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 9: Mitigation Methods for MEV and LVR
41 Plays1 month ago
Professor Ciamac Moallemi (Columbia University) dissects Maximal Extractable Value (MEV), Loss-Versus-Rebalancing (LVR) and associated mitigation methods. To provide more background, MEV is the economic value that can be extracted from blockchain users through arbitrage activity. Of particular note, arbitrageurs can extract value from liquidity providers at Decentralized Exchanges (DEXs) where these losses are known as Loss-Versus-Rebalancing (LVR). After explaining both MEV and LVR, this podcast focuses on mitigation methods including expedited block times and auction mechanisms intended to extract MEV.
Paper: Automated Market Making and Arbitrage Profits in the Presence of Fees
Recommended
Transcript
Introduction to OWL Explains Podcast Series
00:00:06
Speaker
Hello and welcome to Crafting the Crypto Economy. I am Silvia Sanchez, Project Manager at OWL Explains by Avallabs, and today we bring you a transformative podcast series in partnership with the Crypto and Blockchain Economic Research Forum. This series features leading faculty from renowned global universities exploring various elements in the blockchain ecosystem. These episodes are a bit longer than our usual hootenannies, since we will be getting very deep.
00:00:33
Speaker
And also, each episode will have its accompanying paper posted on our website for further reading.
Guest Introduction and Discussion on Automated Market Makers
00:00:38
Speaker
And with that, I will hand it over to our moderators Fahad Saleh and Andreas Park. Hello, everybody, and welcome to another edition of the SEBA and our Explains podcast series on crafting the crypto economy. My name is Andreas Park, and I'm your co-host together with Fahad Saleh from the University of Florida.
00:00:59
Speaker
um We're very happy today to have Siamak Maleni here to talk to us about his latest work on automated market makers, in particular on the avoidance of MEV. With that, um we have ah had Siamak before here when he talked to us about a concept called lever or loss versus rebalancing. And yeah, so we're very happy to have you back.
00:01:26
Speaker
and I'm very much looking forward to this episode where we learn a lot about how ah extractable value can possibly be avoided. now For the audience, I hope the audience is not already put off by by all the jargon that we've used here. What we're trying to do is we're trying to to ease everybody into the topic as slowly as possible because it is a constant, it's ah it's a bit different a complex topic, ah but that doesn't make it unimportant. It's actually probably one of the most critically important
00:01:56
Speaker
ah subjects and topics that we have in the DeFi space.
Deep Dive into Automated Market Makers and Arbitrage
00:02:00
Speaker
um So, maybe we start very, very simply. Sir, Mark, can you outline to us, A, just briefly, what an automated market maker is and how ah extractable value arises in the process of operating an automated market maker?
00:02:19
Speaker
Sure. um and and And thank you for having me, Andreas and Fahad. um so yeah So an automated market maker is an automated protocol that's typically implemented in a smart contract on a blockchain that um kind of programmatically makes markets.
00:02:38
Speaker
So um you have these agents called liquidity providers. They provide assets to an automated market maker. So maybe as a running example, we can imagine trading ETH, the ah Ethereum cryptocurrency versus, let's say, some tokenized form of US dollars. So you would have these liquidity providers. They would provide pools of ETH and US s dollars from which end users could buy or sell one or um the other of of those two assets.
00:03:05
Speaker
And what an automated market maker does is it provides an automated system where based on the quantity of these assets that are available at any given time, it quotes a price or an arbitrary trade of of one of these assets. And um that um ends up intimately relating to this concept of MEV or um or or maximum extractable value. um I'm not going to sort of get into sort of formal definitions. But in my mind, NEV is the value that arises from being able to order transactions in a block on a blockchain. And in this context, what's especially important is there's special value um when trading against an automated market maker to be the first to trade in a particular block.
00:03:48
Speaker
And um the reason for that is arbitrage. And um the idea there is that you know um this automated market maker is providing quotes um that is prices for trading these assets. And but but those prices are static because again as I mentioned they only depend on the quantity of assets that are available in the pool.
00:04:06
Speaker
Whereas in reality, in the real world, information is constantly arriving and um the the the real value of these assets is is is sort of dynamically changing. So when you go from one block to the next, if there's a price movement in sort of, let's say, the real world,
00:04:22
Speaker
um on you know some of these these assets, the price in the the AMM will be stale, and there's an opportunity to make money by exploiting that stale price. For example, if it's too low, maybe you buy on the AMM and you sell um us somewhere else.
00:04:38
Speaker
As you do that, um as an agent does that, let's say you know someone called an arbitrageur or a searcher, that opportunity will disappear because the price will be corrected um ah through that trade. So there's money to be made exploiting these um stale prices, but really only if you're the first one.
Blockchain vs Traditional Market Arbitrage
00:04:55
Speaker
And so that create is you know creates this this maximum extractable value. I think you know there's other types of MEV, but um you know reality is that you know DeFi trading indexes are perhaps some of the most successful um you know applications on on on modern blockchains. So in in my mind, this is really sort of the dominant or prototypical form of MEV on modern blockchains. So let let me let me ah let me interrupt you here and maybe we can before we go um into the the details specifically when it comp pertains to blockchain. um So first of all all I think it's useful probably for the audience to understand um you know how this relates maybe to traditional markets right because we have arbitrage we have traditional markets say an equity market
00:05:38
Speaker
NASA trades usually also multiple different platforms, say for instance, can trade on the New York Stock Exchange, on IEX and also on the NASDAQ. Anytime a price moves in one market, there's an arbitrage opportunity relative to the other market. so In that sense, it's kind of the same as what we have in traditional markets.
00:06:00
Speaker
but there is something special about the way how a blockchain operates right so that that could rise to a particular type of arbitrage opportunity and maybe in this context if If I can press you, can you, because I think the the settlement process of how I see transactions are being processed in the blockchain matters a fair bit here. Can can you maybe ah explain that to the audience? Yeah, so so you're absolutely right. The ah the um same phenomena occurs in and in traditional markets. um It's often called sniping or ah like high frequency traders or you know sort of snipers or so on where prices you know price move on one market and the traders will race to exploit it on the ah the other market.
00:06:41
Speaker
Now, the the the key difference which you're alluding to is um ah pretty much every TradFi market that I'm aware of is first come, first serve, right? So when there's a race to exploit a stale price, the mechanism to compete in that race is to be faster than the other guy. So you see lots of investment in things like you know very fast computers, maybe specialized computers like FPGAs, maybe um specialized communication links like microwaves or lasers or or sort of whatever.
00:07:09
Speaker
On the other hand, the the the process in a a blockchain, you know, the blockchains are are not first come, first serve. They, you know, trades occur in in when blocks are generated. And so, you know, there's various kinds of of rules, but, you know, I think it's just sort of a very high level. um That order in the block is, ah is is you know, typically um determined by a builder or a proposer.
00:07:32
Speaker
So um where you're going to compete in a a blockchain is not necessarily being um ah faster than the other guy, although that might be one element, but you're going to compete in in in that block building um process. So in some sense, if I want to put it really make it really a really gross simplification, it's almost as if The DTTC, which is the Clearing Settlement Corporation in the US, would settle trades in real time. And if they would be actually taking advantage of any arbitrage opportunity, is that maybe a way, an analogy that would be fitting here?
00:08:05
Speaker
um If you want to think about the ah the builders as the ones who are doing the arbitrage, which you know I think is is largely um the case today and in and the larger chains, that's
Impact of Arbitrage on Liquidity and Market Dynamics
00:08:15
Speaker
correct. like it's It's almost like the you have vertical integration so that the um basically it's it's so valuable that the economic incentives have created this kind of vertical integration, which in some sense the infrastructure itself does the arbitrage. I think that's a ah reasonable way to look at it.
00:08:30
Speaker
Yeah, because I think in you know and in traditional markets, the speed, as you say, right is is really derived from the fact that people make investments of linking markets, but there's no such thing because the markets in in the blockchain are not really physically anywhere. It's just basically just the contracts on the chain. It's really a question of the the processing of the of the of the inie infrastructure level that gives rise to the exploitation of some arbitrage opportunity. I see.
00:08:56
Speaker
so now In traditional markets, this is a very well-known paper by Eric Budish and his co-authors on the effect of sniping, where they basically say that sniping itself creates costs. right because you know i mean the The story goes as follows. is there is you know You have a limit auto setting on the market.
00:09:16
Speaker
And you know if you want to if you are the one who has it and there's new information that arrives, you want to cancel this order. But because there are snipers that are trying to you know be faster than you, there's a probabilistic chance and a very large one that you cannot cancel the order and change it, so it becomes stale and therefore get exploited. And that alone creates creating costs.
00:09:37
Speaker
right so i kind of kind of That's the argument, so you have to require the bid-ask spread in order to... The bid-ask spread could otherwise be lower if it wasn't for the snipers. The interesting feature here is that in an AMM, there's no spreads, for instance, that can have it. um But there is fees that people pay. right so if Every time you want to use an AMM, you have to pay a fee.
00:09:59
Speaker
Now, when you think about when i the way I think about it is, okay, so there is a possibility that because there is a trade that happens, let's say, on Uniswap, there could be an arbitrage opportunity relative to SushiSwap. And you would trade against both of these markets. So, in what sense is there can we say, is there a sense in which there is an excess price that liquidity providers therefore need to account for, that that that they have to charge for them not to get taken advantage of or um I think so. I think, um you know, the idea is pretty similar um when, you know, I think these kind of very short term transactions are zero sum between the the two sides, between the arbitrage and the liquidity provider, be it in a limit order book or an AMM. So I think in the same way, the liquidity providers are losing money.
00:10:46
Speaker
Now, in in sort of current AMMs, there's not a mechanism where there's you know um the the fee is ah um you know the sort of the device that could equilibrate. You could imagine that could be the case. You could imagine that you know as you set the fee to be larger, um then that would sort of you know ah like reduce some of this ah um type of a sniping activity. However, the um the equilibration mechanism is essentially how much liquidity is there.
00:11:12
Speaker
because um ah you can sort of view, like again, these these type of arbitrage trades are very profitable and very low risk. So at some level, there's infinite capital available for arbitrage. However much capital is in the pool, like enough capital will appear to to arbitrage ah against that. On the other hand, you know um the other type of trading in the pool is is also important, and that's the noise traders, right? If we only had arbitragers, liquidity providers would always be losing money and there would be no liquidity provision.
00:11:40
Speaker
However, we have these these sort of noise traders, and as you increase the um liquidity in the pool, the amount of noise trading doesn't necessarily linearly increase. It's it's going to be sort of sublinear. So I think the equilibration mechanism is through the, um between um ah you know sort of the amount of arbitrage activity and the state of the pool. is really through the level of of liquidity.
Debate on Market Welfare and Maximum Extractable Value (MEV)
00:12:00
Speaker
And the more arbitrage activity you you have, the more money the LPs will lose, and they will withdraw liquidity until, again, that sort of um the losses to arbitrageurs balance to the amount of money that they're making from non-information-based traders, the so-called noise traders. So I'm i'm going to reformulate maybe slightly what you say, and because you know in some sense, it sounds like, well, you know some party one party wins, the other party loses. Who cares? But I think in some sense, we should care in the following senses, that if it is becomes more expensive for liquidity providers to provide liquidity, that raises the cost. And as it raises the cost, it makes it more expensive for
00:12:41
Speaker
the average joe the noise trader to trade i think that's probably the way how we should think yeah i didn't mean to suggest that nobody cared like zero sun doesn't mean nobody cares um uh absolutely i think the the lps care they will provide less liquidity and what that means is when noise traders come to trade there's more price they experience more tri price impact and higher trading costs. And maybe that higher trading costs would ultimately dissuade them from trading in some cases where they may trade otherwise. And so then you don't get sort of the welfare you know kind of gains that from trade that that you would get if the arbitrageurs weren't there and if there was more liquidity. So and like my view is that um ultimately this is ah um ah you know um a negative. This MEV is negative. Now um I should say that this is not a a universal view. Some people um ah you know like to argue that it's positive in the sense that these arbitrageurs are providing a service that they're correcting prices so that the ah the the that the prices on the AMM are accurate. um I think there's some truth to that. But I also think that this service should be very cheap.
00:13:45
Speaker
Right. um Because you know they're doing relatively obvious things. They're looking at the price on Binance and comparing it to the price on the AMM and you know trading if if if those two are different. And if we had a competitive market um you know sort of people would sort of compete away that profit and be willing to do it for very cheap.
00:14:01
Speaker
So I think, um you know, there is a ah service being provided there. But um the the question is, you know, you should be really paying ah if if it's something that's very obvious and that anyone could do, you should be paying something very close to zero for that service of correcting prices. yeah So here's something which I think is actually maybe it's a useful tidbit for the audience to know.
00:14:20
Speaker
um if you If you think of a traditional market which has a limit on a book like the Nasdaq, the New York Suctions, but also Binance, of course, um it is possible that prices can move in these markets without trades happening. So that essentially nobody actually loses from, ah say, a news event of some kind. I mean, you know and in crypto, there's always a little bit more difficult, but imagine you had a stock, there's all of a sudden there's an announcement of, say, some earnings. That means that the price has risen. um In, ah you know, in traditional markets, that would mean that essentially people who have posted limit orders can cancel these orders and repost them at a higher price. So nobody loses, right? So no money changes hands, no shares trade and change hands, but the price change all the same. In an AMM, at least in the ones that that we normally think of, like Uniswap and the like,
00:15:09
Speaker
Prices really only change when ah when there is a change in when there's a trade. and i mean you You could internally change this. And I think this is something that we have to talk about, because I think this is part of your proposal, maybe. um but But fundamentally, the operation of changing liquidity, which is what happens in the limit order when people repost, doesn't do anything to prices.
00:15:33
Speaker
right so So in that sense, so but here's where coming back to your point that you made just a second ago, um the service that arbitrageurs therefore provide in making the price adjustment is precisely maybe, is that something that is maybe just needed in ah in ah in an AMM? So actually that this can work out.
00:15:52
Speaker
um So you you do need it. And i think I think that's part of the appeal of the AMM, actually, is you can be a passive market maker. You can make um markets and set prices um without having sort of without having a model or views about the future or so on. And I think arbitrageurs make that possible. But um I think the the the point I'm trying to make is in a standard setting, like in you know the AMMs we have now, like let's say Uniswap or or or um ah whatever, um the arbitrageurs get paid too much for this.
00:16:22
Speaker
right they They exploit the ah the price difference and they make that full amount, right? Whereas if if it's something that like really anybody could do, there should we should sort of try to think about mechanisms by which um ah you know either either we move prices without trade, as you suggest,
00:16:37
Speaker
or other types of mechanisms where maybe there is you know some kind of competition or some kind of auction. So multiple people who can be arbitrageurs compete with each other, and they compete with with each other in a way where they provide that same service, but without taking all of the profit from the LPs.
00:16:54
Speaker
ah Wait, on on this point, though, so I'm not totally or i'm not i'm not that convinced by the idea that so the arbitrages are providing as much of a service as some people describe it as. And and the reason for that particularly is, so if you imagine, as you were saying, let's say they're liquidity traders, right? um Now, normally, we think of liquidity traders where there's like one venue, and so they have to trade, and so they will trade there.
00:17:18
Speaker
But if they're liquidity traders and there are multiple venues, like even but even somebody who's uninformed might think, OK, well, all things equal, I should get the best price that I can get. And so if you imagine there are no arbitrageurs, think of a mental model where there are no arbitrageurs, but just liquidity traders who come in and just choose the best venue to trade at. Wouldn't that cause reversion in prices, at least to some extent? So like if you have just natural liquidity in trading,
00:17:45
Speaker
Do you really need arbitrageurs to come and show up and trade it may happen slower but if you have a lot of equity in the market i don't see what real what the real services are providing i think you have a point um with respect to for example to like let's say.
00:18:01
Speaker
um on-chain markets right like if you have like let's say five different AMMs all trading on Ethereum mainnet maybe different fees maybe with different um liquidity curves you know whatever um I think these days most of ah like um ah you know retail traders most noise traders they do come into the market through routers and these routers will pick the combinations to of exchanges to trade on which um have which offer the most beneficial price to to that end user. And by virtue of that they will end up ah correcting the prices of on these exchanges to match each other over time. So I think you do have a point there. But I think the um the the place where that that we we don't have that arbitrage linkage is if you want to try to compare to other markets.
00:18:45
Speaker
right if you want to try to compare to like let's say what is the price on Binance the routers typically aren't aren't looking at the price on Binance because ah you know if if you're if you're trying to swap on Ethereum mainnet you're probably not set up to trade on Binance you need the KYC and you need the the credit relationships and so on right so they're not arbitraging against those um um kinds of other markets. And I think also there's maybe things which are like a little bit less of like kind of a pure arbitrage and more of a statistical arbitrage. So for example you um let's say you're trading a coin which is um a not kind of a super liquid right. And um it's trading in ah and an AMM. The price is static. um But let's say the overall crypto market is going up.
00:19:27
Speaker
Right. So there's not an obvious way for i'm an end user who wants to come in and um you know sit like you know know that that this thing is mispriced. Whereas if you have arbitrageurs they can they can build statistical models they can build things like factor models to say hey that crypto prices are going up.
00:19:43
Speaker
I should buy on this AMM and push the price up um ah even though there's not like kind of a pure pure kind of arbitrage there. So I think and in some sense you're you're correct with like a limited set of markets and so on especially things like routers will correct prices across exchanges but if you want to pull in information um um from other sources information from off-chain and so on then I think um ah you know the arbitragers are providing a service. Sure okay but so i actually there' there's a point that We talk about checkstacks arbitrage, meaning centralized exchange, decentralized exchange ah arbitrage, um i which we were alluding to just now with the idea of, well, bananas is a centralized exchange. Say Uniswap is a decentralized exchange. Routers aren't going to correct for dislocations between bananas and Uniswap, for example. But it it seems to me that a lot of what we call checkstacks ARB is actually value ARB, which maybe it was in the category of what you're saying is not in the same sense a proper ARB. But
00:20:40
Speaker
on it's not clear to me that that by itself is not sort of a significant factor. So like right now we have deeper markets in centralized exchanges. um But there are, for example, assets that don't trade at centralized exchanges, but they do move. you know and And so, for instance, like to your point, like the crypto market is going up. Is this asset that doesn't trade at a centralized exchange you're going to move up? right So it it seems to me like The long run vision, in some sense, I would hope for crypto, right is that decentralized exchanges, for example, don't have the sorts of let's say overhead costs that a centralized exchange might have. right So right now, we live in a world where the centralized exchanges have more market depth.
00:21:20
Speaker
but um If we can solve some of the problems at decentralized exchanges, right like where we're going here with with sort of the liquidity providers taking these losses through things like lever, if we can solve these sorts of problems, is there a reason why the primary price discovery couldn't be happening at the decentralized exchange and then its it's value arbitrage that you you would expect? like it's It's actually the flow is going like, oh, this big event happened. I'm going to trade it at the primary venue, which is a decentralized exchange.
Challenges for Liquidity Providers and Adverse Selection Risks
00:21:49
Speaker
and and And so it's not it's not that arbitrage in the sense that you mean the two leg arbitrage. It's more there's flow coming in like, come on, like that we had really good news just now. The price should go up in the same way that if I look at IBM stock, like why is the price going up? It's not like there's an ARP. It's oh, we had good news or whatever. Right.
00:22:07
Speaker
I agree with that. I mean, i I think people in general use the language of sextex are because I think it sort of makes it concrete. But I think the right way to sort of um ah think about that is is um not that you're actually going to do the um the hedging trade, right? Like if if if the price on Binance is higher than the price on an AMM, likely you will, um an arbitrage will buy on Binance because that's the mispriced part.
00:22:33
Speaker
but they're not necessarily immediately selling on the on on on centralized exchange because there's also transaction costs associated with that, right? So I think the the right way to sort of think about it is the centralized exchange provides a really clear, credible um signal as to the value, right? Because as it happens, that's where most trading happens right now, and that's where um a price discovery is is is occurring. If we switch to, you know, kind of a world um where, um you know, the the volume picks up on the ah on on the decentralized exchanges, I think um people will also construct these sort of ah value signals. And, you know, I mean, this is this is actually what market makers on on centralized exchanges do, right? They construct some, you know, various kinds of signals as to what is the value of this asset now, what's its price going to be five minutes from now, an hour from now, and so on. And they leverage that to sort of um ah build their quotes in terms of where they're willing to buy and sell.
00:23:29
Speaker
That's in the future what what arbitrageurs will do. um they They will not necessarily always just do arbitrage trades. They'll build fair value models based on the price on Binance, based on you know information, based on other information, based on maybe news articles. And based on that, they will um build you know fair value models of where they think the price will be, and they will trade against the ah the AMMs you know um ah you know if if they're inconsistent with ah with with with those those ah those prices.
00:23:57
Speaker
Yeah, I mean, guess I guess I was just thinking about, um to some extent, the liquidity providers are still facing losses, even if it is a value arbitrage, right? um So I guess even if you had even if you had trading only happening at the decentralized exchange.
00:24:13
Speaker
It's still a problem for liquidity providers that the price is moving around because they're still going to face these losses. The fundamental problem is is you know what we discussed earlier, which is that you can't adjust prices without trades. right and so that that creates and And we know prices need to be adjusted constantly in extremely volatile assets like cryptocurrencies. And so that is going to create some kind of intrinsic loss.
00:24:36
Speaker
Yeah, so if if I could chi chime in here as the economist, I think what you're really, what you're discussing is really adverse selection in the sense of that there is better information to be had somewhere and somebody benefits of it. So this is like something which is like of a permanent nature and that's that there's a permanent shift in the underlying and therefore somebody has that information before and can can therefore arbitrage against that.
00:25:00
Speaker
And so the I think one of the issues, I mean, this is this is my own pet little interest, is the fact that the price that you sell something at using an AMM is always lot ah below the marginal price that pertains after the trade. So that's kind of the the correct price or the arbitrage price. Except if you have, I mean, there is a way to do this. I think Cowswap actually has a price where that is not the case, but there's some other issues that arise with that.
00:25:26
Speaker
um is Is that probably the best way to to describe this year? At least for those of you, if anybody here has an economics training, you'll probably know what I'm talking about. um Maybe in this context, can I can i ask you, Seema, can you maybe just define one more time for for our audience also what the concept therefore of loss versus rebalancing is, or lever? Right. So the concept of loss versus rebalancing is to ah look at um ah basically the trades that were done on the AMM, and um instead of doing them at the prices they were done at the AMM, to do them at fair prices at that time.
00:26:04
Speaker
So the idea is the quote unquote it's loss versus rebalancing. What rebalancing means is if you had a um a trading strategy which you know maintained the same position as the AMM but simply did it at um you know fair fair market prices. You know um ah for example let's say at at at the sort of finance prices.
00:26:23
Speaker
And um ah what we see is that um you know because of the staleness of AMM prices, um the AMM always gets worse prices than had it done those same trades on Binance. And again, that's you know um ah you know ah worse prices, I should say, when when they're trading against arbitrageurs. And that's because the arbitrageurs are trading because the um the the prices are better for them, and in in that case, worse for the ah the the LPs. And so, hence, you always have a loss.
00:26:50
Speaker
Right? And so therefore it's a loss versus rebalancing. You're worse off in the AMM than if you did those same um ah trades at, like let's say, fair or reference prices. Yeah, I think that that's probably, and that that kind of captures exactly the issue of theres there's a shift in the fundamental, and that's really the price that that is the price at the fair value price that you have after, right?
00:27:10
Speaker
Now, and so and the tradeit that you and so therefore, the way you describe this is, fundamentally, this allows this difference in prices that an AMM has relative to what the fair price is that that gives some rise to some m MEV extraction. right um Or it is one form of MEV as you describe it. defense Essentially, I think it's probably one of the best ways to describe it is the arbitrage between a fair price, which pertains elsewhere, and what you can get in an AMM.
00:27:38
Speaker
and now And so the reason why we like to mitigate that amount right as ah is as for a ah for liquidity provider perspective is if they could be mitigated, they would be willing to offer better prices. And then you know the little people like us that want to trade like $100 worth of some cryptocurrency times would actually get better
Strategies to Reduce Maximum Extractable Value (MEV)
00:27:59
Speaker
prices. right Now, how would we go about what what could we do to reduce this? So this is a very active area of ah um ah research. It's something that I've been thinking about, something that others have been thinking about, um ah both in you know academia and in in and practice.
00:28:18
Speaker
um I think there's maybe um four main classes of of of of methods to sort of mitigate this MEV and i' I'll list them and then you know maybe we can um ah do sort of a deep dive um ah you know and into some of them. um The first is um of you know simply i'm running faster blockchains.
00:28:37
Speaker
um So um ah what it turns out, and you know I can separately get into this if you like, is um the the more frequent um the blocks are, the more like a quote-unquote continuous market. um ah You have continuous in time rather than being discrete where you only trade when when when blocks are generated. The less um ah this um um ah um form of MEV is, the the less loss versus rebalancing there is. And that's simply because um having faster blocks in some sense creates more competition between arbitrageurs and they compete away um ah some of these profits that they would would otherwise make. so So that's one kind of strategy, having faster blocks. so I don't understand quite the mechanism. So if the if the price of ETH or the the fair value of ETH goes from
00:29:22
Speaker
Let's say $3,000 to $4,000. So that's that's a fundamental shift. right um And there is a particular you know amount of liquidity in an AMM. Say I can trade a million dollars. It takes me a million dollars to move the price from $3,000 to $4,000.
00:29:40
Speaker
How would a faster blockchain change that? So what a faster blockchain does is um because prices have less time to move, the more frequent blocks are, the less prices move between blocks.
00:29:54
Speaker
right And that has to do with the underlying assumption of the of the stochastic process of the price. Right, that's right. So over smaller periods of time, and you know there's different ways to model prices. They could be continuous, they could jump, or whatever. But in every model of prices I've seen, in some probabilistic way, the smaller the interval of time,
00:30:14
Speaker
the less chance of a large price movement, right? And if you think about what is the profit of an arbitrageur in sort of a very kind of approximate sense, it's roughly quadratic with the with the price movement, right? um Why is it quadratic? um Well, the first is the bigger the price movement per unit that you trade, you're going to make more money.
00:30:34
Speaker
right um like and but But the second is, um the bigger the price movement, um it's more profitable to to shift prices more to trade against more liquidity. So the bigger the price movement, you're also going to trade more units. right And in in in some kind of, you know let's say, um you know sort of tailor approximation or calculus kind of sense, for for small price movements, um the the amount of profit an arbitrator makes is roughly the square of of of the amount that the price moves.
00:31:04
Speaker
Now, if we can sort of um um ah imagine a scenario where let's say you have a pool, and let's say it's got a five basis point fee, and let's say the the price discrepancy against that between that pool and what you think the fair value is, is this is actually six basis points, net of fees and arbitrage could make one basis point, right?
00:31:23
Speaker
Now, if there was only one arbitrator in the world, what that arbitrator should do is potentially they should wait. They should not take this one basis point and they should wait because that one, um you know, that pricing discrepancy could increase or decrease. But because of this convexity, because of this quadratic um ah payoff, the the expected value is higher if you if you wait.
00:31:43
Speaker
right Now, on the other hand, because the trade is obvious and because there's a universe of competitive arbitrageurs, you can't wait, because if you leave that one basis point on the table, someone else will ah will take it. right So, you know again, if if there was a monopolist arbitrageur, they would wait for larger price discrepancies before they ah um take advantage of that because of this convexity, but because of competition, they can't. However, if you have discrete blocks, that discrete block creates a friction where if there's a small price discrepancy, no one can trade, right? And so if you have, um let's say, a blockchain with a very long block time, the price movements will typically be bigger. The arbitrageurs are not allowed to sort of trade when they're when they're small. They have to wait till the block is generated. They will end up making more money because the um the block time is a friction which prevents them from ah from trading on the smaller price discrepancies.
00:32:36
Speaker
On the flip side if you have a um a blockchain like maybe something like a layer two with with very very fast block times then what will happen is the arbitrageurs will end up um um you know whenever there's a small price discrepancy as long as it um you know the the price discrepancy is higher than the transaction fee and other other kinds of gas costs or whatever the arbitrageurs will jump on it, but because there hasn't been as much time for the price to move, they will typically be making um ah smaller amounts of money. So, you know, faster blocks is is one way. um ah You know, and I like to sort of think of it as you're you're creating more intense um competition between arbitrageurs by, you know, kind of subdividing that price difference and forcing them to compete on kind of ah smaller price changes.
00:33:21
Speaker
Now, just to give the argument a little bit the other side, um ah you know, so, so, you know, ah you know, from, from, from, from maybe a TRATFI perspective, there's almost, um you know, no downside here. Everyone likes ah um faster blocks, right? like Wait, see, i ah can i can I just inter inter interject here? so i So I just want to provide a little bit of context for the audience. And so first and foremost, I think it's important to say you are talking about a result that you've shown in a paper that you have, right? ah automat Automated market making and arbitrage profits in the presence of fees.
00:33:50
Speaker
which is joint with Jason millllino Milianis and Tim Roughgarden, both at Columbia. And frankly, I think you're a little bit underselling the result, ah both, I think, in terms of the assumptions that that that are probably needed for it, and also in terms of whether it's generalizable to other things, other forms of MEV, other than lever. So I want to reframe it a little bit, and I want you know you to tell me whether it's a fair reframing, because it's your result. So maybe I'm just misunderstanding something.
Continuous Trading and Its Impact on MEV
00:34:19
Speaker
So I think the the best, the the the the easiest way that I find to at least think about it if I'm thinking about it correctly is take the extreme case. Let's imagine that there yeah that the blocks arrive instantaneously. So in other words, I take away your friction that you were just describing, right? Forget about fast blocks. There's essentially like continuous time action, right? And I layer that on with the idea that the price process itself is a diffusion.
00:34:45
Speaker
which is a very frequently used assumption, like you know this is not this is an innocuous assumption ah by and large. right And the way I think about this, um which it's a little bit to the sense that I also think this is more general than lever, I think you can sort of show the same result for other parts of MEV.
00:35:00
Speaker
is that in that world, if there is some value to be extracted, let's say an arbitrage, either value arbitrage is really what it is, I would say in your in in your paper, the way it's modeled, ah but it could be anything else. um Just there's some arbitrage opportunity. Somebody is going to step in and immediately drive it to the edge of the arbitrage bound, right? And so what you basically get is if you have a setting where you can act continuously, you're going to be stuck inside the arbitrage bounds because the moment because if you just happen to be outside of it, somebody would come in and and take that profit.
00:35:31
Speaker
But then what you get essentially is so then, OK, so so whenever you're if if you were ever out of the bounds, which won't really happen in the setting, um you'll end up right at the edge. But then, you know, the the process could move one way or the other. That is to say the ARB could sort of open back up, in which case it'll get shoved right back to the bound or it could kind of go the other way and it'll get inside the arbitrage bound. So basically what's happening all the time is either you're inside these arbitrage bonds, and nothing's happening.
00:35:56
Speaker
Or you hit the edge, and the moment you hit the edge, this is exactly the point that for the arbitrageur, they're they're exactly indifferent. So there's some value there, but they're paying it all back in fees. but And when you go to like longer and longer block times, you're basically saying, like oh, we hit the arbitrage bound, but they can't run the ARB. So maybe it's going to get a little bit out of the bound. And when you take the block times to to sort of to infinity, or the rate to infinity,
00:36:21
Speaker
you're basically sort of stochastically controlling how far out from the bounds it could get. And so to the extent that it sort of gets stuck at the bounds. So the perfect scenario is that the moment it hits the bound, you get the arb and the arbitrageur is actually perfectly indifferent. So yes, he he he he picked off the AMM, but he but paid everything back in fees.
00:36:40
Speaker
right and And so that I don't think requires, let's say, particular technical assumptions about convexity or this or that. right That mental model, ah it's basically a diffusion process um where where really the arbitrage value, the mispricing, like in your paper, the mispricing process following a diffusion I think is important, but I think that's a very fair assumption by and large.
00:37:01
Speaker
And if you can act continuously, and it's exactly as you were saying, like, I'm not a monopolist, right? If I don't arm this thing right now, somebody else is going to come around and arm this thing. But again, that's just basically the reality of blockchain. These are permits in the settings, right? And so um it it it It actually does seem to be like a ah pretty strong result. um Yes, we don't have blocks coming instantaneously. But I don't know if you have a block coming every three seconds on Arbitrum. ah Maybe you were quantitatively close to this. I don't know. That's an empirical question. But the point is, it she seems to me to be a very powerful result and very much generalizable. So for instance,
00:37:35
Speaker
um If I were to think about you know liquidations at a lending ah protocol, right why can't I make the same argument if the value is sort of moving as a diffusion and I can act the moment it hits at the hits the bound? so i i think I agree with you, but I think at the heart of it is some kind of convexity. right um in your In your liquid action example, um ah like if if it's you know I could make a dollar um um if I do the arbitrage now,
00:38:03
Speaker
But hold on. so let let let me so Actually, say I agree with you. But I think the point is the convexity you're describing is built into the reality of the problem. So for instance, if we're right on the edge right of the arbitrage bound and you go right, there's an arm there, like a strictly positive arm. Whereas if you go left, there's nothing there. but That asymmetry is what you're alluding to, really. oh that There is a convexity there. um So I agree with you. I think that convexity is is probably enough.
00:38:29
Speaker
I think in the AMM case, I focused on an additional convexity where if it goes in the direction where the price discrepancy increases, you you make super linear profits. So I think there's an additional effect in the AMM case, but I agree with you that if it's arbitrage in general, your your downside is zero and your upside could be increasing. That is in itself a kind of structural convexity. Let me say, though, by the way that, um so the thing about, I think, thinking about the the second piece of convexity you're describing is If you think about risk aversion, it's a concavity effect. right and so it can sort of it can It can dull that a little bit, but I think the other point like that i that I was getting at, it doesn't go away with risk.
00:39:09
Speaker
right like that jo so you know It's your result, but would you agree then that like if we have you know infinitely fast blocks, um you you are going to have all the arbitrage profits paid back through fees? um ah Okay, so they're like, I mean, we're we're getting a little bit technical there, but I think you you need continuous prices. So if prices were a diffusion, that would be compact. I think once you get below a certain time scale, jumps become important and um you still get, um you know, um price, you know, the the jumps happen and you lose when they happen and that doesn't go away. yes
00:39:46
Speaker
the Yes, no, no, absolutely. no i fully We're 100% on the same page. my My point really more so is that that was the the mental model of the continuous pads with the infinitely fast blocks is is incredibly powerful and so like exactly where I go from there is like this really is something that everybody should be paying attention to and it's exactly the point that okay so fine we don't really have infinitely fast blocks we might have three second blocks I don't know how how fast you know how fast they'll get and then on top of that yeah okay when we talk about really short time intervals then yeah maybe it's important to think about kind of like a more discrete process because like you know prices can't really be pi for instance they can't be irrational non-terminating numbers um but
00:40:28
Speaker
But then I think like the way it pushes research is, OK, let's start thinking quantitatively. How much of leaver is mitigated if all the trades happen on arbitrum tomorrow? like i think that's that And that question, I think, should be dealt with in the framework of your paper, because I think Again, as continuous paths is mostly considered an innocuous assumption. It's exactly when you get into like very ah specific, very tight things, like very short intervals. Now, like you don't want to do that. Microstructure, literature, and so on. Maybe you don't want to do that. But I think there's like a first order way in which your paper essentially reframes this problem in a really powerful way and gives a really powerful insight
00:41:14
Speaker
that you can see through the lens of the continuous paths with infinite block times. And again, I think that's generalizable to other forms of MEV, because you can interpret your like mispricing point about just the value of any other m MEV being between arbitrage bounds. And the moment it kind of gets to the edge, you have the arbitrageous like swoop in. And i so I would personally love to see somebody actually look empirically into like, for example,
00:41:37
Speaker
Yeah, if it's happening on a layer 2, like a first-come, first-serve, the Arbitrum is a perfect case, right? Layer 2, first-come, first-serve protocol, much faster block times than Ethereum. um how How damaging is is Lever there? How much is being paid back? By the way, do you happen to know? I mean, have you looked into empirically the extent to which ah it is mitigated just by the blocks being faster? um I haven't. Others have. So I know, for example, Austin Adams, formerly of Uniswap, has a
Fast Blockchains: Challenges and Opportunities
00:42:07
Speaker
paper. He doesn't do quite a lever calculation. He does his sort of own calculation. um But I think he shows the ah the the the losses are less on ah on on on on the faster
00:42:18
Speaker
L2s. So I think i think that's that's um that's the sort of known. um I think to be fair, the counter argument there um is you know um that you know when you have faster and faster blocks, um ah you know um geographic proximity starts to become important. You start to see the same kinds of things like colocation,
00:42:39
Speaker
and, you know, maybe low latency links and so on that you see in the in the TratFi world. And I think, um ah you know, people like to make the argument that perhaps, you know, if if if you had like, let's say 100 millisecond block times or 50 millisecond block times, um you couldn't be, um you know you have to be geographically co-located, and that destroys centralization.
00:42:58
Speaker
So if you if you want like sort of a global blockchain, there's a certain lower bound on on the timescale you can operate at, which is probably you know hundreds of of milliseconds. and But you know I think purely sort of financial considerations, if if you don't care about that decentralization, purely financial considerations would dictate trying to have as fast ah a blockchain as possible.
00:43:18
Speaker
I mean, so I can I can see all the and it's kind of a funny thing. I'm arguing your your paper and you're and you're giving a little bit of the other side. But but so I can see sort of all the points you made. But I also feel like.
00:43:32
Speaker
It's a huge step in the right direction. like You're talking about stuff that I think are in order of magnitude less important if those are the problems you're trying to solve. I'm not you personally. i mean like you know if If we go to a world where that's the biggest problem that you know there's that that people who are co-located are going to have a slight advantage of whatever, 50 milliseconds or something, I think that's a much better world.
00:43:55
Speaker
then the world in which we go oh god you know these people are able to capture large arbitrage gains because of the amount of price movement that we see in ether over twelve seconds.
00:44:07
Speaker
um I think it's definitely solving the problem like and and probably like moving it an order of magnitude down. yeah yeah I mean i I personally agree with you. I'm just trying to sort of steel man the yeah the other argument. but But yeah I mean I think if you think about it in um in a modern context like Ethereum um trading once every 12 seconds that's crazy. Like what kind of market you know ah does you know sort of ah sort of does price discovery like on on on that kind of timescale in the age of ah of of a computers. Even even you know the the trading pits in the you know New York Stock Exchange or whatever, back when humans did it, it was much faster than like one trade every 12 seconds. yeah Except that Ethereum was not made for trading per se. right right right so
00:44:50
Speaker
Amen trading was just it's just and it's a nice way to to use the blockchain but it's as you as you know is it's not really designed for that. right it's not That was not the first lot of concern. In fact, actually I think Solana explicitly try to be what was trying to be like like you know having the throughput of NASDAQ is essentially on the blockchain. right So that's kind of their promise.
00:45:11
Speaker
Not that I want to advertise for them, but and I remember that it as the <unk> the and as the sales pitch for that. But I think, yeah you know that in some sense, we're all converging on a very nice insight, which is that, on the one hand, there is maybe some advantages to having larger block times, as you say, right because you know you have the latency issues are a problem. On the other hand, if you have shorter term block times, then you start getting all of these considerations of colocation of some form Can you be very close to a particular you know um part of the market which where where all the trades originate from so you can catch them on the line? You would find a situation where the builder probably co-locate with Binance or something like that. right so and I'm not sure if that is ever desirable, but that kind of seems to be like one of the consequences that we could see.
00:46:06
Speaker
ah So that's why I was about to get to that, right? So we've talked about faster block times for now, which seems to be something that is requires a pretty heavy lift from a lot of different parties. and And as far as I hear, ah you know in know, Ethereum, there's no, at the moment, no consideration to change it systematically, right? And, you know, ultimately, but there's a limit for what Ethereum can do. It can go from 12 to 1 second block times, but there's a limit to that. Well, I think we have to add some context, though, right? Which is that Ethereum's plan to scale now is based on layer 2s, right? So most of the activity happening on Ethereum actually doesn't happen on
00:46:44
Speaker
main that And that's one of the reasons why I think they don't care that much about expediting the 12 second block time is that if everything's happening on arbitrary. or optimism or wherever, then it's not their block time that's the binding factor. Fair enough. But you know here's the thing, right? So once we actually run an AMM like an arbitrum, at some point it will look exactly like NASDAQ, right? And then the Ethereum blockchain just becomes a DTCC, right? I think there are AMMs already on there, but. Yeah, no, no, no. but No, no, no, they are. Of course they are. But we're just trying to say it's conceptually we're really moving into a model where you have
00:47:18
Speaker
I mean, it's not, it's, we don't call it centralized, but that's really what it becomes, right? It's going to be somewhere, it's going to be the server that collects all the trades that come and then we have co-location for it and all of a sudden we're moving back to the old world. Well, no, no, okay. So actually here again, I'll get off of it in in a moment, but if I can go on one more time on this. um So in fact, let me say, yeah you know a lot of the other A lot of the people working in mar microstructure like where where Andreas works will tell me things like, this is exactly why AMMs are you know not a good place to trade. that they're not They're not competitive compared to centralized exchanges. right um And so, again, having these kinds of problems, I think, is probably a step in the right direction. But by the way, it's not turning i think AMMs into centralized exchanges, because
00:48:06
Speaker
Imagine if you are, and this is actually kind of in line with Andreas, your work actually, right which is that if you imagine a liquidity pool with a ton of liquidity, because one of the problems is sort of a coordination issue. It's like and if nobody's trading at the decentralized exchanges, why should I provide liquidity there? And if nobody's trading liquidity, why should I trade there? right So if you imagine a world in which these are your biggest problems, which are anyway the problems that centralize exchanges, and you are able to then get a lot of liquidity at these places, then it seems to me Yes, you have some of the problems of centralizing changes, but you don't have the problems. ah you don't have You don't have other problems, for example. You don't have overhead costs, right? You don't have, um and and you know, going to the maybe the paper that I'm alluding to of Andreas' that shows that, you know, if you do have um if you do have a lot of ah people providing their liquidity at these exchanges, you can actually get transactions costs um below what you see at centralized exchanges. So I'm not sure that's a bad bat. Like that actually might be
00:49:01
Speaker
a good path. It doesn't level the playing field for everybody, but it's not like the playing field is leveled right now for liquidity providers, right? yeah i mean you know There's obviously concerns that we have in existing markets anyway. right so um you know um that There's lots of questions around that that require answering. and Now that we're talking, i'm I'm after thinking about my own work, which I don't really want to because I want to talk about CMX work now.
Auction Mechanisms for MEV Allocation
00:49:25
Speaker
but i so because I know i know yeah you mentioned you want to move off to some of the other points that Siamak had mentioned. He said there are four mitigation strategies. But there's one last thing, Siamak, you can choose to go to go along on this or or or short on this, but because it's curiosity of mine, so I want to ask. um is you know As you know, um and the Ethereum blockchain in particular is thinking about ideas where they sort of auction off um slot by slot, the right to do MEV, essentially detaching the role of proposing the block and receiving the senior ridge rewards from MEV effectively. So that at a conceptual level, just think about that. So ah forget about how they do it. We don't have to talk about the specific plans to do it, but the idea being that we have um a way to auction off the MEV value block by block, right?
00:50:18
Speaker
um But then this this gets into i this point that I think you know you've thought a lot about, multi-block MEV. right So going back to this why faster block dimes are useful is um there is a competition aspect. right like I don't have a monopoly to run this ARB, so I better run it now. Otherwise, you or Andreas is going to run this ARB on me, um on whoever's getting ARB. But if you can buy out 10 slots in a row,
00:50:48
Speaker
then ah you're a monopolist, right? You you can do the ARP, right? So how does, I mean, well, I guess one question is, does what you find with Tim and Jason ah have um something to say about whether whether any blockchain should go in the direction of allowing MEV to be auctioned off slot by slot, certainly in a deterministic way. So as you know, there are certain protocols, well, find certain ideas like execution tickets where you don't really know what slot you're going to get, but then there are also like these things called execution auctions where you know exactly what slot you're going to get.
00:51:24
Speaker
So why can't I just buy 10 slots? No. I mean, can you say anything about how your work informs that entire discussion? so It seems to me they're very related. Yeah. So, you know, under our theoretical assumptions, which are mainly sort of Black-Scholes, you know, geometric Brownian motion type assumptions, the amount um that an arbitrager makes per block is roughly the um the length of time um between blocks to the three halves power. And in particular, the important part about that is it's super linear.
00:51:53
Speaker
right And it it comes from that that same effect, right um that you know these these type of ah competition or convexity or boundary effects that that we discussed. Now, if you have an environment where I can guarantee that I control the next two blocks, right what I can do is I can censor the first block in a way such that nobody else does the arbitrage, and then I can um do the arbitrage at the end of the second block.
00:52:18
Speaker
And what I will have accomplished by this, by controlling those two blocks, is effectively I've doubled the um the the block time. And so I can make more than 2x the profits because of this ah um kind of super linear structure. So I think um the concern with these mechanisms is exactly what you say. If there is a way where, um ah you know, I think um you can sort of guarantee that I control a sequence of blocks, then what you will do is you will sensor and you' will you'll effectively through censorship convert that to a situation where the block time is longer and where we have the opposite effect of everything that we were both just arguing for.
00:52:58
Speaker
So you don't want a monopolist arbitrageur. You don't want a monopolist arbitrageur and um you also don't want, um I don't know what the you know the right language is to use, um but you don't want a monopoly on block construction, on consecutive block construction. right but so So every block, is there's always you know some proposer who's ah ah you know ah ah you know a monopolist at that instant, but you don't want that instant to be a that that person to be able to extend it in a ah and in it in a deterministic um um way. And so I think, you know um especially with you know um ideas like execution auctions, if you can sort of predictably bid up to um you know control some sequence of slots in the future, um you know what my model says is there's incentive to do that.
00:53:38
Speaker
And someone will do that and and try to control multiple blocks and and and and use that to extract super linear um ah profits. but sort of one this is maybe tangential the one one curious thing about that is so In that case, there's an argument that the the value may still not go to the so-called arbitrageur because you'd have to pay to win the auction. I don't know how competitive the auction is. I guess it would depend on how relatively effective everybody is at doing the same multi-block MEV thing.
00:54:08
Speaker
but But it still wouldn't go back to liquidity providers, which seems to me to be kind of the fundamental problem here, right? Like if we want these things to thrive, it's not just that we don't want the people who are kind of executing it to get most of the value. We also want the liquidity providers to not get picked off or effectively to be repaid for getting picked off. Is that fair?
00:54:30
Speaker
That's right. and and And I think even now um a lot of the ah um ah or I shouldn't say a lot. um Some of this M.E.V. does go back to um the the proposers right through various mechanisms either through you know I mean depending on the block chain either through ah you know priority gas auction or through you know that booster through the builder market or or sort of whatever. So the arbitrageurs do pay some of that out they do not- you know. By virtue of competing with each other through you know priority fees or bribes to you know builders or or whatever mechanism. They do pay some of that out but the critical thing is it doesn't go to the LPs exactly what you said. Right and so because it doesn't go to the LPs it ultimately- still results in losses to the LPs. And um that in turn will result in less you know-
00:55:17
Speaker
um downstream onm liquidity on the AMMs and ultimately less you know trade and less welfare gains and so on. yeah not Not only does it not go to the LPs in that case, it does also just doesn't go very widely. right like the ah the the The builder market is is extremely consolidated.
00:55:33
Speaker
But the sort of the validator market is not as consolidated, but still quite consolidated. right So who's getting it? Lido, I don't know. ah um but and you Right now, like someone like Lido could potentially, i mean and you know I don't know what they control, but probably 30, 40% of blocks in a probabilistic way, they could be extracting multi-block MEV.
00:55:55
Speaker
As far as I understand, they don't, but um you know they could. And um I think potentially the worry with ah um some of these proposals um for you know various kinds of auction mechanisms is they could accelerate that, right? Like right now the financial incentive is there for someone like Lido to do that. Maybe um they don't have the technical infrastructure, or maybe the social consensus isn't there, right? But um you know I think the worry would be, but you know some of these mechanisms might just you know sort of accelerate that that process.
00:56:24
Speaker
you know but I'll stop beating this dead horse in a moment I promise but but this is why I think that result of yours is so powerful like I don't think enough people understand it because if they did then all these things that we're talking about.
00:56:36
Speaker
I think would be much more research like this seems to me that this this stuff that should be happening, because we do need to better understand, like the block times are hugely important for the reasons that you said. And when we start talking about things like auctioning off slots, it immediately becomes a thing about well,
00:56:54
Speaker
you know, that can sort of reverse faster block times. And to your point, maybe something's already going on, right? Like Lido does, I mean, did they control, I think, 30% of the stake stake either right now. So they probably do get drawn in consecutive slots with reasonable probability. And I don't know, maybe they are doing something. maybe Probably they they would do it cleverly if they did.
00:57:16
Speaker
um but but I don't think there's enough work on it and and hopefully you know hopefully maybe this podcast will but lead to more people looking into this because I think it's i think it's the right direction, practically speaking, certainly. Yeah, I mean, for for the empiricist in me would also say that and you know you kind of need a model to to create the identification, right which is a little difficult in this environment.
00:57:39
Speaker
um But um anyway, maybe we should move on to, so earlier on, you mentioned that there's various different ways how you can mitigate MEV. I think we've um now fully understood the block time issue, hopefully.
00:57:55
Speaker
um what What else is there? Right, so let me list the um ah three others and then um week we can sort of go through them. um The next one I'll list is, let's just say a broad category of what I'll call dynamic AMMs. And so, um you know, that in general means improving um the A. M. M. S. a trading strategy- that could be through a couple of different mechanisms it could be for example adjusting prices- without trades so for example if you had Oracle updates maybe you could sort of leverage that.
00:58:27
Speaker
It could be um ah you know things like um you know more intelligently picking fees, um things like ah um you know dynamic fees, dynamic adjustment of liquidity. um It's in general like quite a vast design space, but like maybe let's do something smarter than let's say just XY equals K or like ah you know a concentrated liquidity pool or whatever.
00:58:48
Speaker
so you know that's maybe the second class and on the third and fourth class have to do with with auctions and those are on basically various flavors of auctions- for the right to up to extract them maybe- and I think that the idea there is- you know that that M. E. B. is there maybe it's it's it's sort of unavoidable but if we can create the right kind of a mechanism where the way that people on compete for it is not by let's say- you know I'm competing in- in you know, I'm paying um ah proposers and so on, but instead competing in some kind of auction where the proceeds of that auction go back to the LP, right, the the liquidity provider. So that kind of, um you know, speaks to the point where um a Fahad was making earlier, which is that, you know, um those are the people who are suffering the losses. And so they're the people who should be somehow, um you know, compensated for ah for for those losses. So there's a general class of auction solutions. I like to split them into two sort of categories. One is ex-post auctions. Those are auctions for the right to extract MEV, but they are conducted during the um process of block generation, right?
00:59:56
Speaker
So it's ex-post, ex-post means afterwards, but ex-post in the sense of you know the state of the world, you know what the fair value of the um the the asset is, you know what the finance price, you can run your models, you know sort of um ah whatever, you see all the users orders and conducting an auction at at that point in time and um that's um contrasted maybe with the ah the fourth form of MEV mitigation which is ex-ante auctions, and the idea there is you would conduct some kind of auction ahead of time. so For example, maybe right now we conduct an auction um among the three of us to extract MEV in a block that's going to be an hour from now, or maybe to extract MEV from 5 PM to 6 PM later this afternoon. right and so I think um ah those are the ah the flavors. and you know If you like, we can we can go into some of the proposals and you know some of the pros and cons Let's start with the thes stoppath the ah the during block, the the exposed auction. So essentially, how should I think about this? is like I notice that the price, so I'm saying I'm somebody who is an arbitrageur, I see the price on Binance have moved, or I know the fundamental value has moved.
01:01:05
Speaker
And so then I put in a trade which would move the price on the AMM to the fundamental value. I know how much money I made. and so now in order big But I'm really keen to get that money. And so now I'm going to bid for this particular block space that I that i need in order for for my my order to go in precisely into this block so that I can extract that MEV.
01:01:29
Speaker
And then so the idea is that there would be a mechanism by which you take my profit that I auction off. And then, so the the buildup, but presumably the the somebody has to pay back the remainder of the, of the of the head has to somehow pay this back to the AMM. Right, so I think the way, um ah you know, this is you know there's there's a bunch of different proposals and they all work different ways, but you know maybe just like at a conceptual level, the way I would sort of view it is, suppose you could auction off the right to be the first guy to touch the AMM in the next block, like the block that's happening right now.
01:02:05
Speaker
Right. So it's ex post in the sense that we all know what the price is. We see what the state of the AMM is like what the liquidity is. We see the finance price. So on you know we can each um ah we have um you know sort of fairly certain um ah you know of value. Let's say of of what that arbitrage is worth.
01:02:22
Speaker
and if we imagine a perfectly competitive auction where there are many of us, um in principle, we should be willing to bid up to the value of that arbitrage and we could um perhaps conduct that auction in a way where the proceeds of that auction are then paid back to the ah the liquidity providers. Can I ask about that? so so i mean I can understand the mechanism of how you could possibly get that to work in terms of how much I would be willing to pay I'm sort of not sure how would that would feed back to the AMM or to the liquidity providers because isn't it that the people that are processing these blocks are fundamentally different? What's the incentive to actually pay money back? Oh, I think the system would be structurally set up by this. so i think
01:03:08
Speaker
you so okay so um ah This auction would have to be run by someone, right? And because um the auction is on the same time scale as Consensus, because you're having this auction occur um every time a block is generated, um it would have to be run separately outside of the blockchain, right? So what we would imagine is we would imagine we have, ah like, let's say an AMM, and the AMM says, hey, the first trade on the block has to be signed by, you know, some committee that ran an auction off-chain,
01:03:39
Speaker
And I will only trade with that as as the first trade. And then separately, like you have some kind of infrastructure that runs it off-chain, and part of that smart contract would also be that the proceeds go to the liquidity providers. So that's the kind of way you would implement it, but that's also it's it's ah highlighting the main downside, which is that you have to have some additional infrastructure in order to run this auction. So either there has to be like a kind of centralized auctioneer,
01:04:07
Speaker
And so like I think that's sort of a little bit the way I've used you know things like um a cow swap. Or um ah you know it has to be um some kind of decentralized ah um auction protocol. So you know there's there's sort of challenges and in in building that. um Also, um you know um composability is an issue, right? So now we're doing all this stuff off-chain, we're running this auction. We can't have ah you know necessarily our trades be composable the same way that they are in a traditional smart contract where, you know, a trade is just a function call like, ah like you know, sort of ah everything else. So I think those are the the the two principal downsides. I think the nice thing is that it's an easy option to participate in, right? Because it's exposed, because everyone can see sort of the value of everything, um ideally it's it's maybe not that difficult to sort of value what that arbitrage is worth.
01:04:57
Speaker
and hopefully you'd expect more participation in that kind of auction and you'd expect it to be more competitive and to refund more of the ah the the value back to the ah the LPs. So in some sense, that's like the idea is there's sort of like a functionality, like a hook in the auction, in the smart contract for the AMM, which says if this the very first transaction in a block has to have a special signature of some form which we have to provide,
01:05:22
Speaker
And we're only giving you that signature. If you want the auction. you If you have actually, if you pay us back. So that's kind of. Right. If you want the auction and you agree to pay us the money that you bid in the auction, you inject it into the smart contract, which then distributes it to the LPs. I'm i'm just going to throw out that that that also, you know, because it has, to as you say, that has to be centralized, right? That all of a sudden exposes whoever runs this AMM also to all kinds of regulatory risk, right? Because at that point, you're becoming an active participant with a protocol.
01:05:51
Speaker
And then you have to pay maybe be answerable to like all the scammy token and scammy contracts that are running in an AMM and now all of a sudden you have to have much larger control that you have to exert here right so there's there's a completely new dimension that would come into play that maybe many AMMs and protocols would want to avoid a.
01:06:09
Speaker
Yeah, I mean, and there are people trying to build these in a ah ah decentralized way also, which would maybe mitigate some of those um ah risk points that you talk
Innovations in AMMs to Mitigate MEV
01:06:19
Speaker
about. But it's it's it's challenging. And it involves, you know, you know, a lot of infrastructure alongside the normal consensus process. So actually, yes, I wanted to add a couple of thoughts here. One is it's not it's not clear it is not clear to me that This can't be done in a decentralized way and essentially concurrent. And what honestly, what it makes me think about is AVS is at restaking. So essentially what you kind of need is like a committee of validators that are sort of in parallel doing stuff. right like If you think about the exercise that you have to go through to do some of the ideas that that that Ethereum is currently thinking about, like even just not having transactions censored in the same slot,
01:07:04
Speaker
It's a similar kind of process that you would need. And yes, like the traditional smart contract, like there are no validators associated with the smart contract, but this is exactly where something like eigenlayer could potentially be useful. And yes, there's a whole bunch of like infrastructure. I think, see, I'm like, you've been saying and saying that. that that's That's infrastructure, right? it's It's not as easy as, let me just write a bunch of Solidity code, deploy it and it'll work. But at the same time, it doesn't seem to me like it's, if anything, it seems to me like something worth exploring in the context of something like v4 plus eigenlayer. And there are projects doing that. like So one um project I'll call out is Sorella. They are building a top of block and you know sort of batch auction system that would be run in parallel with, the like let's say, the Ethereum consensus process and would ah you know attempt to address ah you know some of these issues. So I think it is possible
01:07:58
Speaker
um We have to sort of think about also, um ah you know, if you do that, then maybe you're just moving some of these things like, ah you know, timing games and latency and whatever, and instead of that being, you know, happening in the, um you know, Ethereum consensus process, maybe it'll happen in this auction. So you may have some of the same problems, just sort of move them to a different place. But I think i think the big win is exactly, um you know, what you spoke about earlier, Fahad, is that these type of auctions will refund the money to the ah the LPs. they'll they'll They'll send the auction proceeds to the LPs, um ah who are the ones suffering the losses, as opposed to you know other things like, let's say, MEV Boost, which is also an auction run in parallel, um a centralized auction but one to run in parallel. However, the problem with MEV Boost is the money goes to the wrong people. It goes to the proposer. Right. And so here's another area where I feel like there's a lack
01:08:51
Speaker
of work work in the sense of there's good work but then like the follow-ups haven't been done and I feel like there's a lack of clarity around so whenever I think about but what what we're talking about now these sort of these auctions the the primary thing that comes to my mind about whether it's going to work or not or whether it's gonna be useful or not is how competitive is this, right? And that's actually kind of hard to tell, right? Like, people will look at things like bids in in mev boost.
01:09:22
Speaker
But I mean, really, that's like a lower bound if you assume that the the builders are rational. like It doesn't really tell you how much they're capturing. um So do you have any sense of how competitive this process would be if implemented and maybe as sort of a comp, we could use stuff like how MedBoost is today? um It seems to me that like Malesh, who is also on our podcast series this season, um has work that we didn't talk about that seems to suggest that there's certainly aspects of MEV that are not competitive, maybe not lever, but um essentially value arbitrage, if I'm correct, one of his papers seems to suggest, is really a two player market. And they have heterogeneity across themselves too. So it's not like they're you know price fighting to the to the price everybody sees. And this also relates a bit to like, we've we've been talking about this checks tax R. And one of the really important points that I keep on revisiting is,
01:10:19
Speaker
i I get frustrated with it because it's such a misnomer and it's importantly a misnomer right like the c checks decks if you take it literally is a common value thing. right I don't have a different price at Penance than you do. I don't have a different price at Uniswap than you do. rights It's just a price there. But that's not really what's happening. right these guys these These players have inventory, and they're not actually doing the other leg. and how And if it's about me just buying at a relatively low price, it does make a big difference whether my inventory is very different than yours. For instance, if I'm already like massively long this thing,
01:10:56
Speaker
then maybe i don't want to load up on it even if it i can even if i'm a little bit below fair value where i buy it um and you don't mind that and now that's going to create sort of a heterogeneity that's going to make you and me competing against each other not that competitive potentially right and so i guess my my main question here is um how competitive.
01:11:17
Speaker
um do Would you anticipate these auctions being in any context you can provide around? Yeah, so I think it's instructive to look at um a different type of arbitrage, which is atomic arbitrage. And um so that would be a kind of arbitrage where maybe you trade against different pools on the same blockchain. And it's atomic in the sense that all the transactions are, um all all the trades are within one transaction. So you know when that um um you know transaction lands,
01:11:44
Speaker
you're sort of guaranteed to sort of um um um ah make money. um So so that area is extremely competitive. My understanding is something like 90 plus percent of the profits are paid to sort of validators. Now to contrast that you know whatever you want to call it. um Dex XR. I like to use lever um personally. um of Of course ah um you know ah that that's because that's our terminology.
01:12:07
Speaker
um ah you know so on of their I think structurally there's issues with that boost- where- you know- you know it's it's not competitive. Right and where you've you've essentially got like maybe two to three builders sort of dominating it I think molasses done a lot of work to sort of.
01:12:24
Speaker
explicate that. But I think it has to do with the fact that, um ah you know, there's incentives to sort of ah for the for the builder themselves to sort of do this, right? Because by um by having the builder um um ah put in the ah um like the the the the trade, which sort of generates the arbitrage profit, they can sort of act later and they can act after seeing other people um ah You know bidding for that for that transaction and that gives them an advantage and it creates like kind of a vector for centralization around builders. Right. So um I think that has to do with the specifics of mev boost. I think if the auctions were done differently potentially um um ah you would not have those um kind of centralization um of vectors.
01:13:09
Speaker
But um ah you know again, you know the the the devil is in the ah but the the details. If you imagine maybe you have another sort of decentralized auction system with another set of validators are done in an AVS or whatever, you know you'd have to sort of study that to make sure there's not an incentive for maybe um a node that's acting as an AVS somehow to also be the guy that's doing the arbitrage and somehow create a vector for centralization there. But so I guess the fundamental question is whether you think that there are Like this is ultimately skill based enough that you can anticipate heterogeneity is therefore arising naturally and.
01:13:47
Speaker
And not, you know, not everybody is participating as well. So it's not like a system where you can think about a countably infinite number of players or a free entry sort of thing, or even a large number,
Evaluating MEV Strategies and Market Insights
01:13:58
Speaker
right? Like if you think this is very skills specific, um and it does take talent to do these, at least that the big ticket items for for Mev, then it seems to me that going in the auction direction could end up being uh not as effective let's say as people would like it to be whereas if you think because you like sandwich decks come up a lot in this context where people will say well sandwich decks are trivial anybody can do them but then you know not every sandwich is a front run a victim and a background they're uh more involved sandwiches um and so there are a bunch of people who specialize in that so i i don't know actually maybe the question is do you know of any work that actually does think about
01:14:38
Speaker
like quantifying metrics of competition and maybe like you you reference a different a few different types of mev but where is most of the mev because i guess the point is that if 90% of the mev is coming from stuff that's not ah competitive, where it is very skills-based, then auctions seem like less compelling, I think, of a solution. um Whereas if most of it's coming from stuff where you can imagine that if there would be a lot of competition and homogenous values, then it seems like a great solution.
01:15:10
Speaker
So um i' I'll point out another Sorella project which is something called Bronte's. They sort of released this recently to um sort of quantify um um ah some of this more complicated map. Like atomic map is pretty easy to quantify. um This type of map is not so easy to quantify because if you see a ah trade on Uniswap to know whether that's an arbitrage trade or not you need to have reference to Binance prices and when do you look at that price and it's different data source and so on. So um there are are people trying to sort of quantify that. My general sense is that first of all it's the biggest form of MEV. um this This type of lever or or value extraction MEV. I think also it it is less competitive than sort of um ah you know other types of MEV. But I think um you know again there's there's a question of how much of that is because of the structure of MEV boost that creates these kind of um centralization effects.
01:16:06
Speaker
versus whether it's, ah um ah you know, um ah like sort of specialization or skill. um So, you know, I mean, I think the starting point of um is is that there's these very easy arbitrages, like if you see a big price discrepancy between, let's say, Binance and a Uniswap AMM that are sort of obvious. I think that is sort of um very competitive. But, um you know, because many people can do that, I think the next level is to sort of say like, okay,
01:16:32
Speaker
um can I build more sophisticated models for what the fair value is? and And more specifically, I don't really care about the fair value now. I'm going to buy on Uniswap, let's say, right now. I really care about what the fair value is whenever I'm going to unload this, right? So you start to enter into the area of, you know, sort of more traditional market making and quantitative trading where, you know, sort of people will not only sort of look at what the value is now, but to sort of start to predict what what is the value going to be an hour from now and and, you know, leverage there. And so there I would expect there to be more heterogeneity,
01:17:02
Speaker
right and in in the same way that there's you know more heterogeneity in- trad find market making in quantitative trading and they're clearly. You know some people who are better at it than- than than others. But you know sort of having talked to some of the people who- who do do this type of- a trading today. My sense is about 90% of it is the obvious stuff. Like just look at what the price is on Binance and how different that is. And then maybe 10% of it is being clever. Either being a more clever in terms of- managing the inventory you accumulate and your trading strategy, or being more clever in terms of forecasting future price movements, or you know ah other types of of cleverness. But of course, that's all you know anecdotal. Now, I'm just going to chime in at this point. um I'm going to say a few things. So, listening to your conversation, in particular, some of the things that you did you say sort of in side sentences here, Marc, is that
01:17:55
Speaker
You know, is anybody here in the audience actually is a researcher market microstructure that but comes comes very clear that is an enormous amount of work there and actually a enormous amount of really interesting and fascinating questions about this. um Because there's there's so much information, there's so many little facets, the little auctions and how people actually get their trades included.
01:18:14
Speaker
which is something that we yeah we're not quite used to. um But there's probably, we we're going to see more of this because of the of the nature of and the growth of blockchain and decentralized finance. And there's a lot of work
Reflections on Market Microstructure and Academic Lag
01:18:26
Speaker
that remains to be done. And what I also see is how far behind we are as a community actually in terms of actually having this work and and and casting it into academic work, right?
01:18:38
Speaker
ah So, and with that, I think maybe it's ah it's a good time to sort of slowly wind down the discussion because I think we have learned an enormous amount about all of the different various different facets, where competition happens among different participants, what kind of arbitrage exists and how you can actually benefit off of it. I think in many ways, this this podcast has almost opened up more new questions for me than than it answered, even though many of my questions have been answered, right? I should say that.
01:19:07
Speaker
um But yeah, so this, I think this was, this was really fascinating. Thank you so much for your, for your time and patience for that, yeah like and having this in-depth discussion. and Thank you for for having me. And by the
Conclusion and Closing Remarks
01:19:19
Speaker
way, I should say um you guys are providing a great service in terms of bridging the gap between the understanding in academia and sort of what people, what's what's going on in the in the real world um and and what people are implementing. So so hats off to to to you guys and to CBER. Well, thank you. we're We're trying. We think it's actually, this is actually kind of critically important and also very fascinating for us as researchers.
01:19:40
Speaker
And for you, too, obviously, because you are you actually doing the bridging all the way yourself there. All right. Well, with that, I'm going to conclude. And thank thank you so much, Seimai, for being here. Faha, thank you for co-hosting and for all of your discussions. And we'll see you on then all the next podcast.
01:19:59
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.