an analysis where we looked at all of the pool are at these 10 pools um and look at the average days that LPS would rebalance their entire portfolio and while it of course varies very very strongly pull to pull we actually find

that LPS are fairly efficient at rebalancing their their liquidity we see that on the weft pool that's actually happening once per day LPS seem to be very very fast at identifying when returns change and rebalancing their position uh both in and out of the pool to take advantage of returns that are better in the rest of the market and that goes all the way down to one month and so we still see that LPS might be a little bit inefficient it may take a little bit of time to rebalance but we're talking about on the scale of weeks to months we're not talking about at a scale of years and so if your goal of running an incentives program for LPS is simply to attract a bunch of liquidity and then not attract mercenary capital and hope that the liquidity sticks around the reality is this is

quite difficult we don't really think that LPS are going to stay around because LPS are investing large amounts of money they're smart and they're going to move their Capital as necessary and so the real mechanism that enables liquidity mining to work and enables this flywheel to take effect is liquidity bootstrapping and so this is something that people haven't really talked about before and is something that we've been talking about quite a bit at Gauntlet as well as with the uniswap foundation and the idea here is that you don't want to have incentive programs running forever the idea is not that you're going to continue providing excessive returns in infinitum the idea is rather that you're going to do a one-time incentive liquidity mining program that's going to create larger amounts of liquidity for LPS at a single given time that liquidity is then going to attract volume that volume generates

fees for LPS the fees for LPS increase natural yield for those LPS on the platform the natural yield keeps the liquidity there at which point you can then reduce the incentives over time and through that you create this actual natural flywheel and so the idea is not we want to run a forever liquidity mining program where we think this is going to be necessary forever or we run a one-time and hope that LPS are going to forget about the returns but rather that we can spin up this flywheel between liquidity and fees on a one-time basis and then have that flywheel persist over time and so how do we think about liquidity mining as a problem well if we're going to run this type of a program what we really need to do is maximize future protocol we have a new net token spends and so there's three main parameters

that we really need to tune here so this is like the type of pools that we want to attack uh which pools do we want to incentivize how much money are we going to provide to each of those pools uh how long do we want to incentivize all of that and the goal here is basically determine what parameters do we think that we can set so how much are we going to spend on a purple basis that have a long-term expected positive Roi in terms of future extractable governance value as well as future extractable revenue for a unit swap protocol and so what we're doing right now is we're currently working on building out a statistical model for uniswap that enables us to predict exactly this we want to build a model that enables us to predict if we do run a one-time liquidity mining program on pools that

we think are valuable what is the expected long-term Roi that we can generate from this and so how do we how do we go about modeling liquidity on uniswap it's quite complicated right and so you can think about this as sort of the microeconomic view of liquidity on uniswap it's very complicated it's incredibly fragmented LPS can put very arbitrary positions on uniswap LPS can rebalance on a quite frequent basis you have very complicated interactions like just in time liquidity they make it very difficult on an intraday level basis to model these things out but if we zoom out a little bit the macroeconomic picture is actually relatively simple um tvl in terms of absolute numbers is fairly stable on this wbtc pool despite the fact that tvl moves around quite often uh within the pull itself and so if we zoom out rather than trying

to model on a very micro level basis we should be able to model this as a statistical problem on a macro level basis and so to formalize this a little bit more liquidity modeling is ultimately a Time series problem what you're trying to do is you're trying to create a function which given liquidity for a given pool today as well as an array of other features that might be important and necessary for predicting liquidity what is liquidity going to be tomorrow and why is this important well if I can tell you by adding an incentives program we're going to add a certain amount of liquidity today and then I can predict how much the liquidity today as well as future returns tomorrow will like lead into future liquidity this enables me to very accurately predict what we can expect when we increase incentives or increase other types of variables that

affect LPS and so in reality what we're actually trying to model here is slippage so while liquidity is interesting the real thing that we care about is how much Traders are paying to interact on these pools while it's great to focus on tvl as a vanity metric what we actually care about is what the liquidity is being used for is the liquidity being put for work and is it adding value to the traders that end up using it and so of course slippage is going to include liquidity numbers when we included but really what we're trying to build here is a slippage model of if we change parameters today in the uniswap protocol in the future how much is that going to then affect how much it costs for Traders to trade on uniswap and so if we zoom out a little bit here we can model

this as a sort of linear algebra problem across all pools so if we can build a slippage model that effectively allows us to predict slippage today along with other variables that are important so features that might add value here are price volatility LP returns through fees and or liquidity mining market-wide volumes you can create a prediction a predictive model using estimate standard ml techniques that essentially allow us to predict what is slippage going to look like tomorrow based on the current conditions of uniswap pools today and so on the other side of this how do we then think about the volume side of the equation and again volume on a micro level is very complicated this is a etherscan tag of a very complicated Arbitrage attack that was done on uniswap versus Sushi swap it's very difficult to predict on a micro level

what Traders are going to do on any given basis Traders are going to trade with the markets there's going to be missive spikes it's very hard to tell you what things are going to look like on an intraday basis or how much volume you're going to get but on a macro level again we can see that there do seem to be fairly common Trends in the volume markets that we can observe and so while yes there is intraday or daily volatility in terms of the volume you can see that the volume for this BTC pull is fairly well bounded you almost always have volume between around 10 and 20 million dollars and so again you should be able to build a Time series model that allows you to predict what is volume going to look like over time on uniswap and of course understanding volume and

building this model under requires you to understand the entire Market you can't simply have a simple heuristic model that's going to predict this the relationships between markets are incredibly complex if you decrease slippage for a single pool and then volume spikes to the volume Spike because you reduce slippage or did the volume Spike because the volume spiked across the entire market so I currently what we're trying to build is a very strong predictive model that allows you to denoise all of the market trends that are happening through all of the data and in doing so simply isolate the effects that a single liquidity mining program or other changes to LP pools is going to cause uh towards the future volume of uniswap and so again we can sort of zoom out and sort of model volume across all the pools using this Matrix form notation

and sort of add in additional features to help denoise out other things and so what does that look like well it looks like adding in Market wide volume to try to understand okay is it coming from uniswap is it coming across everything like what are isolating effects volatility metrics we'd expect that volume is going to spike during High volatility periods so it's also an important feature to have in this model and also like user level statistics what are the what are the components of the users that are actually trading on these pools if a single user does a large transaction and take makes up most of the volume that happened over the last week the expectation of whether or not that volume or persist is likely very different than whether or not that volume is coming from a million users or a thousand users and so all of these are

features that we're building into our model to help understand this and so to simplify a little bit we can kind of have this Grand unified theory of uniswap which is essentially a function that calculates any expected volume for a given pool across all of uniswap given the expected slippage across all of these pools in the future as well as the expected as well as other market conditions that are going to impact volume on unit Swap and so through this we can do is we can see how changes to parameters today whether that's a change to fees whether that's a change to liquidity mining things that uniswap can do that will affect users today we can use this to then predict how is that going to then go affect users tomorrow and in doing so really understand how to optimize and decide what requires our attention what requires our spend and

avoid wasting money or making mistakes from the get-go and so how how do we operationalize this so this is the sort of platform that we've been building out and how we intend to do this and it really functions very similar to a PID controller the idea is essentially that we pull on chain data for uniswap ingest that internally and then use that to train our two elasticity models the liquidity impact model as well as liquidity volume model using those two models you can then use a standard convex optimization techniques to allocate Capital across all of these poles whether that's liquidity mining whether that's fees there's a lot of different things you could do with this model and then through this you can then run uh recommendations you can then do Dynamic on-chain optimization you can change things based on the expectation of the model that's going to result in

the best possible expected value for a uniswap protocol and then once you've made the change it's really important to follow up it's not just a one-off change it's not just a one-off analysis once we've made a change we then have this question of did the change track with what we expected our model predicted that the liquidity was going to increase in this way that slippage is going to change in this way did the reality track with the model expectation if yes great we're going in the right direction we're currently tuning a uniswap as a protocol if no it means that we need to reevaluate our model that data is going to feed into the model and provide future training data to make the model get better in the future as well as provides feedback for our internal data scientists to continue thinking of ways to iterate on and improve the model to

make it more accurate and so what does this look like in practice well so the current iteration of the model what we essentially do is plug this into a simulation and so this is how we're setting liquidity mining on uniswap today and the question is essentially given our slippage and liquidity models if we run a one-off uh if we want to if we run a one-off liquidity mining program for a given pool during the program what do we expect is going to happen and so the simulation was run assuming a 30-day liquidity mining program for these pools and then after the liquidity mining program ends assuming that the Traders are going to continue assuming that the LPS are going to continue using the returns that were generated from the fees even after the incentives are going to be removed how do we expect both liquidity and volume to change after the

program is going to end and so regardless you expect there to be some impact but the question is really Roi once you have a predictive model that tells you what is going to happen after you run this program you're now unable to decide is this program even worth running in the first place should we provide the incentives to do this is the expected long-term value to uniswap higher than the expected spend that we're currently running and so what we've done today is we've currently launched this program so as of last Friday we're doing a very limited very small run on these eight pools on uniswap on mainnet and the goal of this is to explore how this is going to uh work so we have a model that works today we believe the model will be fairly

powerful and generate positive expectations we have predictions of how these pools are going to perform and like what the returns these will drive long term for uniswap But ultimately we need to run small-scale experimentation we need to try out a very small set see how well calibrated our models are prove out this thesis that liquidity bootstrapping can be effective at driving long-term value to uniswap uh and that's effectively what we're doing today we're doing a small scale program with the goal of then future iterations on this generating data for ourselves improving out this thesis so this can be a long-term and impactful thing for you to swap to work on and so there's really sort of three good takeaways just come from this and the first one is that choosing what program to run the mechanism that you're going to use to provide incentives and create

long-term growth in your protocol it is very difficult a trade mining program makes a lot of sense on its face if you're going to do that type of a program there's a ton of parallel traditional Finance to pull from it's not a crazy thing to think of but until you do deep analysis on your users you're not going to realize that it doesn't actually make sense because of the behavior of your users that you're observing and secondly is once you run a program it's not just a one and done you can't simply release the program into the world and let it run wild especially when you're spending tens of millions of dollars as we've seen many protocols do on liquidity mining or other incentives programs it's important that you're constantly iterating on that you're constantly readjusting your assertions and you're constantly monitoring the impact that the program is actually

generating and if the program is ultimately not being successful you need to start winding it down the goal is positive Roi if you're not generating positive Roi you shouldn't be wasting the money you're only uh killing your protocol in the long term if you cannot always achieve positive Roi on your span um and last time optimizing a program means means monitoring uh you need to have really strong dashboards you need to have models that are predicting whether or not things are going well you need to be monitoring your impact at all times and if your impact is negative it means you need to start descaling the program a lot of programs a lot of protocols like to go big at first they like to spend a ton of money make a big splash get a lot of traction on Twitter but ultimately the goal is really to

drive long-term growth we're playing the long game here we're trying to drive long-term Network effects and long-term growth for protocols the goal is not to Simply run the biggest program that you can at first to pick a big splash the goal is to run a very scientific program where you're saving money in the long run and generating long-term growth for your protocol and ultimately this is what we do at Gauntlet applied research very excited to be partnering with uniswap to start out this pilot program but also really excited to see how we can apply this type of work to other protocols as well and so lastly if you want to get in touch this is my contact information I'll also take a I'll be hanging out outside the venue as well if anyone wants to chat but please send me a DM I'd love to chat about this but either

the work they're waiting for uniswap or how we could do similar types of work for your protocol as well thanks a lot guys awesome thank you Gabby [Applause] yeah we have some minutes left for questions yeah rhetoric awesome thank you so much um so we here in the community are always discussing how to model things and a lot of attention and focus is on agent-based modeling and so I just want to point out that you've given an excellent demonstration of um uh opportunities that we have to model at the aggregate level and apply typical statistical methods to gain Insight that can allow us to make decisions and you

actually said that trying to model at the micro level is you know it's too complicated and and you'll miss where the information's at so uh thank you so much for giving a clear demonstration of that yeah thanks a lot congratulations for the presentation did you study the the incentive of curve and in concrete the All The Comebacks and all the other things how the impacting the liquidity of all stable coins yes we've done sort of light up lights analysis of other other coins as well um

I don't want to name names but in general I think the ROI is quite low for a lot of these programs a lot of Protocols are spending tens or even hundreds of millions of dollars and yeah maybe they're generating some liquidity generating some volume but ultimately like the long-term growth and extractable value is much much less than the incentives that they're spending um and so I don't have like particular numbers on curve or convex or other things like that um but would definitely be super excited to potentially partner with those types of protocols in the future to do this analysis I think one sort of key takeaway from a lot of this is that this work isn't trivial you can't just hire one guy with an Excel spreadsheet and sort of like call that a day if you're going to be spending tens of millions of

dollars on a growth program if you're going to be staking the long-term value and viability of your protocol on this massive thing that you're going to be doing that can't just be like a one-off dashboard it can't just be one person doing a one-off analysis it needs to be an ongoing and very deep understanding and optimization of your protocol and so that's sort of our thesis with gauntlet is that we we don't want to just provide sort of like one-off metrics we want to provide really deep understanding of the protocol and in doing so help generate long-term value you all right any more questions we still have some minutes left a very quick question on the analysis of a successful project versus not successful one what is your time frame that you look for success

yeah for sure I mean I think I think it differs um so for for the work we did for uniswap we used one year as our as our sort of metric of success is basically uh our simulation model the question is basically if we can increase volume by an amount that the extractable value over one year through things like protocol fees is greater than the spend to do that then this is a program that's worth pursuing now of course like when your time Horizon isn't actually perfect right like you know so it's been around uh for around four or five years right now hopefully it's going to be around for 100 years after that so really like your your time Horizon should be longer than that but ultimately we think it's very important that you can achieve positive Roi relatively quickly especially given that you're going to be

spending large amounts of money the market moves very quickly things change very fast as we saw with sort of the V2 to V3 transition where V3 just suddenly took over market share and so if you're going to be spinning up a flywheel it's important on the long term but it's also important that your Roi is really positive on a shorter term basis and so we traditionally use sort of one year as a benchmark but definitely a lot of options to go with depending on your sort of growth strategy for your particular protocol yeah another question over there yeah uh yeah I think one part of your thesis was that liquidity and users are not sticky after the after the incentive program and I've been reading a lot of like uh analysis around how LPS are facing like net Mark out lvr losses and IL losses

but still uni swap seems to have like a lot of passive liquidity being provided in in general so do you have any insights on how this stickiness of liquidity and LP losses they correlate with each other and why peopleps are still sticking around in the current mm paradigms yeah I mean I think that's a great point right there's a ton of I think you're talking about the sort of lost versus rebalancing paper which is an excellent paper um and I think it's a really interesting point I think that we are seeing in some cases that LPS do add liquidity like potentially in an inefficient way right you're potentially getting negative returns as what you would get by just simply holding the tokens or putting them in other sources of yield and so short term this is good for protocols right if you can model out this behavior

and understand that some LPS are more sticky than others and this is something that we consider um then you should Target those types of LPS if you can onboard more sticky LPS and sort of trust that that's going to stay around as a long-term basis it's definitely something that you should include in your modeling but long term I think that simply relying on that as your protocol is is really like not a winning strategy in a very long on a longer term time basis since I think what you expect is that if an LP is losing a certain percent like yes they might be forgetful for now they might not realize that they're making an efficient decision but long term they will long term if you have millions of dollars that you're responsible for you're going to realize that you're making a sub-optimal decision either

you're going to realize that or someone's going to come to you and inform you about that and so we I think that it's very essential to think of sustainable ways to drive liquidity and drive growth on uniswap as well as other protocols rather than simply essentially dumping on retail and assuming that retail LPS are going to make sub-optimal decisions uh forever thank you awesome thank you so much everyone just in case um you haven't noticed we are Distributing nfts for the attendees of this event it's the first token engineering track at UCC at a large ethereum conference and you'll attend these anti-academy students and nft holders have a say in distributing grants in the token engineering

community we want to have your point of view your perspectives for developing stuff like this incentive optimization for other initiatives in the token engineering space so make sure to Mint your nft you can find the Flyers over there there's a QR code and there you can claim thanks Gabi for this talk really exciting exciting to see how this work evolves thanks for coming today [Applause] and we're getting ready for in a couple of minutes with talking on