Who Wins Ethereum Block Building Auctions and Why? | Devcon SEA

[Music] all right hi everyone I'm borak today we're going to present our work with Danning to you and our co-author Thomas is unfortunately not here today um as you might see from the slide and if you were here for tyron's uh presentation the title is heavily infu influenced from the uh famous book who gets what and why and we're actually going to look into that but in the concept of ethereum uh blocks and if you were around in the past couple of days a couple more folks from flash Bots actually had uh quite nice presentations and uh specifically Phil and Kristoff and one concept they were talking about was the centralization in how ethereum blocks are being built today and in fact we're going to start from right there so how are the blocks getting built today and we know that as an ethereum validator when it's your turn to propose a block you can either build it locally meaning that you will use your own execution load uh execution node and uh pick the transactions which go into your block and this has the benefit that you pick the transactions that go into your block but also the same thing is kind of also occurs because you have to pick the transactions which will like maximize your uh earnings from that block and if if you are not like a sophisticated um validator who is like aware of uh block or bundle merging techniques or block building techniques well this could lead that like some validators could make really good money while you are left behind and to avoid this kind of an heterogeneity in uh block rewards we proposed or implemented me boost which is a PBS implementation that's been active since the merge and the idea is that now the validators could actually just delegate their uh block proposing rights or auction them off uh to a set of external entities called Builders and the builders can will simply bid for the right to earn uh uh this proposing rights and in this um trade-off now the build the proposer simply gets easy access to rewards which is good for the protocol because now ideally we have like homogeneous access to immed rewards from the validator side good for consensus security but the problem is that now the validator has no say in what goes into their block and from an individual validator perspective this might not be like such a bad thing but we see that now this could be bad for our protocol and we see that around 10% of the validators only actually produce their blocks locally while 90% op in for me boost and what we see today is that as you might have seen already only two Builders are actually building like 95% of me boost blocks right now in our study it was like three Builders but right now it's it's a Dooly and why is this bad well if these two Builders simply don't want your transactions in their block or if they have an economic reason to exclude you they are free to do so and what this simply means is that the following slide is not available so we need to avoid this and there are a lot of pro um gadgets or uh protocols uh that's been under uh development for this right now and in this paper we simply wanted to understand okay how did these guys become so powerful in the first place and there are two aspects that we uh tackle this problem one of them is how do you earn market share how do you like get to uh win more auctions and the other one is how do you actually increase your profits or how do you become profitable while doing this because we see that there are some Builders who are like not building that many blocks but are extremely profitable whereas some Builders who have relatively higher market shares but uh they are not actually that profitable and we want to understand how like this can happen and what we see uh here in this slide is that if we classify the transactions based on their objectives like what they achieve uh or we call them um orderflow labels we see that telegram B transactions and me transactions so like your um Arbitrage sexx Arbitrage sandwich and so on they cumulatively contribute 51% of all value to ethereum blocks while only consuming 10% uh block space or 10% of the gas whereas um less valuable transactions like I don't know like your simple e transfers or your retail swaps they con uh consume around 60% of whole gas while contributing much less value and like one takeaway we have here is that the more valuable transactions are usually exclusively submitted so they skip the public manool and the blocks are usually filled with less valuable public transactions and on the right hand side you see a an orderflow composition of the top 10 builders that is like what percentage of value comes from which order flow label and uh like one of the results again we achieve is that the valuable order flows are also like uh available in higher percentage in these more successful Builders and one thing we actually uh find is that the diversity of the order flow is positively correlated with the market share so those Builders who have more diverse order flow can simply be more competitive or be competitive more frequently than those who are dependent on a single order flow type for example like if you are a Searcher Builder which you only get the order flow of your Searcher then you are dependent on that opportunity to arise so that you can win the auction whereas a builder like Beaver build um or Titan who is like getting diverse order flow can simply uh be competitive more frequently the other thing of course is yes you can VN the auction one way or other maybe you can subsidize but how do you actually make profits and to keep profits in the me boost auction structure right now you need to have exclusive orderflow which gives you the Competitive Edge to win the auction but on top of that keep some profits so or in other words you don't have to express your full value to win the auction and we see that if you have an exclusive orderflow provider which means you have a provider that is only providing meaningful order flow to you and no one else and you can think like canonical examples like SCP and beaver build or Vint mute and our sync then you can actually be more profitable and again there is a positive correlation there with having an exclusive provider and the percentage of uh profitable blocks that you can produce as a builder and in fact we find many other features which are correlated in the same direction with being profitable and having a a high market share especially among the top 10 Builders and this kind of indicates a chicken and egg problem that is like if you want to be profitable then you need exclusive order flow which like gives you this Competitive Edge which differentiates you from others but that kind of an order flow only goes into those Builders who already have high market shares because those order flow providers they want inclusion guarantees and hence like as a new entrance to the market you simply have no chance and this is why like we see such a a convergence uh in the number of Builders who are like simply uh winning the auction today and Denning will uh tell more about like reasons of that but this is like a one of the key takeaways and we also uh go further into some Builder strategies so like one of them is you can think it as like how they pack blocks and um what percentage of value they drive from uh their block regions and this is specifically uh interesting for the top of the block region because those competitive transactions which are uh okay to pay high fees they want priority and By Priority this could be like acting first on a state that was stale so like the last block for example or like doing a sex Dex Arbitrage which means that someone shouldn't touch the same state before them and that for this they're okay to pay high money but of course uh as a builder you need this order of flow to come to you so that like you can place them in the top positions and get the money from them and we see that the builders who have high market shares are also the ones again who are making the most money from top of the block and you can also see some interesting behavior in the end of the block region um and we can discuss that later on more but uh simply you can think it like some Builders allow some Searchers to do searching after they construct the block so on the state that's already transitioned before the simply like we move on to the next slot another interesting strategy here is the latency and bidding behavior that is of course this auction happens in a 12C period and um as a builder you need to get relay your transactions to the uh intermediary relay so that they can make them available to the proposers and of course latency matters in this uh in this structure because even if you have like a really high value bit which like you just uh maybe created in the last like 200 milliseconds of the auction if you cannot relate this uh information to the network then simply it's worthless and we see that this uh structure simply incentivizes latency and collocation with the uh relays and like one strategic behavior is um cancellations that these uh Builders use which is simply like lowering their bid in a strategic way which is maybe because they see that like a signal like an external signal is not in their favor anymore like if they're doing a XX Arbitrage now they want to lower their value that's available in the auction and now Denning is going to continue with uh so how do we end it up like with such a centralized Market structure yes so with all these learnings of all the behaviors and strategies the builders are doing to win the market share and to make more money back to the critical question how do we end up with a such centralized Market structure today um if we revisit this data if we compare how the trans transparency label or basically meaning if the transactions are going through public or private mle we can see this also interesting 8020 rule here on the right side you could see pretty much uh only 20% of transactions are going through as exclusive signal meaning they're going through priv Min pool while they are providing more than 60% of value to the final payouts to the proposer meaning the me value of the block um pretty much as Brooke already hinted The Exclusive Auto flow is providing much higher value than the public uh mol transactions with all these learnings what what's the outcome we Al we we can see that like from the right side of the chart uh from the bottom to the top we ranked the block Builders from the biggest market share to the lower one we can see pretty much the bigger the Builder is with the market share the higher private order flow percentage is everyone learned that okay we need to make sure get to acquire more transactions orderflow privately and not have other people to access it so we can basically be able to keep more value to ourselves rather than competing in the bidding to the proposers we can also see pretty much that on the left side most of the most me valuable transactions are pretty much all private today what does that mean if we look at at a time zone Horizon you can pretty much see that this Behavior or learning has been pushing all the transactions on ethereum towards private mle more and more over time during a 2 years time frame if we look back from the beginning of 2023 in the beginning of the 2023 we only see 3 to 5% of private mle transactions meaning every day when the transaction land on ethereum only 5% are going through private mle but this number has going up to 10% and then 15% in end of last year and then actually bumped up to 35% right now so today there are pretty much how to say onethird of the transactions on etheron are never seen by all those public mle nodes so why did it happen if we what happened to the jump in 2024 Q2 if we exam on all those top block Builders uh block and basically break it down by the orderflow content you could see that uh besides the orderflow that belongs to themselves exclusively for example the thing we identified as the beaver build who is also a sexex uh trading firm that has the SCP flow our Sync has their winter to Mu flow we could see some critical order flow for example like the orange one banana gun has been uh going up in Titan block um very interestingly it correlates with this like dotted line on the right side which represents the profit of the Builder that has shoot it up crazy in the beginning of this year um which correlates with the growth of the banana gun flow um so the answer to the question why did the private order flow went up over time is that we have seen this move movement of block Builders trying to compete and almost like competing in orderflow acquisition War um which happened quietly everyone is trying to get deal to the basically most valuable telegram Bots or potentially researchers who is the winner uh we can see that Tian has got the banana gun flow here Beaver build had an increase in the Jared from Subway the famous Searcher uh who is doing sandwiching attack on ethereum um the result is that basically we have a more centralized Market structure one interesting to highlight is that there's another thing that happened which is in Q2 metamask the wallet they starting to monetize their transaction order flow to block Builders which then drove the crazy 35% up um so to revisit how did we end up here today you could see that with the orderflow acquisition thing that we mentioned in the beginning of this year um you there is a huge change or like a pirad and shift of how the block Builders started from pretty much three builders in the past have uh half years become now only Titan and beaver um our sync pretty much gave up after they lost the biggest deal and decided to not to like try to keep their market share up because to do so they will have to give out more profits from their searching flow and also subsidizing potentially um so this is our takeaway me boost have successfully enabled uniform access to me rewards for ethereum validators however the current market structure is not ideal and it undermines ethereum censorship res distance and usuality properties um and we know that everyone in the market want a more decentralized Mar uh Market structure including the top Builders um by talking to the top Builders um the top Builders are also sick of this basically uh orderflow acquisition competing uh game and because there's a lot of inefficiency in current design for example if the top block Builders are trying to acquire orderflow uh exclusively to each other um meanwhile while these aut flows are actually not really conflicting in terms of the state accessing of the blog space they are basically competing say for example the centralized exchange that uh and uh exchange how do say uh searching flow competing against the telegram Bots flow which can totally exist in one block but today because they belongs to different block Builders they will have to compete to land in the next block and so to pay more to the proposer which also cause proposer Monopoly um another thing is that it's really hard for any any new Builder to come to the market without uh adding a lot of subsidy to Kickstart and markets in the inclusion rate charts which have evolved into a pretty toxic zero sound game where everyone is trying to subsidizing more which also then causing the top Builder to subsidize more and which is very unsustainable for the whole game um so what if we are able to find a way to decentralize a market set and redistribute value to all the people in the game um so our team Flash B we believe that um decentralizing the block Builder set is one solution towards uh solving this centralization uh of the market structure uh we we believe it has to be achieved step by step one is we have to basically design a set of the multi-operator set where multiple Builders can run um and you can trust them to be run by any of the neutral entity and another thing is that we can we can have this C how to say neutral platform to allow any block Builder to come to the block building algorithm and just to bring their order flow keep them private potentially in a tee um and we can be able to uh fairly redistribute value for the autoflow they provide uh one thing we have implemented so far is a gas refound rule um so the the idea is basically like uh whoever you have a a set of order flow that you want to build and land in the next block you have a measurement of the realized value which is the onchain payment um how much these transaction are paying including gas fee and tip to to the Builder um you also have a block value Delta or we call it marginal contribution value for the block so meaning for the block when we simulate uh to build a valuable block without this order flow how much is the value uh compared to how much the value will increase including this order flow the Delta would be your marginal contribution to the block and uh then we can basically calc how much refund we can potentially refund back to the autoflow providers um based on how much profit the Builder can take uh we have already um collected around 60 ease refunds for the orderflow providers that are sending to flash B Builders uh we have processed uh three e of them and we're hoping to see more and more adoption for this um thank you that's our talk and if you're interested in all those details of the data feel free to check out the paper uh which is uh includ in the AFT conference this year thank you thank you [Music] everyone thank you everyone and thanks for the speakers I I do agree a monopoly is never good right um if I could invite you all just to stand here uh to answer some of the questions that's posed by the audience so the first one and most upvoted what do you think about epbs yeah um so if you have attended uh Kristoff's talk yesterday simply uh his his theorem shows that or suggests that the centralized structure of the market would not change under epbs or any APS designs um and hence that's why we are at flash BS trying to put the building logic inside a te where like the outside uh interference is not possible and thus achieve some kind of a decentralization in the process because again EP BS yes could has benefits in terms of like providing an enshrined um market for the same purpose as meev boost without relay intermediaries but it won't be a solution for centralization okay thanks uh on to the next one which is how do you get the metrics for order flows for example the telegram board yeah so I I guess we should have defined the metrics meev what does that mean in our context um it basic it simply means how much value are paid out to proposers so we're calculating how much basically U each transactions are paying in gas priority fee the base fee is burned so it doesn't really count and there is a function called coinbase transfer in M boost um the software which pretty much looks like a value transfer in traces so them all together would be the value or the payouts for all the transactions um then there's a layer of uh bidding involved so when the block Builders are collecting all the transactions this is how much value they're receiving which is the true value of the block and then um we are looking at how much build are bidding to the proposers so the value of the telegram Bots is simply basically looking at their uh payouts on chain and labeling based on their routers okay thanks thanks for the insightful one uh okay on to the next one is the code publicly available for labeling autof flows um we have all the uh um queries that have uh used on Dune available and actually referenced in the paper but then some um some of the internal logic is I think not yet publicly available but we can definitely share if you reach out to us okay um next one did introduction of block markets have any effect on Builder Market um yeah one thing we observed and also we note in the paper is um it's not directly on the Builder market in terms of like who wins Etc but we see that after the introduction of uh blobs um the the gas uh space or the block space consumed by these um rollup transactions simply like which uses call data and so on actually lost significance so you can see I guess a positive impact in that sense okay next pretty interesting one isn't he just trust me bro it's definitely not TR trust me bro there is like remote adast station and many other another thing you can verify like if the code running in Te is actually what you what they're saying uh we actually have a teammate uh dears he had his talk about um this more specifically so I would suggest you guys to check it out um I would say te wouldn't help with Central uh censorship resistance and there might need to be other solutions to patch to that but and also just to add on to that I mean aren't we trusting anything so everyone so far we trust the relays I don't know we trust the build who we send our order flows to uh maybe this time we can try trusting like cryptography a bit more and give a shot to te okay I think we have time for one final one so can gas refunds increase spam as it would get it was mean the spammers can get the most of the gas back it's an interesting question um here's your take too but I I think um I don't think it will increase spam um or maybe you meant a different meaning of the spam but like if you're paying that much money um like the refund is kept at how much you're paying uh as of today so I don't think I don't know if you're what are you getting out of spamming uh us I mean I mean um so to get a refund you simply have to end up Landing in our block and uh have a have a contribution and of course depending on your unique contribution you get refunded so uh like you can spam but it will be just like bad for us in terms of like dealing with that but it won't like affect anything your refund as a matter of fact spamming might actually hurt your refund because we then we'll see similar order flow that is uh competing with your other flow and which makes the marginal contribution of your flow less all right thank you so much for the q&as there's a lot of questions thank you very much we will be back at 12: p.m. so feel free to get a toilet break stretch your legs see you later