In this episode I speak with Grug, an anonymous MEV searcher on the Ethereum blockchain. If that sentence made no sense to you, I promise this will be a fun episode.

To begin the conversation, Grug explains the basic architecture of the Ethereum blockchain and how its structure allows for the emergence of MEV strategies like sandwich attacks, arbitrage, and liquidations.

He discusses some of the criteria he looks for when identifying a profitable MEV strategy and provides examples of some of the long-tail approaches he has deployed in the past as well as some of the risks associated with them. We discuss the pro-cyclical nature of some of these strategies, the role of retail flow, and the edge in being able to deploy rapidly.

Grug also provides his thoughts on the impact of alt-L1’s and L2’s on MEV, airdrop strategies, and the end game of MEV if Ethereum infrastructure becomes too centralized.

Please enjoy my conversation with Grug.


Corey Hoffstein  00:00

321 Let’s jam. Hello and welcome everyone. I’m Corey Hoffstein. And this is flirting with models the podcast that pulls back the curtain to discover the human factor behind the quantitative strategy.

Narrator  00:19

Corey Corey Hoffstein Is the co founder and chief investment officer of new found research due to industry regulations he will not discuss any of new found researches funds on this podcast all opinions expressed by podcast participants are solely their own opinion and do not reflect the opinion of new found research. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of newfound research may maintain positions and securities discussed in this podcast for more information is it think

Corey Hoffstein  00:50

In this episode, I speak with Grug, an anonymous MEV searcher on the Ethereum blockchain. If that sentence made no sense to you, I promise this will be a fun episode. To begin the conversation. Grug explains the basic architecture of the Ethereum blockchain, and how its structure allows for the emergence of any V strategies like sandwich attacks, arbitrage, and liquidations. He discusses some of the criteria he looks for when identifying a profitable MEV strategy, and provides examples of some of the long tail approaches he’s deployed in the past, as well as the risks associated with them. We discussed the pros cyclical nature of some of these strategies, the role of retail flow, and the edge and being able to deploy rapidly. Grug also provides his thoughts on the impact of Alt L ones and l twos on MeV, airdrops strategies, and the end game of MeV. If Etherium infrastructure becomes too centralized, please enjoy my conversation with Grug. Grug, welcome to the show, excited to have you here, we are just going to do something a little different than we’ve done in the past, we’re going to dive into the world of MeV, which is I think, for anyone coming from a traditional finance background, hopefully going to be a pretty exciting and unique look at strategies in crypto that maybe they’ve never heard of. So I appreciate you joining and being willing to do this. I know there’s a meme out there and crypto Twitter about you know, Grug teach me MeV. So I’m glad you’re actually coming on. And maybe you can teach me MEV here.

Grug  02:28

Thanks so much for having me. Yeah, hopefully I can be of some assistance in that regard.

Corey Hoffstein  02:32

So I know we want to be careful about providing too much information. Your anonymity is important to you. So I do want to maybe start with a little bit of background, what you’re willing to share about how you got into crypto and MeV. Specifically?

Grug  02:47

Yeah, absolutely. So my journey into crypto kind of started in 2017. But back then I was more just dabbling and behaving like a very unsophisticated market participant. I was speculating on all sorts of shit coins on binance. And things like saya coin and other crap like that. It was very much a hobby back then. And I pretty much just lost money. And after the market crash, I didn’t really look at crypto again. For several years, I worked in what I guess people in crypto today would call web two, specifically in the machine learning space. But I ended up pretty burnt out on the tech industry, and was just doing some boring web dev contracting stuff in the interim for a while before I ended up going full bore into crypto and nmbd after the fact. So yeah, after the big crash in 2017. And abandoned in crypto, I determined I was a pretty bad speculator. And to be completely honest, I really still am. Although maybe a little bit better than before. I did have a decently sized eath bag I bought near the top back then, but was too stubborn to sell that I ended up holding all the way down and back up again. So I guess I managed to squeeze a little bit of upside out of that. But when I really got into crypto though, was near the tail end of defi summer, somewhere by happenstance, I was just browsing Twitter. And I stumbled across a tweet by some guy who had made a killing farming this Wi Fi token. And that kind of sparked my initial interest. So I decided I would give it another try. But this time I throw myself at it in an uncompromising manner. So I made this my grog Twitter account, because that’s kind of where I intuited that everyone who actually knew anything about crypto hung out and operated on and I started aggressively using defy products like uniswap and shitposting on Twitter and finding my way into discord servers with like minded folks. Yeah, so I was determined to lock myself away in my attic and make life changing money along the way, like so many of the people that I had observed on Twitter had to defi summer previously, and my center of attention was primarily on communities with a focus on development, and where most of the members were engineers. I figured since I had an engineering background, surrounded myself with others who did too, or were at least decently technical, so to speak, would be the best route to quickly hone my skills. And foreign above the most impactful community, I found my way into was FEC, which is, of course, where we initially met, and FEC was chock full of a bunch of super smart and terminally online individuals. And it seemed like just the place that I was looking for. And for a very long time, my daily routine basically consisted of waking up, hanging out in FEC all day and drinking from the firehose of information there. So yeah, a lot of the FPC alums actually also went on to start protocols or businesses in the crypto space that ended up being super successful, I would say probably the, the best example, there would be gem, which was an NFT aggregator that I invested in, and that very close friend of mine founded, that sold to open sea in 2022. So that was a bit of how I stumbled into crypto. And relatively early on during my time in FEC, I ended up reading an article I think, was on coin Telegraph that someone posted about MeV. And it caught my interest. Prior to this, I just been toying around learning about structured products and the ribbon finance server on options trading. And I was pretty unfocused, I distinctly remember the moment that really sparked my MTV journey, being watching one of Robert Miller, who is steward over at five bots, one of his early Twitch streams, which was kind of just a walkthrough of a GitHub repository on a simple arbitrage bot. And after that, Robert and I began chatting about all sorts of things MeV. And he sort of took me under his wing. And this is pretty much where I started to more exclusively focus on MeV, I would say, the zero sum aspect, and it kind of been a explicitly player versus player game was what was very attractive to me. You know, I used to play a lot of chess, and I’ve always been hyper competitive. So I guess it sort of makes sense that that’s where I ended up within the crypto sphere.

Corey Hoffstein  07:20

One of the things that I make an assumption with with this podcast is that listeners have a general finance background, we don’t go into standard definitions, I assume they sort of have that sort of CFA Level masters and finance type knowledge. But my guess is there’s a lot of people who are listening right now that come from a traditional finance background, whose head might be swimming a bit, right? You’re throwing out things like defy some or Wi Fi, token uniswap Flash bots, all of which they may have never heard of, and have no context for. So let me start by saying to those listeners, hold on, we’re going to get there by the end of this conversation, you’re going to know what all that means. But with that in mind, Grug, I’m going to ask you to do a little bit of table setting if you can for me, because I think for the rest of this conversation, it makes sense. We have to sort of start from basics. And so I want to start with the most basic question, which is at a high level, can you explain how the Ethereum blockchain works?

Grug  08:18

Yeah, absolutely. It’s probably a pretty in depth topic. But I’ll try to give a good overview and one that appropriately sets the stage for diving into more nev specific discussions. So Aetherium is essentially a blockchain with a computer inside of it. The computers called the Ethereum virtual machine, or the EVM. And everyone who’s a participant in this network collectively agrees on the state of the EVM at any given time, so anyone can make a request to perform some sort of arbitrary computation on the network, like making a transaction or deploying a smart contract, which is the foundation that all these defy apps are built on top of, and whenever someone does, they pay some amount of eath, or rather gas, which is just a denomination of eath to the network, and one of the most important actors involved in Aetherium. And in any blockchain or block producers, this would be validators now, and previously, miners before a theory comes semi recent transition to proof of stake so validators propose and validate new blocks of transactions. This ensures consensus among participants in the network, and also secures it. And probably one of the most important factors with regards to MEV is that validators have free rein to order transactions however they want within a block. So when someone submits a transaction to the network, it’s also not instantly included in a block, the transaction gets held in this waiting area, or buffer zone that is called the mempool. And the mempool is publicly viewable. So while transactions sit in the mempool, anyone can see it, including searchers, who are sort of an actor in this nav. ecosystem and validators look in the mempool for transactions that pay the highest gas fee, and they then include them in the block. So it’s not really like a first in first out system, like in Trad phi, there’s sort of a gas auction that’s occurring here. And new blocks are created every 12 seconds as well. So the basic breakdown here is that block producers create and confirm blocks, blocks are composed of a sequence of transactions based on fees, and block producers possess the ability to order these transactions however they want within set block.

Corey Hoffstein  10:33

So in the context of that description, can you explain what MEV actually is?

Grug  10:40

Yeah, so this is still a rather hotly debated topic. I mean, there have been a lot of efforts to come up with a more formalized definition of what MEV is, the acronym has shifted from standing for minor extractable value to maximal extractable value has activities that might not have been classified as MeV, at some point in the past now are or at least considered sort of MEV adjacent. So we don’t muddy the waters here too much. I’ll stick with a more rigid definition, which is also the one that I believe flashlights uses are similar to it. And that would be that MEV refers to additional profit, aside from block rewards and transaction fees that a miner or now validator can make by auctioning off transaction ordering rights in the blocks that they mined, more specifically prioritizing, or entirely excluding certain types of transactions. And this will make more sense when we jump into what flash bots is. But I think that’s a decent, broad definition, in my opinion.

Corey Hoffstein  11:39

Yeah. So let’s go right in there. You’ve mentioned flash bots a couple of times, what is flash bots? And why is it so important, at least historically to MeV?

Grug  11:50

Sure. So before I get into what flash bots is, and what they do, I think it’s important to establish the types of participants in the nav ecosystem that make this whole sort of grand digital dance work. And these would be users searchers and validators. So users generate the flow that creates the value to be extracted. Searchers hunt for profitable MEV opportunities and build the bots to submit profitable transactions to the network, and validators confirm the users and the searchers transactions. And I think walking through the evolution of flash bots and MEV over the past several years would probably be helpful here too. So prior to flash bots, and the advent of private meme pools, most MEV opportunities involve searchers competing by sending multiple transactions in real time often triggered by other searchers chasing the same opportunity, there was really no way to effectively communicate with block producers, how to granularly Express transaction ordering preferences, and flash bots change this. And the reason flash bots exists is because during the second half of 2020, and the start of 2021, big spikes and Aetherium usage brought with them a bunch of negative externalities that were sort of incurred onto users, and were caused by me the extraction. I think PGA bots are a good example here. They cause just mountains of spam and transaction reverts that lead to network and chain congestion, which made gas fees become artificially inflated, and block space artificially scarce. So flashlights saw that there was an inefficient communication channel between bots and miners for order preference, which was leading to all of these negative externalities incurred on to users and they set out to fix this problem. But ultimately, at its core, Flash bots is a research organization that’s focused on mitigating the negative externalities, meth incurs on users. Really though, Flash bots builds MEV infrastructure. I would personally say they mostly build software for validators, but we’ll be charitable here and say that they build me the infrastructure for searchers, and the ecosystem in general, their primary product, and the ones that see the most use are flash bots, auction and nav boost. So flash bots auction basically establishes a private communication channel between searchers and validators for communicating transaction order and preferences within a block and flashpoints auctions it’s a sealed bid auction, so you can’t see other searchers bids. The basic flow of how a searcher interacts with Flash bots, is that they build a bot to identify or capturing MEV opportunity. The bot finds that opportunity and submits a bundle along with a bribe. If they’re bribe wins the flash bots auction their transactions are included on chain and they capture the opportunity because any the opportunities are a zero sum. The more competitors there are trying to go after a given opportunity. The greater percentage of a searchers profit ends up giving away to the validator. via the auction and flashlights auction is built into MEV Boost, which is the software that validators run, and is flash bots implementation of proposer builder separation on Aetherium, which we can get into at a later date. But I think the last thing that would be instructive here is exploring an example of how a searcher might use flash bots. So say a user makes a trade on a decentralized exchange, like uniswap. And he’s trading some very illiquid shit coin, because the liquidity for the pair is not very deep, the user has to set his slippage high on uni swap in order to make sure that his order gets filled. This leaves him vulnerable to something called a sandwich attack. So a searcher who’s built a bot sees this user’s transaction sitting in the mempool. And he submits a bundle to flash bots to sandwich the user. A bundle is just one or more transactions that are grouped together and get executed in the order they’re provided. And they can include both the searchers and the users transactions. In this case, the searcher wants to sandwich the user. So he places a large buy before the users transaction, which increases both the slippage and the price of the token that the user is trading. He puts the user’s transaction in the middle, which drives the price of the token up even further, and gives the user a worst quote. And then he places his transaction, his sell transaction immediately after it, and then pockets the profit. As a result, it’s generally considered a very malicious form of nav.

Corey Hoffstein  16:32

And unlike traditional finance, just on that last point there, because it’s all done via smart contracts, The Searchers can actually calculate precisely how much they can move price with the transaction and and optimize the trade. Right? Exactly. Yeah. So one of the things you mentioned there was this idea of proposer builder separation. Can you explain what that is?

Grug  16:57

Yeah, yeah. So prior to Aetherium is transition to proof of stake block proposers built the blocks and proposed the blocks. And at the protocol level, they still technically do. But now majority of validators run flash bots nav boost client, which essentially separates proposing and building into two different actors. So validators running me V boost sell their block space to these specialized third parties called Block builders, who then collect and sequence transactions to produce a block and pass it on to them. And block builders want to produce a block that maximizes fees collected in the form of priority fees, and MeV. And this sort of just leads to the most efficient, the most efficient way to produce a block and it also prevents MEV from centralizing the validator set and mitigates against censorship, which is generally a concern as most people regard any v as a centralizing force. So if the block building competition is very fierce, builders are required to bid up as their bid approaches total revenue. And this reduces their net income, while increasing the net income of the validators in general, which makes it harder for builders to centralize their potential percentage of total eath staked previously validators could use any vi to bind more validators to have a greater chance of extracting any V, and then use the revenue to hire better searchers etc, etc, which is sort of like a centralizing risk. And on the other side, separating builders and proposers from one another makes it much harder for builders to censor transactions because all sorts of different complex inclusion criteria can be added that ensure censorship isn’t taking place before the block ends up getting proposed.

Corey Hoffstein  18:46

You mentioned sandwich attacks is one example of an MEV strategy. Can we maybe take a step back and move to a higher level? I’d love for you to talk through maybe the topology of MEV strategies, the general way you would consider categorizing them and maybe provide an example of each.

Grug  19:06

Yeah, absolutely. So I think the best example of a longtail distribution is generally the most accurate. So on the left side of the distribution, we have the nav that occurs the most. So this would be like sandwiching, arbitrage, and liquidations and these categories encompass the most met that occurs on chain. Some people would probably lump stat ARB in here two, which has become a larger and larger percentage of total nav so far, but to keep things simple, we’ll, we’ll stick with these sort of big three categories. And then on the right side of the distribution, we have a ton of different longtail strategies, which encompass just a wide range of activity where there’s much less competition, there’s a lot fewer searchers going after these opportunities. So sandwiching as I touched on before, is Probably the most and malicious and attested form of MeV. And it occurs when a searcher buys an asset immediately before a targeted users transaction, and then sells that same asset immediately after. And one thing that, as you talked about, that’s interesting when it comes to savage attacks is that you can use the pricing formula as detailed in the uniswap docks to calculate the exact price impact that any order will have on any given trading pair. And this is a pretty important advantage that sandwich attacks have over front running and Trad fi, where calculating slippage is totally probabilistic. So because individual trades on decentralized exchanges are all deterministic, searchers can derive and calculate the optimal trade size to front run a purchase with, which allows them to extract the maximal amount of profit from any given sandwich attack. Most people use sandwich tax, as I said, is totally exploitative. The broad opinion is that they’re parasitic, and they prey on uninformed retail users who have little to no concept of the blockchains underlying mechanisms. But there is another take on the opposite side of them being a net positive or at least neutral to the ecosystem. And they take here’s kind of twofold. The first point being that sandwich attacks incentivize protocols towards better mechanism design, like MeV, shielding the private transaction channels, which I generally agree with. And the more niche take here is that when a user submits a swap via an automated market maker like uni swap, they’re declaring their maximum willingness to pay. And so if they get filled at the better price, that they’re declared willingness to pay, there exists this consumer surplus and thus the absence of a Nash equilibrium. So the sandwiches actually moving the system towards the Nash equilibrium, so that the market meets the consumers maximum willingness to pay while also extracting some value for themselves. And in doing so the sandwiches generating more utility, and reaching the Nash equilibrium where no other action would produce additional utility given the current state of the system on the declared economic preferences of the participants. Not sure I really agree with this take, though, because I don’t necessarily believe that users are actually declaring their maximum willingness to pay, and they’re likely just ignorant, or sort of being led astray by UI and UX design on some of these are the front ends for these decentralized exchanges. My stance on sandwich attacks is a bit more of a moral argument, I think that they are fine or at least tolerable, because crypto is just the biggest and greatest casino ever created. And in the end, almost all of the activities that people are engaging in on chain or zero sum anyway, the second type of MEV would be arbitrage and the concept here is pretty straightforward. And one that your listeners or anybody with a Trad fie background will probably be the most familiar with. So we’re essentially moving money from one market to another and we’re pocketing the spread is the profit arbitrage on Aetherium functions the same way it doesn’t track fi with one crucial difference transactions on Aetherium are all atomic, which means that either all the conditions are met or the transaction reverts. So in Trad Fie, when you’re executing a multi hop orbit are you run the risk of the market moving against your position while you’re halfway through your arm, which would force you to sell at a loss with the existence of atomic execution on Aetherium. Not only does every leg of your hop, execute and be guaranteed to execute one after another, but you can record in your code your initial starting balance. And at the end of execution, you can assert that you want your ending balance to be greater than your starting balance. So then if the assertion fails, the smart contract will revert and it basically undoes all the trades made and leaves you with your starting balance. So this means that atomic ARBs are effectively riskless. And another one of the ways that searchers make money via ARBs is called back running. In this case, you want your transaction to land right behind someone else. So let’s say Bob buys X token on decentralized exchange a when he does so he moves the bonding curve of the AMM that he’s trading on. And this means that the purchase token is more valuable now, while the token that Bob just sold is less valuable. So a searcher Alice, who’s back running his transaction can then utilize this imbalance that Bob has just created to perform arbitrage with another decentralized exchange. So Alice buys the soul token from Bob for very little and sells it on another decentralized exchange for a premium. And that premium minus the transaction fees that Alice incurs ends up being her profit. The third type of MEV in the big three would be liquidations so just like in Trad, fie and defy there are lending markets, except because we don’t have the State to chase people down if they run away with the money all the loans in defy are over collateralized. And at their most basic level defi lending protocols consist of users borrowing assets against some amount of collateral that they deposit into the protocol. But the value of a user’s deposit to collateral can drop sharply. And lending protocols need to ensure that all the outstanding loans are always fully backed, or I’ll stand up risking insolvency. So in the event that the value of the collateral drops below a certain threshold, the protocol actually lets anyone liquidate the users position. So to incentivize liquidations, these protocols pay a flat fee, or they allow liquidators to keep a cut of the collateral being liquidated and the liquidator needs to repay some or all of the loan amounts of the protocol. At which point the protocol transfers the borrower’s deposited collateral to the liquidator. So the value of the deposit and collateral it ends up always being greater than the value of the loan, because otherwise there would be no economic incentive to for a searcher to liquidate the loan. And while you can liquidate a loan by just repaying the debt from your wallet, it ends up being quite capital intensive, like you’d need a million dollars to liquidate a million dollar loan, of course, and so you can expose yourself to losses if the value of the collateral you receive continues to drop. So lots of searchers instead, leverage something called Flash loans to front the necessary capital to liquidate the position and sell all the received collateral and then repay the flash loan while keeping the extra profit for themselves. And a flash loan. Essentially, it’s just a tool that allows someone to borrow any given amount of assets available, as long as they repay them in the same block for a small fee. And then the fourth type of MeV, which sets you know, sort of on the narrow side of the longtail would be, these are opportunities that fall outside of the scope of the majority of MEV activity on chain. And these would include like sniping a fat fingered listing for a valuable NFT, like a crypto punk, or as well as like a litany of other strategies.

Corey Hoffstein  27:15

Can you provide an example of some of the longtail strategies you’ve pursued in the past?

Grug  27:20

Sure thing, I of course, won’t talk about strategies that are currently generating profit for us. But I’m happy to discuss those that are deprecated at this point, and more, the alpha has totally decayed to give your listeners a picture of kind of how this all works in practice. So circling back to the criteria I outlined before, I think highlighting a strategy that ticked all my boxes, and proved to be one of the more successful ones that we have pursued would probably be the most insightful, and this would be NFT sniping. So in the early days, the NFT Bull Run, there was such an abundance of dumb flow. And block space was in such high demand that lots of users couldn’t get in on these initial sales for teammates, and then they would be forced to buy on secondary markets after the fact. And as a result, most of the higher quality NFT drops, they would Mint for around point 1.2 eath. And then they’d immediately sell for many multiples. So we would basically bought these mints via flash bots, and by hundreds of entities at warrants, and then sell them soon afterwards, you would think that you’d want to dump most of these things as soon as possible. And if you’re doing this today, you really would, and even then you might still lose money because the dumb flow and the NFT space is largely gone. But when everything was just going up, and we first started doing this, we would actually just sell as quickly as possible. But we soon realized that a lot of the inventory we unloaded quickly just kept going up. So we stopped doing that. And we started holding them for longer periods of time and continue to make even more money. So that’s one example. And another strategy that we’ve done very well with is what I would kind of term searcher versus searcher strategies. So running nav lots involves deploying smart contracts on chain. And like any piece of software, these contracts are occasionally deployed with vulnerabilities. So if you can find and exploit those vulnerabilities, you can basically trick a bot into handing its entire balance over to you. Sometimes this is a couple $1,000 Sometimes it’s much, much more. It just depends on who’s running the pot and how well capitalized their operation is.

Corey Hoffstein  29:34

Some of these longtail opportunities seem to appear really sporadically. So one example you gave was fat fingering a listing for crypto punk will recently someone listed the sale of a board ape NFT that had associated staking rewards. And MEV searcher bought the ape liquidated the staking rewards and sold the ape in a single transaction capturing more than $60,000 in those rewards by Otherwise, I’m not saying that person was you. But I’m also not saying it was not you. But these sort of fat finger events seem to be pretty rare, you could build the searcher bot, and you could just never get a hit, or it could miss that one opportunity per quarter that shows up. So my question to you is, do you think the edge in longtail strategies is in depth within your strategy or deploying a breadth of strategies?

Grug  30:25

I would say definitely, in deploying a breadth of strategies like that’s generally been our conclusion, if you’re going to pursue longtail opportunities that only come around once in a blue moon, you should probably pursue as many as possible simultaneously, so that your hit rate is as high as possible. One thing I will say though, is it’s often the case that the rarity of these opportunities actually ends up having some benefit. And this is because most searchers won’t waste their time spinning up infrastructure for a bunch of them. There are several very profitable opportunities that have been captured, where there were almost no competitors, or absolutely no competitors. And a good example here that comes to mind off the top of my head would be last year, there was a user who fat fingered a global offer on an NFT marketplace that was rather new called Lux rare for this collection Moon birds, they basically accidentally placed a bid for 240 eath, for any moonbird When the floor was actually 48th. So this made it possible for a searcher to buy a moon bird off the floor, sell it into their offer, and pocket the difference. And as far as I know, there was no competition at all in this instance, and the searcher walked away with around 200 eath.

Corey Hoffstein  31:37

So some of these longtail strategies seem pretty capital inefficient, right? What you just mentioned that searcher needed to have 40 eath at their fingertips. But you’re just waiting for the right moment to execute in that moment might not show up. So why don’t you see more searcher bots employ flash loans and that sort of situation.

Grug  31:57

So some of the time it’s just not feasible to use a flash loan based on the strategy being executed on so like, if you’re minting a ton of NF T’s via flash bots and bought in some sort of NF T man, and you want to then go dump those NF t’s on open sea or blur right afterwards, you can’t use utilize a flash loan, because you end up with a bunch of JPEGs, which you obviously can’t actually pay the flash loan back with. And then the other reason would be that flash loans are less common would be that most of the good searchers have inventory now, and flash loans are inherently gas inefficient, which means that you’re less likely to win the flash bots auction and capture the opportunity you’re chasing, if you’re using a flash loan. Now, back running isn’t really a longtail strategy. But lots of back runners do still use flash loans, as far as I know, but well capitalized teams will avoid them if they can, it’s better to use your own inventory if possible.

Corey Hoffstein  32:55

What do you see as being the main risks to these long tail strategies? Why don’t you see a lot of other searchers competing in this space.

Grug  33:04

So other searchers do still compete for longtail opportunities. They’re just less competitors than for things like atomic arms or sandwiches. And there’s a number of reasons that they’re less competitive. Some are non atomic, for example, so you end up holding risk, like you could end up holding 100 tabby cats, for instance. And if you don’t have, you know, the best knowledge of like the NFT markets, we are worried that all of those heavy cats will go to zero before you can sell them. This was less of a problem during peak mania. But it’s a large problem. Now, like most searchers are not super clued in to NF T markets. They don’t want to waste their time staying plugged in to what the latest and greatest drop is. This was actually a distinct advantage that we had over other teams when we were actively Bahding tons of drops, as you know, based on some of our prior interactions and keep very up to date with the NFT ecosystem, or at least I did for some time. So that was a huge aid to us there.

Corey Hoffstein  34:04

These longtail opportunities also seem pretty cyclical. You’ve mentioned a few times one of the primary drivers being retail flow, cure curious how you manage the opportunity cost of developing a strategy versus the risk that the cycle is going to end before you can even deploy it.

Grug  34:22

Yeah, this is definitely a problem we’ve run into. And we have had bots that we wasted our time on only to find that the return on investment wasn’t worth the effort expended on building and deploying the strategy. In terms of managing opportunity cost. We generally scope out the expected value of a given strategy versus the amount of development time that we think it will take to get it up and running. But this is obviously not a precise science and we still you know Miss sometimes as well. Little anecdotal

Corey Hoffstein  34:53

story there. I think I accidentally bumped into you while you were developing one of those bots with a friend In real time over voice chat in the BC Discord server. You guys were trying to mint I think it was the times cover NF T. And you were building the bot like, minutes before the mint was supposed to go live. And you didn’t know whether it was going to work or not. And you just said, Screw it, we’ll go for it. And there was quite a bit of capital at stake there. So I think it sort of shows I think you’re probably more buttoned up now. But those early days, there was a little bit of flying by the seat of your pants, perhaps to to take advantage of those cyclical opportunities.

Grug  35:28

Yes, yes, absolutely. There are still examples of times like that, where sometimes we just hear about something, you know, a couple hours before, and maybe we don’t have all the infrastructure up and running already. And so yeah, that’s still a very common occurrence where it’s sometimes we only have three or four hours and it’s go time. And you know, a lot of the times it works, but some of the time it doesn’t.

Corey Hoffstein  35:53

We spent a lot of this conversation talking about Ethereum based MeV. So far, a big part of the last cycle was alternative l ones. So a VAX or Solana or the emergence of L twos that are trying to make Ethereum or other l ones more efficient. How does the emergence of these Alt L ones or L twos change the opportunity set for MeV? searchers?

Grug  36:18

Yeah, so I’ll preface this answer by saying that we don’t do a lot of searching off of the theory, we did do a ton of stuff on a ivax for a while, but the chains relatively dead now. And so this is one of the domains I know probably less about. But basically, each new chain adds to the me the opportunity set, because the markets and the participants on them tend to be pretty inefficient at first. So there are fewer people that understand the intricacies of each chain. And these intricacies can have a big impact on nav opportunities and strategies. So yeah, generally when a new chain emerges, as long as the flow is actually there, it’s a goldmine at first for those that dive into it. And this was kind of what we found with a Vax. Although the timeline was not very long there, the more complex the chain is, the longer the alpha tends to last. And then as time goes on, competitors, see the profits that are being made. And they move in, and they make the market more efficient. So at one point, in the summer of 2021, there was this back runner that made roughly $30 million dollars in less than two months on BSE. And today, that same bot maybe makes like $5,000 a day, because competition significantly reduced his edge. I think that’s probably a good example of the cycle of MEV on relatively new chains,

Corey Hoffstein  37:38

given that MEV profits tend to be correlated with retail flow, and retail flow is updated quite a bit. What sort of strategies have attracted your attention in 2023?

Grug  37:48

Yeah, so this is absolutely correct. And as I said earlier, the number one question I tend to ask myself when I’m pursuing an opportunity, or looking for one is where’s the dump flow? And as the amount of dump flow or retail flow decreases, the amount of value available to be extracted also decreases? And so in terms of stuff that we’ve been doing lately, and strategies that have attracted our attention, I would say primarily like PvP, and with other nav bots. These are the searcher versus searcher strategies I touched on earlier, as well as farming airdrops, which I don’t think I would really classify as MeV. But as probably somewhat MeV, adjacent.

Corey Hoffstein  38:30

What skills do you think are necessary to be a successful MEV searcher,

Grug  38:36

I would probably break them down into three different categories. So you need to understand how Aetherium works at a low level, and have a good handle on smart contracts. You need to understand how defy protocols work. This includes things like automated market makers or decentralized exchanges, lending markets, things like that, because these are basically the building blocks that make the MEV possible. And you need to have a good grasp on math, computer science and economics. And obviously, you need to be a good software engineer. And I would add one more on the long tail route. If that’s the direction you decide to go, that you need to be good at rapidly identifying new opportunities moving quickly and not falling down optimization rabbit holes, because you’re going to need to jump from strategy to strategy very often.

Corey Hoffstein  39:25

Taking a quick tangent. It seems like air drops have been all the rage lately. And I know that’s another area you’ve been quite successful. You actively participated in the blur AirDrop, as well as did quite well recently in the arbitrary AirDrop. First, can you explain what an airdrop is to folks who are listening that have no idea? And then can you explain your general philosophy in terms of finding and maximizing AirDrop opportunities? Is it something that you can do systematically or is this something that has to be done by hand?

Grug  39:59

Yeah, Absolutely. So an airdrop is an event where crypto protocol gives out tokens to users. Generally, this is based on prior use of their product. And it’s sort of a method of bootstrapping community as well as instantiating. Governance. And governance is basically where users with a given amount of tokens can use these tokens to vote and proposals by other users, or the protocol themselves. And this kind of ties into the whole decentralization narrative of Defy. And in terms of like, strategies, I want to tread a little carefully here and not reveal too much alpha, because sibling airdrops is a strategy I see with a much longer time horizon than, you know, other MEV opportunities, but I can speak to it somewhat. My general philosophy in terms of finding AirDrop opportunities is that protocols that are well funded by top tier VCs and capture a large amount of flow, or are expected to capture a large amount of flow, land near the top of my list. So things like layer ones, layer twos, bridges, or just protocols in general that are very hyped up on Twitter, and as a consequence, have a pretty good user base when it comes to maximizing opportunity. And most importantly, not being excluded or identified as a civil attacker, we take the approach of making all our accounts look like a regular defi user, avoiding cross contamination, like not commingling funds, things like that. And a lot of this can be done systematically, although early on we did almost all of it by hand, and it was incredibly arduous. Probably can’t speak to it too much more than that, or no, because I tip my hand a little bit. But that’s a good general overview. I think,

Corey Hoffstein  41:43

for a lot of the atomic MEV strategies, the opportunity math is fairly straightforward, as we discussed a little bit earlier. And so the game seems much more likely to be played out in auction theory. I know this is a place in traditional finance, for example, trading desks competing for flow, where machine learning has actually been employed. Curious if you think machine learning is going to play a role in MEV going forward.

Grug  42:10

So at least for the foreseeable future, not really, we’ve actually utilized some ml based techniques with some of the searching, we’ve done. But diving into that, as it is something we did that was pretty niche. I don’t know if I could touch on it without revealing far too much alpha. So I’ll just say that there are some use cases, but they’re pretty rare. And most searchers are not leveraging any sort of mL from what I’ve heard of, and Trad fi, you know, most of the ML type stuff that’s been done, it’s just kind of like people running a lot of regressions. I don’t know if it’s like the complex Annelle work that’s maybe being done at somewhere like open AI or other like big large scale AI companies.

Corey Hoffstein  42:52

Do you think MEV is a net positive or net negative for the Ethereum blockchain? And if it is a negative, do you think it’s something that can ultimately be solved for?

Grug  43:02

So I think there’s a lot of nuance to this answer. Things like sandwich attacks and front running are a net negative for the user, whereas things like arbitrage are probably a positive help even out user quotes across exchanges. So some MEV is good and some is bad. Overall, though, I would say nav is a positive for blockchain security budget, because it disincentivize validator on staking for things like yield farming by actually increasing the block subsidy. And it allows for chains with no block rewards at all to be reliably secure, and makes markets just more efficient overall. So in terms of solving MeV, though, my take is that it’s a bit like trying to solve for gravity to me. MEV is an emergent property that’s inherent to distributed consensus. So no, it can’t really be solved for in the sense that it can be eliminated. However, you can absolutely reduce the negative externalities inherent to MeV, which is basically fleshpots chief goal.

Corey Hoffstein  44:00

What happens to me v if infrastructure becomes more centralized? For example, what’s the end game if flash bots are likely to end up controlling all the validators?

Grug  44:09

Yeah, so flash bots won’t ever end up controlling all the validators per se but all the validators are, the vast majority of them will probably inevitably run MEV boost lyoto is more concerning though, because of the inherent network effects of staking derivatives, they could theoretically gobble up a huge majority of the total validator share. And if Leto has all the validators, they could extract all or most of the MEV and further boost their staking rewards and use that revenue to hire the best searchers to extract more and the V and so on. And there are have been a lot of people who have raised these concerns about lotto and tried to curb the amount of eath that they’ve been gradually acquiring. There are some competitors like rocket pool and other protocols that have their own like liquid staking derivative But at the moment lotto is definitely in the lead.

Corey Hoffstein  45:03

All right, Grug, last question of the episode for you. It’s the same question I’m asking every guest. Every guest shows a tarot card that informs the design of their cover and you chose the magician. Curious what drew you to that card?

Grug  45:19

Oh, I chose the magician because I think a lot of people think MEV is a little bit like technical wizardry or on chain magic. And so it seemed the most fitting to me. It kind of gets mythologized or has this like bit of mystique to it within the crypto community, and it just seemed like the best fit.

Corey Hoffstein  45:40

Well, Grug, I appreciate you coming on and teaching me MEV.

Grug  45:44

Thanks so much. It was great. Very much appreciate it.