DAPPCON 2025: State of DEX Markets- Execution Quality, Frontends, an Trends - Danning Sui

Transcript

found it. Okay. Um, hi everyone. I'm gonna share a talk about a review of the state of the DEX market. Um, I'm Denning Su. I'm a data scientist at Flashbots. Before flashbots, I worked at Zorx, which is one of the DEX uh platform in the space. And so today in the talk, I want to cover a bunch of topics. So I have a bunch of slides and bunch of charts. So hopefully I can cover all of them about looking into say like what are the front ends that are mostly dominating right now and like what are the use cases that we've seen in the past one or two years and how does the execution quality compare across different decks which actually has been a thing that people always talk about marketing about on Twitter but no one

have really seen the data to really justify it and some crosschain trains as well. Um so um yeah this is agenda rounddown. So first we'll have like a recap of how did the DEX space evolved over time and then we'll look at who's the king of the DEX orderflow right now on Ethereum specifically and compare across the specifically solver based decks who has the best pricing this is from one of the paper we published earlier this year and uh then a few charts about you know other use cases and across chain and last but not least a very interesting study we just published yesterday with flashbots um about the spam problem on L2 specifically base. Um so this is my

view of how DEX trading has evolved over time in the space. In 2016 2017 when Ethereum just came out there is order order book uh decks came out like Ether Delta um and they were basically following the Tratify model where basically if there's a taker and order uh and maker there's potentially also a relayer trying to match the order and if there's a match then the order settles on chain. Um, initially people were putting order book on chain which turns out to be very not gas efficient. And then everyone kind of adopted the model of like um how to say off-chain order book but onchain settlement. And so around 2018 when like unis swap finally came out they proposed a really novel model to basically solve kind of like the fragmentation of liquidity issue

where they basically price all the assets in the curve and where everyone can just deposit it in and it's a unified liquidity pool. So it's much easier to basically take liquidity from the pool and but that comes out with a lot of problems right like uh when you access price on the curve it's very possible that there could be other people actually end up having a better price before you if they paid a higher gas or they were basically trying to front you in the mele when they saw your transaction. Um so and like going on there is um more and more how to say more professional people who are coming into the D5 space and say you should also adopt like RFQ liquidity. So request for quote as for RFQ which is basically where there's a market maker

who they have a lot of inventory on centralized exchange and they say hey I can provide a better price than unis swap aml liquidity and um I would like to you know like provide a better quote for you and so RFQ how to say liquidity comes into the play and at this point we again have liquidity fragmentation problems so um 0x 1 in they all kind of like shifted towards a dex aggregator model so Now they're saying like there is order book, there is AMM, there's RFQ. How can I really like price the best for the user? I'm going to try to aggregate across all the liquidity sources and trying to find the best price for user. So that's when you know there's RFQ adoption and there's like aggregator model starting to shift and pretty much around 2022 when cow swap

came out they were saying aggregation makes a lot of sense but what if we also created some competition around solvers and outsource that job from just the one you know single entity say one inch or zerx using their own router. Instead, we just basically put an auction on top of it and say, "Hey, uh, anyone if you have a good routing algorithm, tell me what's the best price." And so that's basically the whole sorro model DEX uh, model that came out. Later on, 1 in adopted as well, pushed their fusion model. And then Uniswap also had Uniswap X. Um, and at the same time there's also, you know, like wallets and others who are basically starting to trying to basically uh, put this use case within their interface, right? So MetaMask

started this like metab aggregator model where they will check all the aggregators and they find the best price and they will keep the user within the UI and say if you want to do swap do MetaMask swap and they charge high in fee but user really don't care and they like the convenience. So this is where we are right now and 2025 what are the new things more and more decks starting to adopt also orderflow auction. So basically say okay we've had you know price aggregated but now we could also have searchers potentially background our users uh orders potentially extract me in a not harmful way and potentially give a kickback or like refunds back to user to enhance the pricing. So now it's a trend of like dex with offa for

example like cow swab they launched a thing called map blocker uh which is one of the biggest competitor to what flashbots launched which called mesh share uh it's a similar idea but you know the end goal is to increase execution quality for user and this year another big topic would be use v4 hook pools which is potentially innovating and like free up a lot of customization for user uh how do you say for projects to basically improve and how to Okay, potentially also protect the liquidity providers. So, we'll see if they can have, you know, like a bigger adoption in the coming month. So, that was the rambling pretty long, but I want to take a quiz in the room. So, with this, you know, like kind of like recap of all

these player in the space, who do you think has the most retail adoption today? And when I ask this question specifically, it's not about just like the DEX volume settled on chain because that could represent more like you know AML liquidity can be easily multihopped multiple times when user swap a less popular token a to pop token b you know like the router will do like a to west and then west to b and then so the volume easily inflated but here we're talking about you know like how much uh trading volume were requested as a demand from retail front ends. Uh, does anyone have any guess to take? Yes. UniS swap. Unis swap. Oh, okay. Um, I think you Okay. I think you know the answer,

but you're helping me to play a straw man. So, um, so we have answering orderflow.art and, uh, this is a website. We're trying our best to basically index all the front end um, volume. I want to make a like caveats here because I'm a data person. I don't want to do chart crime. Currently it's a little bit out of maintenance that it's representing one inch fusion charts wrongly. So there's a portion of 1 in fusion is not represented there but um per my manual check in even combining that cow swap turns out to be the biggest one uh in the front end right now and uh I checked that there trade sides buckets as well it turns out it's not really like I was suspecting if it's like one entity trading a large amount trading not really the trading how to

say sides bucket actually grow pretty pro uh proportionally so seems actually very organic growth um but yeah one thing in This diagram I want to highlight though is that when you zoom into the whole like dex supply chain today it's very sophisticated and it's pretty mature as a segment where what we mentioned earlier when the uh front end actually got the user request they go to a bunch of solvers and a bunch of servers might like uh route across AMM and RFQ market makers they may simply just be a market maker directly you know giving quotes um and vertically integrated um and then they actually all goes to like a orderflow auction to potentially get another batch of like background attached and then it goes to

block builders. Sometimes the solver may have a direct deal with the block builders and so a lot of these transactions are going to private meool now. So like it's never seen in public meool because people now realize more and more that me is harmful and you should just not like giving up all these value freely. So one data here is showing that basically um over time the private meo adoption has grown up to 40% transaction today now across all the ethereum transaction. So basically the meo are kind of drying up or like dying a little bit and that actually gives a lot of implication as well. I think um it's good and bad. good in the way that like the apps and front end are trying to internalize and recapture and

distribute the value of MEV rather than just giving it up to searchers and but the the bad part is that I think like with less and less transactions are being in the meool it actually give a lot of problem for a gas estimation for people who to trying to understand like what's the next blocks gas actually going to be um yeah and then another uh chart here just basically to show that so if we want to look zoom into the offa market share. How does it look like? Um there is not very good data out there except that Merkel offa has a public endpoints that flashbots we are scraping. We're trying to like put them into public goods into dune and also map block has their data on dune and we also

upload uh flashbots protect data there. We can see pretty much like Merkel right now has a 3 to 5% of all the Ethereum transaction flow ex exclusively. uh M me blocker also have around 2% similar to flash protect but there is actually a 30% chunk there that we don't know who they are but we know that it wasn't seen in public me so it could be another offa from like the meta mask uh smart transaction off uh could be uh how to say blink which is another big big offa it could also be a direct how to say exclusive deal that they're sending towards blog builder like tiden or Bieber um but yeah I would love to see more data on that Um if you how do I say measure this percent

percentage in a different way by looking at how much retail transactions DEX volume it's way more pro prominent it's around 70% and more uh and this trend is just like going up and only higher um so yeah so with all these sophistication right and we still don't know if any decks is doing better than others are providing better price you see all these decks fighting on Twitter if you're in that you know Dex corner of the sub like you know Twitter crypto Twitter you can see like unis swap thinks like cow swap settlement is too slow and actually is basically uh a how to say false index of like having good price um cow swap thinks no like unis swap x is like this and that and then the unis swap data

scientists came out be like this is wrong um so that's pretty much part of the motivation why we wanted to do a study in terms of like a pricing comparison um a few other reasons just to lay about the context for from end to end is that so for any Dex user today right like you go to a website and you have a quoted price shown in the UI maybe like if you you are a slightly more sophisticated or advanced user you might even go to multiple front ends and trying to check if the price looks better on the uh on one or the other um but what does that really mean it doesn't even guarantee still the final execution um if you look at this mental model from like Zerx a few years ago it kind of lay out how

does the pricing go through the whole process. So first of all you see a mid price midpoint price and then you have to add like a LP fee of the pool and then there's price impact which means like the bigger the trade size you have you have to go down the ticks of or basically like go down the curve and your average price or basically like final price of the order is going to be worse and that at that point that's a quoted price you're seeing in the front end UI but that's still not guaranteed um if you're accessing AMM liquidity right so then now you send it to pool and then there's searchers looking at your transaction being like, "Oh, this transaction is moving the price off this uh this pool compared to another pool.

There's arbitrage opportunity. I'm going to basically do something maybe push up the price further and then like dump it after or maybe I'll just simply arbitrage." And so there could be also another person just, you know, not malicious at all, but like just trading at the same direction, but they paid a way higher gas fee and they will go before you and then they will push the price further. So you will have a lot of like slippage and coming from both trade collision and also me. So at in the very end the amount you see in the wallets is like very different from the UI. So another thing is also like we see all these like solver model um that's going on but none of them really achieved the how to say the initial idea of like a decentralized auction because the even

today unis swap x 1 in fusion cost swap they get the quotes from the solver their back end decide which one is the best solver how do we really know if it's like transparently that solver actually had the best price or you could imagine maybe one of the solver has a deal with them they might promote then when even they like two bips force and others, right? So, another thing we usually talk about is that how decentralized are these platforms are because that could be a potential approximation of how uh much of the competition are and the pricing as well. So, when we look at unis swap x two market makers making 90% of volume, one infusion slightly better, three market maker making 80% volume and cow swap actually is the most decentralized one.

There's one top one market maker obviously we know the name is wintermute uh is making 25% volume and then next five indie uh solver I call them which means just like simply they're not market makers are making around 10% each and so you would maybe suspect okay might have the best price maybe comparing across all these models um so yeah actually one update on this data is that I checked it um just two days ago and in the last 30 days barter which is not a market maker now actually taken over winter mute. So I'm like really happy actually cuz like when we look at all these like me dex PBS data you realize all these market makers like everywhere and like they're kind of like

the problem of everything being centralized. Uh but I'm really happy like seeing one like non-market maker team is actually taking over the place of a you know like a market maker. So this meme is just basically saying like everyone's like these days when they see price go down it's like okay what are you doing are you dumping so I hope it's not going to be the end of the defi story. So um yeah we know a few other things about um these solver models in like a descriptive level. So when you look at like group by the liquidity type across cow swap in one infinus swap x you realize actually all these auction platform doesn't really possess um exact the same like pro liquidity provider and like composition right and so unis swap x in my opinion really was like a

strategy move for unis swap to incorporate rfq liquidity because they came out unis swap uh amm pool design and their universal routes are pretty only routes across unis swap v2v3 pool over the whole time in the past few years they didn't really have access to market maker so I think X really is a way for them to tap into RFK liquidity and at the same time K swap you know like started with all these advocate in like hackathon and have like indie projects to uh start building as a solver and so you simply see a lot of AMM liquidity on their platform instead um we also know overall that like um what um type of profile assets different liquidity likes to how to say market

make on um we know the professional market makers they would like to avoid like high-risk profile assets then easily you would see most of the longtail tokens are provided by AMN type of liquidity and uh when you go down to like the um major token type like Ethereum or BTC uh backed how do they pegged uh pairs there will be more and more presence of uh private market makers um So we finally did this like study to basically compare the pricing uh with a theoretical price uh or a simulated price. So we looked at six months of data across all these platforms and we did a simulation against the V2 and V3 pool for the most

how to say um highest liquidity pool for this asset. Obviously this approach is still very limited in my opinion where you can't really simulate the routing algorithm across um the current best like more than aggregators. So what we are measuring here as a benchmark really is just a sea level of naive routing across one biggest major uh like V2 or V3 pool. Um and also we're again simulating against basically the end of the pri end of the block price for last block or you can think it as like the top of block price for the current block. Um it's also kind of tricky if how you measure the gas assumption because like it can actually flip the final results and stuff. Uh but yeah,

here's are the results. So we picked also two major uh token pair as the symbolic pair. For the major asset, we looked at USDC and west pair. And this box plot kind of shows that basically uh cow swap 1 in unis swap all have basically positive uh execution quality wfare compared to v v2 uh amm pool and barely or maybe even slightly worse compared to v3. But uh we also compared that Binance markout. I wouldn't say it means that it's comparing with Binance execution because it's only really mid-pric comparison. But just as a benchmark as well um okay but if you look at this chart you would say okay

unis swap seems like even worse um than just go directly to V3 right? Um I would say not really um we really need to look into like the distribution controlling the trade size as well. So here we kind of see actually if we uh expand the charts by like trade sides all of them actually have a pretty good uh price improvement when the trade size goes bigger and especially go up to like 300k. Um but it a lot of those negative results were actually coming from the noise from small trade sides which is likely because of like you know gas inefficiency and etc. Um so yeah and then if you look at a V3 the results is very much less prominent um because basically V3 pool are very highly how to

say concentrated compared to V2. So it's actually much much harder for all these solver platform to compete over um even like naive routing of V V3. So yeah, we also then looked at the Pepe pair. Uh the results looks very different in terms of the shape, but overall it fits our assumption where um when you see the uh liquidity pool is very concentrated on B2, it's harder to compete it to have a better price than B2. But meanwhile V3 pool for Pepe is much less, you know, concentrated and that's where you see a lot of like price improvement for those.

Um so yeah, that was pretty much the conclusion. And if you are interested, feel free to check out the paper. Um so I actually have a lot more charts to go through so I'm going to speed up but um so what are the other new use cases uh growing on decks? Uh we look at this like whole chart trend across time. Um pretty much a year and a half ago when I was showing this chart we were very excited saying oh you know like Telegram bots become a new thing now but little a little over a year after you see they're kind of dying down particularly the only Ethereum um because really like when you look at Telegram bots volume and it's mostly on Salana and some are on base and actually Binance recently um but also overall if you look at the trend

they died pretty fast they come fast and they die fast after Trump and Libra uh um how to say peak around the beginning of this year. Okay. But also even though they died, the Telegram bots are still printing a lot of money. Um the biggest bot by volume today is Trojan bot and they are still making around like a quarter million every year uh every day and they're projects to around like 100 million annual revenue even on these flat days. Um so yeah another trend I'm seeing is that there is more and more integration of swaps in crosschain bridges and all these uh crosschain swap solutions. Um I showed this chart about like layer zero having all these crazy airdrop snapshot and then like die down right after. Um somebody actually

pointed out that this number might be wrong because um layer zero have a lot of like bridges under it. So for example like Stargate they actually are still the highest uh volume uh how to say bridge today uh per my check on deflama. Um so yeah and when I zoom into like the crosschain swaps data uh it turns out I would say most of these are still like stable stable. So even it's crosschain swaps it's still more like a bridging use cases as I understand today. Um yeah, so one last thing I want to talk about within the last one minute but very interesting is that uh we look we try to look into all the deck swaps uh profile into layer 2 as well but we realized actually spam is actually the

most important or basically pressing problems right now on L2 because they're cheap. Um so here are a few charts just to go through. Uh we know that base has been scaling up their gas throughput the whole time and now it's like 35 million gas per second throughput. But if you look how to say um respectfully and like if you identify all those spam transactions and you realize actually more than half of these transa uh gas usage actually are consumed by spam and even like um around end of 2024 to beginning of this year there's like a flat face where all these scaled gas usage are all used by spam. There was no actual scaled usage by real users transactions. Um so now you would think actually when at layer 2 measuring gas

throughput they're scaling they're not really scaling it effectively because of all these like meb problems they have to address. So the effective gas throughput here is really around like 20 million gas per second even today. Um so yeah we um just published this study yesterday and this is a problem on all these L2s not just base and whenever you have like TVL that's big enough and trading volume big enough there's MEV and spam problems so highly recommend uh to check out this post um and uh yeah that's the talk and uh any questions Please.

Yes. Yeah. Thank you very much for an interesting talk. Uh based on the data you've analyzed, uh do you have any insights on what in a DEX attracts uh users and trading volumes? Yeah, I I want to say um so first of all like um lot of these decks they are providing like also the API and solutions. So a lot of time it's a competition among integrating more front end as well. And I want to say I think the DEX pie overall is still very small and it's very narrative based because I I feel like a lot of time people just leaning towards the narrative. Oh okay un uh cow swap actually has better pricing because they have uh they the their model theor

theoretically going to benefit for bigger trades and you kind of basically also see accordingly a lot of like whale trades just come to cow swap. We see like Ethereum foundation also mentioned they swap on cow swap. Um so I think a lot of time it's still how to say doing research marketing and also integrated front ends. Um yet we don't know the truth to be honest. Yeah. Thank you. Yeah. Uh great presentation. The 90% of crosschain transactions of stable stable. Uh, do you think a lot of that's wash trading or you think it's real? I I doubt it's wash trading when there's no

token incentive, but I think it's possible that could be a like rebalancing use case for like market maker or actually like the solver who is trying to land like a crosschain uh how to say bridges and like fills. Um, yeah, I feel like a lot of those volume may be mixed with like solvers trying to move their assets around. But that's my assumption. I would love to see more analysis. Uh great presentation. Thank you. Um what would you need from uh from back end to make sure that there is no collusion between market makers and uh indexes,

right? Yeah. I think um I'm not trying to show TE but I think um I think there uh could be benefit from like if the how to say solver auction back end posting potentially what are the uh quotes they got um without anonymously without like mentioning who are the providers and potentially or some sort of benchmark base saying hey this was the highest price this was second price and um posting it uh like either on chain or some like data stream. Um I think that will be more credible at least. Yeah.

Yeah. I think that's it. No more time. No. Okay. Thank you very much, Danny. Thank you.