MEV after EIP-1559 - Alejo Salles & Kristof Gazso

awesome thank you um all right so i'm going to talk about uh near future which is uh 1559 upcoming um this is joined work with kristoff from nethermine and let me also say that this is very much a work in progress and this is hopefully we'll have a write-up of this soon um so perhaps some of you already thought about the question of mav in one five five nine and the first uh conclusion one might arrive too there's nothing new uh under the sun here uh we're gonna base burn the base fee but there's no uh new uh mav extraction opportunity that's uh raining on us so it's nothing nothing really strange but still we think there's some so this is perhaps correct at order zero but we still think there's uh some interesting things to be said about 1509 and med so hopefully this serves to spark some further thought on this so i only have 15 minutes and i already spent my talk index time so it's gonna be three surprise topics coming um first one is on minor economics so naturally there's a question of okay now that they will be have less revenue after one five five nine will they be incentivized to extract more nav in particular uh in a fair way like uh charlie pointed out before is there gonna be more reorgs and time bandits and so on so okay this is an interesting question first we realized we need some better understanding of minor economics actually so the first question we ask ourselves is will miners switch lanes meaning will they be incentivized to switch to other blockchains after the the reward cut so we built a super simple mod i'm going to run through it uh to try to assess this this question this is an equation for profit so the profit is just reward minus cost there are some equations for a hash rate there's a total hash rate that's distributed between ethereum and x other gpu based blockchains after uh 1.59 this looks again distributed between post-159 ethereum and some other blockchains assuming the total hash rate stayed the same and there's a revenue drop uh due to the burning of uh fees in in one five five nine parameter parameterized by some parameter gamma um then we state some equilibria here we say like the the profit per hash rate is is the same in ethereum and in other blockchains otherwise it would be a non-equilibrium situation where miners would be switching and another equilibrium holds after the london fork so we plug in um some algebra here we do some uh rearranging and we can find an expression for the actual fraction of a hash rate that stays in ethereum after the london fork let me say that this is of course a very naive uh model it doesn't consider in particular the cost of switching or the fact that the revenues are denominated in different currencies and miners might have more or less trust in the different currencies so again this is a very course first approach to to this question um but we can do some nice colored graphs here you see the fraction of miner staying in ethereum after the forecast function of the revenue drop uh after 1.59 and we have another parameter which is again the size of this other uh uh blockchains in terms of revenue for miners as uh in proportion to ethereums right so we can uh plug in some numbers here with actual uh data from the different revenues the sizes of our blockchains and we actually find that we're in the red cross there in the in the graph it looks like most miners will stay in ethereum according to to this prediction it's 0.98 so 99 of miners would stay would be would the equilibrium would be at 98 percent of minor stain of course um again these numbers are super tricky they change by orders of magnitude so you're all aware of how gas prices change so this is uh take this with a with with a grain of salt so okay this is for again the minor economics and switching to other chains how does mev play a part here so uh first question that comes to mind is how much extra mvv uh miners will have to extract too much profits before one five five nine and this is a whopping 150 percent more so they would have to go through a lot of effort to try to uh increase their their extraction uh widely to actually compensate for the drop in in in minor revenue um but of course this this might be like prohibitive or it would be easy of course to try to come up with strategies to do that so next question is try to integrate the two things and and and compare this uh the extraction with the switching so we can ask how much extra mav would need to be extracted to match their revenue height from switching chains okay so so when so to achieve this equilibrium uh minus in achieving this agreement they would increase their profits a little bit and how what what would that be compared to if instead of switching they would extract more maybe and this is a more reasonable number of 22 percent of course these numbers don't tell us what miners will actually do it's not enough to think about these numbers in terms and and try to come to conclusions but it gives us a framework to compare the two different things that are uh switching to other uh other chains and extracting more mvp so we think it's a it's a first starts to work a step would be incorporating costs more uh more quantitatively and try to model that and perhaps agent-based modeling and so on so again this is a first approach take it as a working process i think it it illuminates a little bit the the the different ways of this of these two factors of the equation the switching chains and the med one okay the second topic i wanted to uh talk about is uh the fact that there's actually many options uh going on in ethereum despite the picture we have of uh running one auction for inclusion so what uh 1559 was designed and mostly analyze in a pre-med world uh charlie already gave a timeline for how mvp developed and although the the mev idea was there earlier the extraction was much less and was a fringe activity now we're in a world where this is commonplace we have like flashback suction all the time uh but but what happens here is that this uh this design assumed there was an auction that the ethereum option was for for transaction inclusion in blocks while in reality we know that users speak for much more than that they also be for ordering of transactions they bid for privacy and and so on so how does this change things um so in particular i'm going to focus on team ralph gardner's analysis of one five five nine and so he posits three uh things that are desirable from a transaction fee mechanism the first is a minor myopic minor incentive compatibility saying that mine is incentivized according to to actually according to according to the the mechanism in the first place in the short scale one block uh second is user incentive compatibility meaning they'd achieve the desire ux improvement of users expressing their true preferences for inclusion without needing to strategically speculate on what other users are are doing and off-chain agreement proof-ness meaning that uh users and miners cannot outplay in the system by uh making agreements outside of of the chain so so he shows that uh mic and oc proof are satisfied by one five five nine uh but the ux improvement is only achieved outside periods of high demand in that same paper he posits another alternative possible mechanism which is the tipless mechanism which is chose to be myopic minor incentive combined compatible uh user institute compatible and only oca proofing in periods outside figures of high demand right so here we are uh we're left wondering okay right now we know that ocas are commonplace these days via flashbacks for instance for for ordering of transactions so perhaps having a tipless mechanism would have been a a better idea like where users are all the time guaranteed uh that they're they don't have to speculate on another uh user's price setting uh when they only care about inclusion and separate the auction for ordering right so so again this this leads us to think okay there's this many options going on perhaps we could think of the inclusion one as like optimized for user experience and have the the ordering one um be taken be done elsewhere um in particular we have like already mentioned this we have the inclusion uh auction privacy the ordering one all like a hundred different levels if you if you will we're in fact that starting to discard tears okay some some searchers care about ordering but uh perhaps uh regular users only care that their trades are not front run so they only care about uh privacy so should we treat this this separately and in fact if you look at traditional finance they run different markets for different indus instruments there's one for stocks and one for options and these are all the different separate places so they can be more efficient and and if you look at the ocean it looks like it's perhaps more attuned to arbs than liquidations and there's like we do due to our pricing formula and this is all open and you can look at this formula in our documentation and there's discussions in the github but so this leads us to the greater question of can we efficiently encompass all preferences with a single auction and here the key word is uh efficiently again this is many things going on at once can we make it in a way that it's efficient for all the interests there for all the all the to all the users and so this is one of our key research problems so if you're interested in that please please uh please come talk to us and yeah now that mev is commonplace more generally we need earlier involvement with the eip process i'm happy to say that we are uh we're talking thinking a lot about x2 and how mev will interplay with this too all right final topic of my session it's on flashbox ethics so ethics is something that was on our research agenda from the very beginning it's really critical for us to be transparent and in particular right now eighty-five percent of uh the ethereum's castrate is running our software so it's a new shelling point and by this uh i'm thinking about a situation where miners could passively collude by just adopting flashbots default this is not different from like the situation before where geth was this this shelling point but now that we have that responsibility we will be open about it and and think about the implications of that and here i'm thinking about uh within protocol collusion so nothing like 51 attacks or or or something like that but but instead to give you a concrete example i think we have we have in maybe a parameter that's the max merge bundles so it's a parameter that says how many maximum bundles can be merged in in a block and so could we uh we flashbots lead to having only bundle blocks so no no vanilla transactions by playing with these parameters and feeding something to to the miner so this is clearly not visible because there's opportunity cost of eventually of leaving transactions out so this comes back to the first slide there where there's really nothing new here but this that brings yet another question so it doesn't seem like flash watch is more powerful after eap 1500 but can flashbots defeat uh 1500 and to wrap up uh my talk i'm gonna show you some ideas along these lines or one idea along these lines rather which is what if flashbots by default decided to fill blocks uh only up to a a size slightly smaller than the target side block size s0 right so we can actually simulate this again assuming 85 percent of the cash rate is doing this instead of what's uh what's uh what's meant to to be uh feeding up to a zero and this is what we find the this is the base fee as a function of the time for different values of this epsilon parameter for some demand dynamics this is based on notebooks by the robust incentive groups at the ros incentive group at the field foundation so you see here that if for some value of this parameter if all miners run this this this modification to the software they would be able to drive the base fee to zero of course the question is is this uh is it sustainable um economically right they would be forfeiting a profit immediate profit in the long run this would be good because they would have no there will be no burn fees so it will be all tips uh i think they can they can they can get revenue out of but it will be in the short term and profitable because it would be uh leaving transactions out so what is the critical number of underfield blocks uh for this to be for this to turn a profit for miners okay and this is the the part where uh it's a work in progress this is uh unfinished work so um uh hopefully we'll have a an answer for this question soon um and it's important to think about this question for for uh for how 1559 will uh ultimately have an impact on the network so stay tuned for an upcoming post on on this again this is trying to work with christopher in gaso and thank you all the people that that contributed to this conversation barnaby monroe and the ross incentives group team baker leo and for the entire flashboards crew all right thank you all for tuning in all right thank you alejo for and you made it perfectly in time yes so uh do we have any questions from eath global.tv live stream for the speakers if not um i suggest the speakers who just finished speaking can sign in to each global.

tv where you can directly interact with our audience on the live stream