Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Ava Labs x CBER Ep 14: Price Discovery at Decentralized Exchanges
36 Plays4 days ago
Contrary to popular belief, price discovery can occur at decentralized exchanges instead of at centralized exchanges. We discuss the underlying economics and how this affects bidding for block space. We also discuss how private memory pools affect bidding behavior and the price discovery process.
Guests: Agostino Capponi (Columbia University), Ruizhe Jia (Stanford University), Shihao Yu (Singapore Management University)
Paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4236993
Recommended
Transcript
Introduction to Blockchain Podcast Series
00:00:06
Speaker
Hi everybody and welcome to Crafting the Crypto Economy. I am Silvia Sanchez from the Avalanche Policy Coalition and today we bring you a special podcast series in partnership with the Crypto and Blockchain Economic Research Forum, also known as the CBER Forum.
00:00:22
Speaker
This is season 3 of 5 new episodes featuring leading faculty from global universities exploring various elements in the blockchain ecosystem. From stablecoins to DAOs and so much more.
00:00:33
Speaker
These episodes are a bit longer from our usual ones since we will be getting deeper. But again, it's kind of challenging to fit in a whole research paper into just one hour. But don't worry, each episode will have its accompanying research paper posted on the website and episode description for further reading.
Moderators Take Over: Blockchain Paper Discussion
00:00:49
Speaker
Anyway, I'll hand it over to our moderators, Professors Fahad Saleh and Andreas Park. We hope you enjoy.
00:00:58
Speaker
Hello everybody and welcome to another edition of the Crafting the Crypto Economy podcast series. I'm your host Andreas Park together with Fahad Saleh from the University of Florida. um We're now in our third season and we're very happy to have repeat offenders on our podcast.
00:01:13
Speaker
um We're talking about today about a paper which has been written by Rujo Jia and Shihao Yu. Rujo has recently been appointed to Stanford University's OR r department, so congratulations again on that.
00:01:29
Speaker
And Shihao is at Singapore Management University. The third author author is Agostino Caponi. who is not with us today. um He's professor at Columbia University.
00:01:41
Speaker
um Guys, welcome to the podcast. We're very happy to have you here. Thank you so much. Thank you for having us. Thank you so much. Happy to be here. Excellent.
Decentralized vs Centralized Exchanges: Key Differences
00:01:49
Speaker
So today we're going to talk about an empirical paper that he wrote on price discovery on decentralized exchanges.
00:01:56
Speaker
um So I think there's a lot of things already in the title that we may have to clarify for the audience. um Let's start with maybe something simple. Can you explain to us and to the audience um how a decentralized exchange works and how it is different from a centralized exchange?
00:02:14
Speaker
Yeah, so yeah we can start from the title. So basically, decentralized exchanges are exchanges deployed on blockchain. So for example, we know Uniswap on Ethereum blockchain or on L2s, for example, Polygon, Arbitron, Optimism, etc.
00:02:31
Speaker
So because decentralized exchanges are deployed on blockchain, they inherit the infrastructure of blockchain. So one key difference we want to emphasize in the paper that sets the decentralized exchange from centralized exchanges So if you look at, let's say the major exchanges in the world, right trading equities, for example, LIC, OIC, LasDac, almost all of them converge to one single model.
00:03:00
Speaker
That is the continuous central name order book model. So basically trading is continuous, right? So whenever it gets to the exchange, get to trade without any delay.
00:03:11
Speaker
Whereas on blockchain, we know that blockchain process transactions in batches or in blocks. For example, in Ethereum, we know there will be a block produced every 12 seconds.
00:03:22
Speaker
So basically trading or transactions are not processed continuously on blockchain or on decentralized exchanges, but they're in batches. So that is the first, we think, important difference between decentralized exchange and central exchanges.
00:03:39
Speaker
And another key difference when I emphasize the paper is we call in traditional finance, TrapFi priority rule and DeFi, we call it order sequencing rule.
00:03:50
Speaker
That speaks about how orders are sequenced or traducted a particular order. So on central exchanges, we know because trading is continuous, so basically who is fast enough to get their order to the exchange in the first place will get to trade.
00:04:06
Speaker
Whereas on blockchain or on decentralized exchanges, orders that arrive in the same batch will be prioritized based on the traction fee or on Ethereum for example gas fee that the traders are willing to bid.
00:04:20
Speaker
So basically the miners or validators will look at all the tractions received within a block and then to order tractions in the descending order of gas fees.
00:04:31
Speaker
So whoever is willing to bid a higher fee gets to trade first. So two key differences again. First, trading on decentralized exchanges is in batches or discrete times, whereas on decentralized exchanges is often continuous.
00:04:45
Speaker
And second, the order sequencing rule is very different. On centralized exchanges, it's a first-come, first-served. who Who is faster gets to trade first, whereas on blockchain or decentralized exchanges, whoever bids a higher fee,
00:04:59
Speaker
where we get to trade first. So these are the two key differences that we want to highlight in our paper that actually sets these two different types of exchanges at apart. If I may, I can add the third one we care a lot about.
00:05:11
Speaker
So just in summary, I think Shihouse has the biggest difference, actually infrastructure.
Infrastructure and Speed: Centralized vs Decentralized Exchanges
00:05:16
Speaker
You go from a centralized entity around data center goes to a blockchain.
00:05:23
Speaker
And there are three key differences in my understanding. And the first two are covered by Shihao. The first one, you're on a block by block settlement. Second one, you arere actually have a different priority rule compared to traditional exchange.
00:05:35
Speaker
But also the third one is also quite important. You can see different things. There is way more transparency on the decentralized exchanges. You can see who does what, how much they're happy to pay for their immediacy, which is how quick they get the order executed.
00:05:50
Speaker
You can see which order has more impact in the market in the later. So all those things, and you can tie to that specific person. So the transparency also plays important role, not only in our paper, but also in the whole decentralized exchange system.
00:06:04
Speaker
So that's the third thing want to add. So let me chime a little bit in here. So ah so um my research area is actually also trading. um So shocker here. um I just want to maybe for the audience, for the benefit of the audience, maybe also point out a little bit how, generally speaking, crypto exchanges, centralized and decentralized, differ from traditional exchanges like, say, the NASDAQ and the New York Stock Exchange.
00:06:26
Speaker
um Because I think there's some subtleties that can be quite important to understand. um in In particular, when you think of centralized exchanges like NASDAQ and NYSEE, they both have big data centers that sit somewhere just off a highway in New Jersey.
00:06:42
Speaker
And people choose to put their computers there to trade as fast as possible. And then they try to connect these marketplaces with usually very high data very very very fast data connections based on microwave towers or sometimes even on lasers.
00:06:56
Speaker
just so they can get the um just you can get the data and the trades from one exchange to another as quickly as possible. Now, decentralized, and not decentralized, but crypto exchanges actually work differently because they're actually not located, as far as I know, on a known data center. They are located in the cloud, which changes the calculus for even for these high-frequency trading firms because they cannot actually put their servers right next to a server and on Google Cloud. They're simply you know physically not possible, at least not... um
00:07:27
Speaker
at least not persistently. and So that's actually very interesting because that changes actually the speed game nickpher and significantly. And then the second thing is that um and ah you know that if you think about it, so in traditional markets, people um try to be faster at the exchanges and they're essentially paying for it with a fixed cost, which is the speed that they invest in.
00:07:48
Speaker
What you're, I think, telling us here is that in a decentralized market, you actually do this more on ah on a case-by-case
Price Discovery Mechanisms in Crypto Markets
00:07:54
Speaker
basis. So you're basically trying to submit a trade and then you're trying to increase the bids that you pay for it to be processed first so that you can take, say, advantage of any particular trade as fast as you can.
00:08:06
Speaker
Now, no i have i have submit I have a lot of questions when we when we talk about this, but maybe we Before we you know drill into all the details, we should have some conversation where you tell us a little bit about what it is that price discovery actually means for you and what it means in in the context of, say, crypto markets.
00:08:25
Speaker
I'll go for that. But also, I just follow up on your previous comment. I think the speed game... in crypto now goes to, as the infrastructure evolved, goes to a different phase. Like, for example, now you have layer two, which currently have no no longer a very decentralized system. They have a sequencers and they're relatively quick.
00:08:45
Speaker
So sometimes the game actually went back to the game used in traditional exchange. So they actually have co-locating, figure out where the AWS service is They try to be close with them and they want to be executing transactions on-chain with smart contract to speed up the whole ho game.
00:09:02
Speaker
So they want to, so yeah yeah that's just a comment I want to add. So the game also has evolved. um But yeah, so in terms of- Yeah, we have to we have to talk about the layer two solutions too actually in some detail because we have lots of questions about that. But let's let's just go and you maybe explain to us but what price discovery actually means first.
00:09:18
Speaker
So this is the other mystery in the title. Yes. So for press discovery, think ah one sentence is how information is incorporated into the price.
00:09:31
Speaker
But on the other hand, to me, is really how the market forces buy and sell, intention to buy and sell, cancel and add new liquidity eventually help us to shape a value of the asset in the market.
00:09:43
Speaker
And that's quite crucial for everything you do. it helps you to see um what is a fair price for people to transact now. And then there you also have to to see how do you efficiently allocate resources.
00:09:57
Speaker
for For example, if you trade something like resources or any financial asset, people are based on the price to redo their portfolio and so on. Or for someone who are retiring, they're based on price to do other things.
00:10:11
Speaker
So I think price discovery is crucial um in any market, including financial market. and then If I could just, can I can i interrupt for one second? So um when we have prices covered in a traditional market, it can happen essentially in two ways, right? It can happen because somebody submits trades and it can happen because people change their orders.
00:10:30
Speaker
So maybe you want to, well, so what's, is this the same mechanism that happens and in DeFi or on Dexas or is there a difference here? So um I think there, this is also a great question. Shahal, let me answer that. I think In decentralized exchanges, there are many different exchanges.
00:10:45
Speaker
So we're looking at a specific one called automated market makers. In that specific exchange, ah the only way to move price on that exchange is based on trade. Whenever trade happens, they move the price in this specific ah exchange.
00:11:02
Speaker
But then if you think about maybe now these exchange have been evolved, there are hyperliquid, asters, all those things. And then for those exchanges, they're more like limit order book a different structure, then their price discovery rate can happen very differently. It can happen by people at removing quotes.
00:11:19
Speaker
It can happen by people which is they add and remove liquidity, can happen become trade. So like I think there is a distinction here, which um for our paper, the trade is what's important on these new exchanges, but now it has evolved different forms.
00:11:33
Speaker
All right. So um tell tell us what have you found about price discovery? How does it actually work? I mean, so we talked about it goes by by but by trades essentially. And so I think you have found, i mean, you drilled really deep into the data, right? so So is there any way, I mean, is there by the way, do you have actually ah any comparison to centralized exchanges and the speed thereof, or is it exclusively what you have um within in a particular decentralized exchange, or automated market maker, or is it across various different ones? yeah Let me first start with the general picture and show how we'll into details.
00:12:06
Speaker
So I think for us, we look at both central and decental exchange exchanges. For example, if we say decentralized cause price discovery, what we really mean is, for example, let's if you trade, because if you think decentralized trade have this price discovery for themselves, kind of mechanical, if there's a trade happen, then decentralized exchange also move based on that trade.
00:12:30
Speaker
But we actually link both central exchange market and digital exchange market, central exchange, which is Binance. And then we can see that actually ah the one big its takeaway is people typically think ah in the crypto market, central exchange is the one drive almost outpriced galleries. In a lot of theory paper, they think,
00:12:50
Speaker
Central exchange, there is price go away. And then what happens is that people do arbitrage and then move this exchange this exchange. I think in general, our one message is actually that's now the full story.
00:13:03
Speaker
You can actually see these central exchanges trading flows also push the price in all market. um they're So basically, there are also some information originated from this interaction itself.
00:13:18
Speaker
So that's kind of one misconception, one first correct. And then, Shahal will get into more details of what do we find from price discovery. So you're saying essentially, i mean, we would normally, you know, if you could think that a market like Binance, Kraken and so on, they're relatively fast, right? Because they are, um i mean, they're not really...
00:13:38
Speaker
They're only loosely connected to the blockchain in a way, right? Because they're really just a laptop sitting on top of it. Nothing that happens on Kraken and on Binance has any bearing on what happens on the blockchain, right?
00:13:48
Speaker
um but And blockchains are slow by design, right? So in Ethereum, all blocks are created only every 12 seconds. um And you're saying essentially that even so, we see that there is price discovery happening on the slower market, on the design slower market.
00:14:05
Speaker
So How do we interpret that? Should we think about it as people make their trades first in in automated market maker and then they go elsewhere? or um Because I mean, i'm I'm always trying to think about this as in terms of the the legacy world of finance, right? Where you have a choice between various different markets and you try to split your trades in such a way to minimize our impact essentially, right?
00:14:27
Speaker
Or if you have to trade a lot, then you want to get out of the door and be be done there and be first. I can answer and show how a compliment and show get into more detail of price gallery. So I think you're asking why price gallery, like, do you have any rationale why that happened in this drugs exchange?
00:14:42
Speaker
I think
Motivations and Strategies in Decentralized Trading
00:14:43
Speaker
there's economic concern. There is actually non-economic concern here for people do that. Economically, ah you can actually see for many humans, there is actually more liquidity on decentralized exchange.
00:14:57
Speaker
And then so when people want execute, they are caring about their cost and liquidity make them have less price impact lot of time. Second, I would think that for sophisticated players, they sometimes just use both places, right? It's just like you have 18 exchange options.
00:15:15
Speaker
If you want to get all of the quid, you want to send synchronized order across all of them. And then that's actually the same for centralized disinformation. They're also somehow happening um in the same time, but then randomly one is quicker, one is slower. So there are second possibility. So if the first one is liquidity, some is indeed better, so okay like might maybe some people just use both.
00:15:37
Speaker
And third one, I think, sometimes actually cheaper on decentralized exchanges. And then for non-economic reasons, I think there are two, at least in my mind. First one is... um No KYC, right? For example, if you are a hacker, what you typically do, you quickly switch using decentralized exchange to trade all of the assets you hold or you hacked.
00:15:59
Speaker
So no KYC is pretty big in decentralized exchange. So that's maybe something lawyers care about. But second thing I think is equally important, probably more important to um the world is composability. If you already like trade on decentralized exchange, you're already doing some decentralized exchange, trading on DEX actually also not that bad. You just have to click a button and then you trade. But then if you think about you want to actually move that to a centralized exchange, as you said, someone's laptop, you have to do all those processes, send asset around and start trading. So as in composability, some convenience
00:16:33
Speaker
also play a role here for the non-economic concerns. So that's kind of how frame it. So I just want to quickly chime in because there's ah this is a few things that I want to unpack from what you said. So first, I think we should, if possible, move away from from the from the illegal hack and so on narrative, because you know that's actually probably not very helpful because we we want to see this as a natural market and what really happens, you know maybe beyond what is specific to blockchain.
00:16:56
Speaker
um Then the second thing um i want to just mention, there's another paper actually by Angelo Ronaldo and Andrea Bourbon and they actually have done the analysis of trade sizes and I think maybe for the benefit of the audience, they found that trading on a decentralized exchange, particularly for large orders, is often significantly cheaper, right? So they've done the work on various markets, various times and so on.
00:17:18
Speaker
So what you just said is there's actually also already published evidence on on these differences, right? So don't have to speculate on that. So, um So, okay, anyway, so, Fahad, I think you have you have a question here.
00:17:29
Speaker
Yeah, so before we get into the depth of your paper, I want to sort of properly understand maybe the first part of of your answer, Rija, specifically about kind of the economic channel. And I guess, so one thing that I'd like to understand is sort of what is the typical impetus for this ah trading? Because so one mental model would be something like okay, essentially a liquidity trader, I have a need to trade, and I'm just going to think about the all-in cost of execution in each of these venues, and I'm going to, well, I could even split, as you said.
00:18:00
Speaker
um But, you know, you've alluded to price impact, you alluded to splitting, but ultimately, that seems to be an exercise, you know, a mechanical, mathematical exercise almost of, like,
00:18:11
Speaker
given that there are these two venues and given that I need to trade a certain amount, I'm just going to sort of split my trade as optimally as I can to get the lowest cost execution overall. um Another sort of model is more of not so much a a liquidity trader, but actually an informed trader. And, uh,
00:18:27
Speaker
Obviously, that's very important to what what you're what you're working on here. And in that context, though, you know, there's this, you know, there's's there's, let's say that there's certain ways you could write a model where you would find essentially the the informed trader will trade both venues um such that the sort of marginal cost of trading at the end of the trades ah aligns with the the sort of, let's say, the fair value less the transactions costs at that particular venue.
00:18:55
Speaker
um So could you maybe flesh out kind of um what is the right mental model here? Should I be thinking of informed traders? Should I be thinking about liquidity traders? So when you're talking about price discovery, I'm usually thinking about informed traders, I guess.
00:19:09
Speaker
But but there's all it also feels like there's there's probably some more complexity than just thinking about like, oh, I have some information and now I know the fair value is this particular number and so i'm going to trade to it.
00:19:20
Speaker
um Yeah, so so maybe I could say a little bit about the impetus for trading and why um there might be more complexity than just this idea that prices should trade to fair value, less costs of trading.
00:19:36
Speaker
yeah Yeah, thanks a lot, Fahai and Andreas. Four great questions. um I think I'm very happy you brought up informed trading. We've been talking about price discovery, right? Just want actually put a link between price discovery and informed trading.
00:19:51
Speaker
So as Rita mentioned, price discovery is the process of how information got incorporated into the prices. just want to mention it's not like magic, right? In order for that to happen, we need somebody to trade on information.
00:20:03
Speaker
So essentially we need informed traders to actually trade in order to move the prices to reflect the information they have, right? Essentially move the price to the fundamental value that they believe the security should have. um So that's why a lot of part of the paper is about how we should actually think about, or as Fahad said, have a mental model of how informed traders should trade, right? um ah Especially on decentralized exchanges, as we already said at the beginning, the trading mechanism is are very different between the two centralized and decentralized exchanges.
00:20:35
Speaker
But that's the first thing want to actually mention. and So we actually think more in the shoes of informed trader. And second, want to go back a little bit to Andrew's question. um i think when you talk about price discovery or information,
00:20:50
Speaker
There are many different sorts of information, right? You can think about long-term, short-term. And I think you were saying that Binance is much quicker, right? It's kind of off-chain trading. It's pretty much like a decentralized exchange in traditional markets.
00:21:03
Speaker
So of course, I think if there's a type of information that's a super short-lived, right? Think about public information. For example, if there's a news breakout. And then everybody wants to trade on the news.
00:21:15
Speaker
And I think in that case, you might just go for Binance, right? Because that's where you get to trade much quicker compared to additional exchanges, at least on Ethereum, right? We have to wait a few seconds until the next block. Shihau, can I just interject there just to clarify a point? Because you said, for instance, like you what you just said sort of had a dichotomy in it. It was like, okay, so in this case, I'd rather trade in Binance. But if I have information...
00:21:39
Speaker
And both the decentralized exchange and Binance are not pricing it yet. Why don't I just trade on both? Yeah, I'm not rooting out. Yeah, I think the best model would be with the order splitting, right?
00:21:49
Speaker
Because if I just concentrate my order on one, I will create a larger price impact than compared to just splitting into half and half and trade on both. So totally agree. I think the optimal execution for you might be actually trade on both.
00:22:01
Speaker
But wait, wait, sorry, can I interject? but So even order splitting sort of seems to presume a certain order size, right? So what I'm saying is like one mental model is that, you know, I know the value is 100.
00:22:13
Speaker
ah Both of them are trading it somewhere underneath that because there's just some, you know, positive information that I'm aware of that the market isn't. And so I'll trade both up until... the marginal cost of trading at both venues. and And in that sense, like that mental model is so simple that it sort of detaches the centralized exchange, the decentralized exchange. In fact, it detaches all venues. It's like, look at any particular venue, I would just keep trading until, you know, essentially.
00:22:37
Speaker
and and so ah that's what I'm trying to clarify about the mental model. It seems like your mental model is a little more sophisticated than what I'm outlining. And I'd like you to sort of flush that out. Right. I think that's based to, again, I think to the different trade mechanisms between central exchange and decentralized exchanges.
00:22:51
Speaker
So to us, if you're an informed trader um and you want to trade on central exchanges, I think the execution strategy will be different compared to trading on
Informed Traders and Market Dynamics
00:23:02
Speaker
decentralized exchange. For example, on decentralized exchanges, you have to think about the optimal gas fee bidding, right? How would you actually bid the optimal fee to prioritize your transaction?
00:23:14
Speaker
we Whereas on central exchanges, we don't have such VBA mechanisms, right? Maybe you can choose your order size, how you maybe gradually fit, sort of split your parent order into a small orders and gradually fit into the market to minimize price impact.
00:23:30
Speaker
So, I would say just because the trading model is very different, even though I think you might have the same fundamental value when we I should trade on both until the price get to that level, but still in the process, you might trade differently on these two, right? One, have decentralized exchanges have to consider the fee bidding, decentralized exchanges don't have that.
00:23:50
Speaker
I'll add base more target to what the has. I think what you have in your mind as a mental model, like works pretty well for maybe 60-50% of the reality, but then I think in the true reality, though like the the word is too much too much too too complicated. For example, what you essentially see I have information, if that information is known one to pretty much everyone in the world, maybe my ultimate thing is just send a big market order to Central Exchange and then send another swap, which is essentially like a market order in Central Exchange to the marginal price that my cost is eventually equal to my marginal gains.
00:24:27
Speaker
But then in reality, you think about what if you know that you may not be the only person knows that information, but then you know like, but you're maybe not everyone knows that information, but maybe then maybe I can slowly position in, right? I can instantly change trace slowly.
00:24:44
Speaker
And then in this exchange, can send a small order. People will kind of do like they think I'm just a retail doing trade. They can reverse my order so that they always have a good price to execute.
00:24:55
Speaker
So like in reality, I think it's very unclear um how people who are sophisticated, who are informed, would, how would they position in and how how would they feed that information to both exchanges?
00:25:10
Speaker
I think it's hard to, our paper eventually, after show houses, like answer part of the question, like they actually do care about like how quick they get their order down, they actually have those concerns. But overall, I think there is no mental model just like perfectly explain how informed trader does. And there is a lot of small details for them to source as much liquidity as possible.
00:25:32
Speaker
And there are many concerns they care about, as we'll go into later, like who is competing with them, how many other people know that are only going to be selection bias. And then all those things will be quite important when you have a mental model. So eventually we we'll give you mental model to explain some of the results, how informed traders behave.
00:25:51
Speaker
Does that answer part of question? Yeah, I think there's so many different ways how you can think about what why people trade, right? i think it's I think it's more important that, I mean, what you're really after is you just look at, you know, people do things and therefore the but and then the price moves thereafter, right? I mean, that's kind of what you're after, generally speaking, right?
00:26:09
Speaker
So, you know, I think in some sense you're very model agnostic in a way. Is that maybe fair to say? Yeah. and So how about, okay, so let's let's dig into your paper a little more. So what exactly is it that, what is so let's let's let's outline your key findings, right?
00:26:24
Speaker
Here we go. So I think the main takeaway for our paper is, so we do find actually informed traders, they actually bid higher fees for the execution on taxes.
00:26:35
Speaker
In the first place, you might actually find kind of intuitive. But if you think about it, there are like two features we think about DAX, right? First, of course, there's a fee-bending mechanism. Informant traders will would be the fee to do their trades.
00:26:48
Speaker
And second is the transparency. Right. So when you are actually trading on DEX, you have to balance two things. The first is actually, if you beat a higher fee, of course, you can get your execution ah quickly, quick, uh, quicker.
00:27:02
Speaker
And at the same time, when you beat a higher fee, your fellow competitors can also see that, right? If you're connected to the P2P network and if you observe the, uh, what's happening the main pool, you know, there is a order that attach a higher fee, uh, uh, uh,
00:27:17
Speaker
but for for for the trade. So I think that's pretty much a sign that but that there's a higher willingness to pay by some trader. And that's very likely that the trader is actually informed or trying to trade for profit and some information.
00:27:30
Speaker
So basically, when we actually take our analysis to the data, what we find in the end is we do find that informed traders, they bid higher fee, right? So they actually care more about the execution priority versus the sort of risk of leaking their information.
00:27:47
Speaker
So can I chime in again? yeah As you know, um so I have a particular interest, which is the MEV extraction. um And just for the benefit of the audience, what I'm referring to here is as follows, so is that trades, as they generate it, they generally enter a mempool,
MEV Extraction and Its Impact
00:28:06
Speaker
right? So you you you submit your transaction to ah to a network and then um This can be viewed from the outside world and then other traders can see trades and then they can try to, ah let's say, benefit off of these trades. Could be in a variety of different forms.
00:28:21
Speaker
But the key part is is that the mempool is a source of information and then that there are other people that may want to ah submit orders based on information that they offer extract from this mempool.
00:28:33
Speaker
So that's number one. Number two is, of course, these other traders. They ah usually have a particular competitive technology advantage, and they can generally also submit trades really quite quickly and get them included in blocks.
00:28:45
Speaker
So now I'm looking at my my my graphs here. So, you know, there's a nice website called The Block where you can look at trading data, and you can see that actually a very large fraction of trades are generally submitted by what's called MEVBots.
00:28:58
Speaker
that be agnostic to what they do, whether they provide a service and all. But it seems like these are very particular type of traders that that are very prevalent in index trades. So is it possible that the ones that you a catch, if you want, as informed traders, that these are actually precisely the MEV bots that try to either take advantage of quick arbitrage opportunities with decentralized exchanges or that they just try to front run or in any way other take advantage of other orders that exist already in the mempool?
00:29:28
Speaker
Yeah, maybe i can get to that question. Yeah, a great question, Andrea. So in our paper, we spent actually a of efforts trying to kind kind of not identify those arbitrage trades that you actually mentioned in our analysis.
00:29:43
Speaker
The reason is actually we care more about the private information part of it. The reason is actually we know that, especially for Uniswap B2, right, there's no price of discovery on that platform per se because the credit providers cannot control like the quotes, quote-unquote, of the security.
00:30:02
Speaker
So basically, all the price changes have to be done through trades. So I think there is a kind of understanding from the traders that the Uniswap B2's prices kind of always lag behind Binance, right?
00:30:15
Speaker
That creates the so-called centralized and decentralized exchange arbitrage trades, right? Essentially, there are price differences between these two and arbitrage becoming trying to profit from the difference.
00:30:26
Speaker
So We try not to capture those trades because we think this kind of price discovery is kind mechanical, right? just because the price on DAX will lack behind Central Exchange. When you see the price actually converts to Central Exchange, that's already not new information, right Because we know that the Binance price has already moved.
00:30:45
Speaker
and The reason why Uniswap price still lacks behind just because it takes time for the arbitrage to actually bring the two prices aligned. um What we really care about is actually when you have private so Just one out of curiosity, have you ever looked at whether or not the the arbitrage trades that you just described, are actually other is it possible that they're actually done by the liquidity providers themselves? um So they're trading against themselves?
00:31:10
Speaker
um just It's just out of curiosity, may not have looked at it, but it seems to be something that you could easily imagine that you would want to do. We didn't look at it, where even though we identified all of them with our best effort.
00:31:21
Speaker
But then we actually never look at whether they are ah taker or also the makers. Or are they really trade against and themselves? We never look at that. um But then Shehal get into more in into detail.
00:31:33
Speaker
ah essentially, just give you two summaries. First, we want to only look at how the student exchange flows impact future price, not only the existing price on the student exchange, otherwise it wouldn't be called price gallery. I think that's first point Shehal is trying to say.
00:31:51
Speaker
And second thing is we actually did, i didn't we're going to do some robustness check, just as you said, like you don't want to, all your result is actually based on something completely mechanical, you misattribute that price gore. So we end up trying to identify all the MEV trade where with our best effort.
00:32:11
Speaker
with like sandwich attack, central X tax arbitrage. And we try to remove that from the decentralized trading flows and to see how that impact the price. So our hypothesis that if they're indeed the biggest driver of our result, where we find decentralized price gallery, then after we remove that, those effects should at least weaken or change.
00:32:32
Speaker
And then you actually see that's not the case. So maybe those are more responsible for contemporaneously, decentralized decentralized exchange, where something happened there. But then for the main result, which Shi Hao says, we're looking to see how digital exchanges move the whole future of both digital exchange and digital exchange.
00:32:51
Speaker
Those are actually not impacted by those MEV trades. I think hopefully that answers your question. Okay, interesting. Okay, so you're yeah you're really focused and zooming in on this particular subset of the price moving trades that that really predict the future.
00:33:04
Speaker
Yes. yeah Plus you want to avoid something else may potentially make you have a misdittribution. Sorry, can I just clarify though? So, um okay. So yeah, you violated very clearly that you're not sort of focused on Chextech, Arb and so on, right? Sort of you're you're really digging into price discovery in a meaningful way.
00:33:25
Speaker
um But just from a sort of from back to that mental model sort of thing, what are the sorts of ah things that sort of correspond to the information that that that these people might be trading about? So I'm not necessarily saying um I'm not so much asking you about your data exercise, but I'm just more saying, like can you give a concrete example of the type of things that you think might be serving as the private information that is then causing these people to trade?
00:33:53
Speaker
that makes sense. For example, we all just one example, which I probably know is better, which is big information may come from you can predict flows and you can predict predict returns. right If you, as a sophisticated player, you know there's more flow coming into certain asset, you pretty much know there might be price change. Or there are other features, or we call other factors that may affect price.
00:34:22
Speaker
Then there are people who are very so sophisticated, sophisticated, have statistical signal no to say, oh, that's the market may move. That is one way to generate information, even though it's not the only way. And then people may position in for those kind of signal they have, and then they can express that view through trading in both markets.
00:34:41
Speaker
So that's kind of one example of what information is that's generated by statistics signals and based on some real um economic factors or real market factors.
00:34:53
Speaker
But there are other um informations as you can imagine for them equity analysts have also has informations. their information for alternative data and so on. But I think that's a general picture of... Just to clarify question, though so um like for instance, generating signals, is that gonna be however based purely on public ah data or like, is it so is there like a real private information aspect or is it more like I have, you know I'm a hedge fund and I have better skills and so I'm able to kind of extract information from public data that is... I think it's both, right?
00:35:29
Speaker
ah Price volume data still have a lot of information, even for equity, such an efficient market. People can still generate prediction if you have good skills, maybe with better machine learning, with something else.
00:35:40
Speaker
She has a new machine learning paper. On the other hand, like there are people using alternative data. They want to say, let's don't fight the like the the fight for competing with better skills. Let's find Signal, have the best analyst who have a view on the market, and then let's put that into a bet.
00:35:57
Speaker
So like there are just very different information source and there are people using alternative data, right? Like people use satellite, credit cards, FinTech firms' data. So all those things can potentially drive information. So, but yeah, I think like all of them are possible if that answer your question.
00:36:12
Speaker
Right. um If I can just add one point, I think financial markets, I tend to think like there are maybe to different types of information. There's news, which we often label them as public information, right? So basically that's type of information that everybody can see.
00:36:29
Speaker
And the second, I think it's more kind of traditional, traditionally people think as private information, right? If you think about equity trading, if you do some fundamental analysis on the company's financial statements or their operations, you might generate some private signal to yourself, right? if you think the company's actually do well, can actually trade on stock.
00:36:48
Speaker
And I think that some information between, for example, the order flow information is kind of interesting because and if you're on a centralized market where everybody trivates each other in the same place, if you can buy the order flow right from the exchange, for example, the only one central exchange, then the order flow is also kind of public information, right? So if you subscribe to the data feed, you can analyze the order flow, you can actually, based on that, generate some training signals.
00:37:14
Speaker
But in reality, trading is often very fragmented, or especially in crypto. We have central exchanges, we have decentralized exchanges, we have a different decentralized exchanges. And just want to bring in, for example, a bit more details. Currently on Ethereum blockchain, we have the PBS, right, the proser-builder separation.
00:37:30
Speaker
And I think a lot of builders, they acquire private flows from some searchers. So basically, those flows are only seen by one particular builder who has connection with the searchers.
00:37:42
Speaker
So I think if you only got if you're the only builder who gets to see the order flows, they also can get some private information from order flow only. right As we just said, if you know how to translate order flow information into training signal, I think you can also profit on that.
00:37:58
Speaker
We now open another Pandora's box there with ah with a new jargon. ah so I'm not sure if you want to go down that
Blockchain Infrastructure Evolution and Trading Strategies
00:38:04
Speaker
route. I think we want to come back to your paper a little bit more. so Yeah, i just want to summarize, recap what Shehal says. Basically, um what happens is with there are a few findings. We find that if we see Dex have price gallery, right then you want to see, oh, where that is, like, give us more details. So we actually see lot price gallery happen around people who are paying for high fees. They are trying to be executed executing quick.
00:38:31
Speaker
So that's the first thing Shehal mentions. And then they actually don't bid just high fees. They bid way more than ah the median people base. So they're actually quite serious about that for pay for immediacy.
00:38:42
Speaker
And second thing, I'll tell you more kind of a mental model, which PayPal cares about. And second thing is we go really, you look at people who actually bid high fees you you think they might be informed traders and then basically you divide all the traders by historically they pay high fees and low fees and you look out of sample.
00:39:01
Speaker
You look into the future and see if they indeed are quite informed about the future price. So you actually see that's actually the case for people who in the history have pretty high fees. In the future, which is completely out of samples, you can still see there are the people who are very informed who who move the market.
00:39:20
Speaker
So like there's also like something ah like basically, high-feed flows are informed and ah information in high-feed flows, or there are separation here. And last thing before, butha like I think we also have a mental model for that.
00:39:33
Speaker
So think about, like there is a, I don't know like how many people in the audience know the paper by Hao Xiang in MIT on Darkpool. They think about Darkpool has blockchain bias, people, if they're in the same direction, the crowd market and so on.
00:39:47
Speaker
So you can think in blockchain, there is something for similar, right? If you're informed trader, conditional on you want to trade. There's a selection bias in the sense of if you think some people are going to compete with you, whether you are the first thing to trade matters a lot because the person before you are likely being in the same direction of you.
00:40:05
Speaker
And then if you're behind that person, you may have much worse execution. But for more most uninformed people, think about when you're about to submit orders. And then people ah before you maybe say, like um they can be informed in that direction with opposite direction. They can be informed in different directions.
00:40:23
Speaker
Maybe the selection bias and not the problem for you. So eventually, like you can see that if you have a selection bias model, mental model, you can see like that actually has a motive for informed trader to be more serious about avoiding that and bid high fees.
00:40:38
Speaker
So that's kind of like summarized what Shihau says about our paper um yeah before we move forward. have I see two hands. Yeah, go ahead. Can I? So this is actually a small question, but it's just a clarifying question that I'd like to sort of get out, given to riha what you were just saying. So um obviously the Ethereum blockchain changed their transaction fee mechanism, I think, September 2021. Right.
00:41:01
Speaker
And I think particularly you were referencing sort of how much higher, for instance, these informed traders are bidding than, let's say, the median ah ah transaction fee.
00:41:12
Speaker
i So I was wondering whether you could talk about a little bit about how you see that change in the transaction fee mechanism affecting what you're finding. And actually, before you say that, i actually want to say something, if if it's okay for hard. I think this is an important its discussion to have, but I actually want to i want to actually may point something out here, which is, um you know, there's a lot of narratives that go around traditional markets about whether it's fair that some people are faster than others and how this takes away opportunities from others and all.
00:41:42
Speaker
I just want to point out, I mean, one of the things that is and strikes me about the crypto market, it is an incredibly capitalist a way to to deal with the market as a whole. but So here, I mean, what you're basically describing you have people that want to get their trade done and they're willing to pay for it. right So they're willing to pay for the priority, which is kind of how we normally think about how how stock markets work. This is why people pay for what they call immediacy. So they pay a fee, usually to liquidity providers.
00:42:10
Speaker
And I think here it is very noticeable that they're willing to pay a fee to liquidity providers, right? So because they they have to pay more for their trade. Okay, so number one. Number two, they're also, as I understand it, they're also willing to pay higher fees to just get their trade processed.
00:42:24
Speaker
So instead of making an investment in speed technology, they're making an investment here on ah on ah on a trade-by-trade basis. And again, so they're basically, on you know, on a transactional sense, just willing to put more money down in order to get take advantage of it.
00:42:39
Speaker
which in the blockchain world has an interesting side effect, which is that it doesn't go, ah that it goes actually to the parties that maintain the system. right So they're contributing in this way to the security of a blockchain, which I think is something that's worth noting here. um Anyway, and I think this goes a little bit into also the trend the the discussion that Fahad wants to have of of priority fees and so on. so but Actually, but before before before you answer, though, I do want to add a little bit there.
00:43:06
Speaker
So we were using priority a lot, right? And so one of the, I mean, but so in some sense, the big point of EIP 1559, the exchange transaction fee mechanism, was to distinguish priority across blocks versus priority within blocks. And so part of what I'm curious about is like when you're talking about people bidding up,
00:43:24
Speaker
is that, you know, in a pre-EIP 1559 world, ah that that could just be like, oh my God, all these like crazy NFT, whatever collections are there and people are are sort of, you know, jamming the network with it. And I really want to get on this block um ah because because the literal time, you know, immediacy is important. And so I'm going to bid really high just to make sure I do get on the next block.
00:43:47
Speaker
um But then essentially EIP 1559 was sort of supposed to take care of that. And so I'm I'm i'm really trying to understand kind of why the bids are are so high. And this also relates a bit to the conversation we werere having earlier about, for instance, like what is the private information? Do other people have it?
00:44:02
Speaker
Why don't I just bid to the point that, you know, the the the sort of the the marginal cost equals the the marginal value. um But yeah, sorry. So, Riju Shihau, you all should take over.
00:44:13
Speaker
will say one thing I should talk about how I deal with EIP559. So just to target your question, I think, there is a difference between being high enough to catch this batch, catch this block.
00:44:25
Speaker
And there is a difference on which position you're in the block, right? The whole um selection bias theory, as I was saying, is if the person care about who is trading before him and how much they're going to raise his impact cost or other type of cost or like just execution cost,
00:44:48
Speaker
then how much they're willing to pay for that. So essentially, like in my mental model is they care a lot of about what they're ordering in one block, not only just which block they're on.
00:45:00
Speaker
And then for especially for informed trader, because they're especially worried about someone trading in the same direction with relatively of large size happened before them because this is very likely going to generate higher cost for them to put down their positions.
00:45:14
Speaker
But on the other hand, for uninformed traders, Maybe there are less of worry about the within va block positions because there are less worry about people trading before them are highly likely to be the people trading in the same direction of them.
00:45:27
Speaker
Then, Shahal, go into more like EIP 1.59, how that impact research. Pretty much it doesn't. Yes, that's actually my my thinking too. So I think EIP 1.59 basically separates fees into two parts, right? The base and dip.
00:45:42
Speaker
So I think of course the idea behind IP 105 and I is actually going to adjust the base fee according to the congestion blockchain, right? But when you think about um informed traders competing with each other to trade a very, very profitable information.
00:45:57
Speaker
I don't think that will, like the increase in the base fee just to kind of account for the congestion will actually it be enough ah to deter them actually doing a competition, right? Because they're actually trading on a very profitable information. So they were going to actually increase the tip fraction of the fee a lot, right? Trying to actually win over each other to get to the first ah place. For example, we just talk about the importance of putting your attraction at the top of the block.
00:46:23
Speaker
but I think within the block, then the tip really matters. right that but How much you're willing to pay for the tip will actually determine so where your attraction will be located and in the block. So if the competition between phone traders it's very fierce, right? Information is very profitable.
00:46:38
Speaker
I don't think, yeah, people are going to lie, but increasing the base fee based on congestion will actually kind of deter that kind of behavior. To me, like it's just like a reserve price for a transaction in this block, right? You at least have to pay that much.
00:46:52
Speaker
Then, like, sure, this is a threshold if you have not very profitable information, tradition you may not do it. But on the other hand, like typically those people are a bit much higher than not only the base fee, but by base fee plus some other priority fee, the median one of them.
00:47:10
Speaker
Then like for those people, I really like sure, maybe there are some people who are kicked out of the game if the base fee is high enough. But on the other hand, I think the competition really, like as I said before, is really about how much the bid ah above their competitors, how much they're relatively positions in a block.
00:47:30
Speaker
And that's actually way above base fee, right? Right. Okay. but So, okay. But that gets right to the mechanism, right? So, so you are, the the point essentially is like, if you're looking at data before EIP 1559, then in some sense, the fees are The bids are sort of conflating two things, right? They're conflating, I mean, in the data, like just as a matter of, like you know, this is why EIP-1559 was put in place is that before that people had to think simultaneously about these two things.
00:47:55
Speaker
ah They couldn't really detach the sort of priority across blocks from priority within block. And I think what you're saying is that you think the sort of high bids that you were observing even pre-EIP-1559 were driven not so much by people being concerned they're not going to get on the next block.
00:48:11
Speaker
but rather that they're they're concerned about their order within the block. But this, I think, also then gets into the discussion, um which we sort of alluded to, but maybe we could go into a bit more deeply, about sort of, say, private pools versus public pools. So I think you know earlier, Shihao, when you were explaining it, maybe at the start of this conversation, you you sort of made a reference to the ah the public memory pool.
00:48:34
Speaker
um ah but ah So how does how how do private pools sort of change the story here um Yeah, maybe maybe let's let's leave it open at that. um because Because, so for instance, um i you're not, at least for a given builder, um you know, it's not all the same people are are bidding for that block in the sense that a builder is not going to take transactions from a competing searcher, I take it, right? Or something like that.
00:49:04
Speaker
So can I actually chime in here for a second? Because I think before we talk about builders and private pools, and now maybe it's it's useful if we actually explain what we're talking about here what for an audience that is not really into the into the entire infrastructure of how blocks are being built.
00:49:19
Speaker
um Should I take this? Do you want to take it? Anybody wants to take it All right, okay, here here comes Professor Park. um Okay, so just for the but for the benefit of the audience, the way how blocks are built is, you know, there's, so the the old story is you send your orders to a group of miners and then the group of miners somehow run the consensus algorithm. One of them gets selected at random. That person builds the block. That's the end.
00:49:44
Speaker
In practice, that's no longer the case. In particular, this has also to do with proof of stake, generally speaking, because you know the whole idea of, say, in Ethereum in particular, is that anybody really can participate.
00:49:56
Speaker
and It means that for block building, you don't have to be particularly sophisticated. um But on the other hand, of course, there are you know the the ordering of transactions can be beneficial. The selection of transactions that have in some form or other are profitable, for instance, because they pay higher fees is not entirely trivial.
00:50:13
Speaker
and So since then, we see sort of a separation transactions of duties here if you want. and And so what Fahab was referring to was searchers, there's also builders, then there's relays, and then there's the validators. So let's take this as simple as possible.
00:50:29
Speaker
So on the one hand are the validators. The validators are unsophisticated, but they have a block available that they have to fill. okay And on the other hand, there are people who know how to fill a block profitably, but they don't always have the slot available. So, well, what do markets do? They bring these two sides together.
00:50:45
Speaker
So there are searchers. Searchers pick bundles of profitable transactions and then give them to builders. Builders then put these for profitable transactions together and they can also do search services and also put their own bundles together and build a blog.
00:51:00
Speaker
And then they essentially sell this block or rather they um you know they bid for this block so um to be included by a validator. and So they basically what they do is they pay the validator a fee for that block to be included.
00:51:14
Speaker
and so And that involves in particular, because some transactions are profitable, that involves that they share some of the profits that they create from the building of the block. And there's a relay in between so that, you know, the validator can't steal the ideas of the builder, essentially. But that's really just a neutrality part of this.
00:51:30
Speaker
So anyway, so sorry, this is a very long explanation just to for the audience to understand. There's this back office operation, which actually is its own little market where where people, where lots of money flows around.
00:51:42
Speaker
And this protocol leads to what's referred to something as MEV boost. So this is payments that are being made from the builder community to the validators. There are also two phases for that and Shihao will tell basically how we actually have a robustness check for that in our paper. There's the first phase before Andrea mentioned like you have PBS, it's actually there is a Syntrust platform called Flashbolts. That's a platform where miners can have a basically say, oh, it's a platform, it's a two-side market ah where miners can join for people who supply the blog space. And people can also join others as a demander. So they can send transactions to that platform.
00:52:21
Speaker
And then eventually ah there are some privacy guarantee that like they hope they don't leak that information to other players. And if they bid high enough, and eventually the order can be ah potentially prioritized, or um there are other factors on that. There's something called bundle, which I don't want to get into, but there are actually two phases of the um what we call private pool. Before, the private pool is more centralized, it's like, maybe like mostly flashable. Now, like the private pool is kind of more decentralized, where like different builders you have you can privately connect with them and send their transactions.
00:53:01
Speaker
In our paper, we actually only look at the first stage, where the private pool is still very centralized, it's still not PBS, and then Shihau will explain like actually like what do we you find, how private pools affect our results.
00:53:15
Speaker
Probably for that, I just want to maybe have a mini sort of history lesson on like the evolution of the private pools. I think even before, I think Andreas did a fantastic job is planning the current PBS sort of proposal builder separation mechanism.
00:53:30
Speaker
Even but before the Ethereum blockchain moved from proof of work to proof of stake, where we have PBS coming, and even before that, we have already private pools.
00:53:40
Speaker
So basically, if I get like the timeline right, I believe flash bots were first introduced to maybe back in February 2021, correct me if I'm wrong with the timeline.
00:53:51
Speaker
so Essentially, as I said, so before, Ethereum was like a public P2P network, right? If you're a node on the P2P network, you have your own mempool. So if you listen to the network messages, essentially you get everybody's transactions, right?
00:54:06
Speaker
And those transactions will be pending your mempool. So basically, everybody gets to see everybody. But then some traders think, okay, don't want my orders to be seen by others, right? If there's a way for me to reach the miners, who is the one who has the final say over which tractions to include, that would be better, right? So I don't need to go through the P2P network, propagate my tractions over the whole network.
00:54:27
Speaker
So I think Flashbots introduced the service back then to connect traders directly to the miners, right? So the so-called private pool. So basically, if you want to execute your order, you don't need to go to the P2P network again. You just need to, can just, you can just send your order directly to the miners.
00:54:46
Speaker
So During our sample period, so basically our sample started, I think, from November 2020 to August 2021. So in the second part of our sample, we do have the flash bots.
00:55:00
Speaker
So basically the rise of a private posts. So during the second part of the sample, informed traders can theoretically actually go to flashbots, right, to send their transactions directly to the miners.
00:55:13
Speaker
So in the paper, we sort of actually use the second half as a business check. So our question was, if informed traders are given the chance of using private pools, right, sending their transactions directly to miners, will they still actually send any sort of this kind of informed orders through the public pool or the P2P network?
00:55:33
Speaker
So what find is actually, even if that the public pool existed, so they can use that. A significant part of an informed trader still actually choose to trade the public mempool. That was actually quite a surprising a result.
00:55:50
Speaker
And I think... Wait, wait, wait, wait. That shouldn't be surprising because you have paper on that. And don't you? Right. I was about to. QA, Ritzer, yeah.
00:56:02
Speaker
you want you yeah You have a paper. So so I'm not sure if you're on the paper too. I forget that. But I know Rizzo and Agustino, you have a paper on that, right? So that, you know, because you have you you want to get your, you really want your ah your trade to be done.
00:56:16
Speaker
If you send it to a private pool, you face possible delays, right? And so therefore, you may actually put these trades still into the public pool. I'm so sorry to interrupt, but I just have to, you know, plug your own work there for moment.
00:56:28
Speaker
Yeah, I'm gonna say that even if we go through the private pool, that doesn't mean there's no competition, right? because I as informed trader can connect to a miner, you as a competitor or competitive informed trader can also connect to the miner. So essentially you're still competing with each other.
00:56:43
Speaker
just instead of, I can view your bids in the public event pool, now I cannot even view your bids, right? We're actually engaging in kind of um a private auctions. And I think Reza has a paper and there, I think it shows some statistics that even if traders actually should go through the private pool,
00:57:02
Speaker
they still actually bid a very, very high fee to the miners, right? I think, I forget the number, I think you showed like maybe 80, 90% of the profits actually are bidded away, right? You have to actually still... But that's more public information. So I want to separate it. That was more front-running and other things, like it's a public waterfall comes in, people front-run it, and eventually most of the...
00:57:23
Speaker
like the fees are going to the infrastructure, which is the miners back then. But I think it's kind of separate from our study of informed trading. So like I think the main takeaway here is really like, oh, private pool doesn't affect the fact, at least the phase one of private pool flashbought platform doesn't affect the fact that there is a lot of information in both markets and trade-in exchange coming from the high fee flow in public pools.
00:57:49
Speaker
So I think that's ah one sentence takeaway. And then this can change. like but before This can change a lot. Because after PBS, as I said, the private are fragmented. Also, like to me, Andrew, I don't know if you agree, because the analogy is tricky. Because I think now the builders like brokers, but then they also have their own orders they can insert, and they also like have their own priority rules.
00:58:13
Speaker
So eventually, like I think they have a lot of bargain powers, and they also do lot of things themselves. So I think that things can change a lot how informed trader behaves. um Yeah. but So I think, Rija, what you said there is really important, right? Like, well, one, you were distinguishing public information that you were discussing in another paper to to private information. but But so what I'm really trying to understand deeply is like the nature of the private information that we're describing, because I think you've already said that um it's not necessarily the type of thing like, oh, one player has...
00:58:44
Speaker
all of it and nobody has anything associated with it because in that case you know they they wouldn't need to bid so aggressively anyway um uh and you also at some point in the conversation were sort of alluding to information leakage right um in that like in the public mempool if i send in a really high bid maybe i'm tipping my cap that i you know maybe i'm i'm revealing information um and so Could you say a little bit about, again, when we're talking about the private information, so let's set aside again, checks, attacks, ARBs, how much attacks, et cetera, et cetera. the type So let's focus on what you are focusing on on this particular paper.
00:59:16
Speaker
um How, you know, ah can you give a sense of like how many players have highly correlated private information here, right? Like, yeah, it's not public. It's not everybody knows it.
00:59:27
Speaker
But are we talking about... a few players who have highly correlated information, a few players who have weakly correlated information, are we talking about many players who have highly correlated? you know Because for instance, like when you switch from private to public,
00:59:41
Speaker
If, for instance, there was, you know, very few players um who have any kind of relevant information and they're maybe they're not that correlated, then like information leakage seems like it's a much bigger factor.
00:59:54
Speaker
And so if I go in a private pool, well, the other players don't, like there's no mempool there, right? So it's not like just because, you know, you and me are both using this private pool that we see that that each other's transactions. We don't.
01:00:06
Speaker
um And so that dimension starts to become a much bigger factor, but then it all goes back to like, what is the information structure when we're talking about these so called private information transactions?
01:00:19
Speaker
um How correlated is the information for people who have this private information and how you know how um how many of them are there? yeah I will first say once and then Shihao answer like from the... So we actually looked very closely at how informed traders bid against each other. There is something interesting will share with you, like their strategy.
01:00:39
Speaker
But on the other hand, I think so we only know what we can observe. um So I actually don't have a good answer of the whole ecosystem, how correlated people's signals are.
01:00:50
Speaker
And because you only see like, for example, Shihau tell you, we only see people are bidding against each other. He can tell you how many people are bidding together to get their order executed. But there are just a big chunk of flows like nobody are bidding so maybe like majority people are their signal not that correlated so they don't have to bid high fees so the true answer is we actually don't know because there is potentially a selection bias if I give you some number but Shehal can tell you for the people who are actually bidding bidding high fees how many people are competing and what their strategy
01:01:22
Speaker
Right. So I think the last or and also the last interesting result of our papers, we also look very closely at how informed traders bid. Because I'm not sure if you remember the very famous paper at the very beginning of the whole literature, there's a Flashpoint 2.0 by authors.
01:01:40
Speaker
I still remember there was a very nice graph where they showed there's a bidding war, right? So two sort of addresses that gradually increased the bids the way until the end of the block. so So that's also something we want to actually look at in our data set, right? Do observe the same sort of bidding behavior or bidding wars between the informed traders?
01:01:58
Speaker
So what surprised us is actually, even for those trades that traders bid eventually very high fees, we don't observe this kind of gradual bidding war. Instead, we found what they do is so-called jump bidding.
01:02:11
Speaker
So jump bidding just means that you do not actually start with a very low bid and then observe what your opponents is actually bidding and then counter bid and gradually bid up, right? Instead, we find out that informed trainers, they actually place a very high initial bid, right? So they actually bid a very high gas fee already for their first order.
01:02:30
Speaker
And there's a literature in comics essentially talk about jump bidding. The reason for that is actually they use it as a tactic. They strategically beat a very high field at beginning to kind of signal to the other competitors that I value the traction very much, right?
01:02:48
Speaker
so event and So it's a basic say, okay, don't compete with me. There's no way that you can win. So by using this kind of strategy, Even if, right, if you think about public manpool, it's like a double-edged sword, right? In the one hand, if it would be the high fee, you're leaking information potentially.
01:03:07
Speaker
in the other way, is also something you can use to signal your trading intention, right? Your willingness to pay. So basically, if you can use the public manpool well, right, by using this kind of jumping behavior, can you can even ward off some competition.
01:03:20
Speaker
So I think there are like two different balancing forces. of for using the public mempool. So our people find that even the private pool actually existed, traders still actually use the public mempool.
01:03:30
Speaker
And one potential of reason might be we think that they want actually use that to use some more sophisticated bidding strategy for that jump bidding to help them to trade. And also, Baha, just add one more thing. like We see the bidding is very different from like Phil's paper, because probably because sandwich is very public information.
01:03:48
Speaker
But then for the private one, we look at First, like they don't bid that often and normally just a few rounds and a few players and the interval is much bigger.
01:04:01
Speaker
um What is the second part I want to mention? so I forget it. So, sorry. I think really like that you can see, um but then eventually because all those forces, we actually don't really know if people are scaring people off each other, if we don't observe people who are informed but don't have a competitor. So I really don't can't give you answer on what is the relatively correlation between people and what they behave. We can only tell you, hey, there are people be competing, that's what they do, and then i think that's it.
01:04:36
Speaker
Well, that's still very informative, right? I mean, I think what what we see here with your paper we get a sort of a glimpse under the hood of what probably also occurs on many other markets. ah Because we do, I would imagine we see aggressive bidding and aggressive behavior, um trying to take advantage of information asymmetries or of short-term information in particular in every single market. But it's just very hard to see, right? Whereas in your particular case, we get to see a lot about how people react to to news and so on.
01:05:06
Speaker
And that's kind of insightful. um I mean, you know, look, I've i've actually dealt with ah proprietary data, right, where sometimes you can see similar phenomena where under certain circumstances, ah certain subgroup of papers traders really aggressively try to bid.
01:05:21
Speaker
um Anyway, so and and that's actually quite telling, right? You can actually learn something about the individuals there, too, which is great. um So, um you know, I think we were pretty much covered everything that you have in your paper. So I want to thank the two of you for for joining us in the podcast. And i hope that the audience enjoyed it too.
01:05:40
Speaker
Thanks guys. Thanks so much for your time here.
01:05:44
Speaker
We hope you enjoyed this podcast. Thank you for listening. As a reminder, you can find additional materials on owlexplains.com and can stay updated by following us on social media. That's all for today.