 Welcome, thank you all for coming. A friend of mine recently shared this picture with me with the label MEV. And sure enough, it looks, it's very reminiscent of the mempool wars between searchers trying to snatch an opportunity from each other, try to get early, try to overbeat the other one. But is this a good metaphor for MEV? Is this actually MEV? People in the community now are very MEV aware. Proof of that is that there's 18 talks in this devcon around the topic of MEV. There were zero in the last, so that's quite a testament for this. Despite this, I would argue that we're not 100% sure of what MEV even means. This is a poll from last year. So which of the following is not MEV? There seems to be rough consensus around front running, but much less so around transaction fees, liquidation rewards, and all of these are MEV. It's very evenly split. This one is more recent, right? Like from last year to this year, there has been many developments, as you all know. This is a recent one from last month. Are block rewards a form of MEV? This is pretty impressive. It's almost half and half, right? So it's really a contentious concept, right? This is perhaps a more trolly one. Conference swag is a form of MEV. I won't go through this chart. It's a lot of fun. I encourage you all to check it out. Essentially, there's two dimensions that are taking into account here. One is value and one is extraction. So depending on where you see it on each of these variables is what you should consider MEV or not. But are these the only dimensions that are relevant? So again, doesn't seem like there's a lot of consensus. This is the ultimate one. I hope nobody got robbed. I really hope so. Yeah, so I think this meme illustrates... This is also my favorite meme format. It's not mine, by the way, but I think it illustrates what I'm trying to get through to you today, right? We've been calling MEV a lot of things. So in this talk, in my work in trying to formalize MEV, I've been argued that there's no consensus on a formal definition of MEV, and there's many around there and we're using different things, and it's hard to come up with a formula. But in doing work for this talk, I realized there's no consensus on the concept, on the intuition of what we want to call MEV. So in this talk, I want to go through different ideas and say why many of these are not MEV, and crucially, this talk is tightly coupled together with my friend Shin's talk next who will tell you what actually is MEV. So I encourage you to stay. He's going to roast me and tell you why the things I tell you now are actually wrong. So you should definitely stay. I'm going to be a bit of a shorter talk and he's going to give a longer talk, and perhaps if DMC allows us, we can do questions at the end. So since you all said that you were going to stay for his talk. All right, let's get going. So yeah, I forgot I have a cool animation transparency. Okay, just to tease Shin's talk. Okay, let's get going. So we're going to start with a trivial example. Are all the things that are MEV also all the things that are value? Is value the same as MEV? This is a trivial example. We can easily argue that this is not the case that there's more to MEV than to any value. You can think of picking up a $1 bill out on the street. We wouldn't want to call that MEV. Hopefully you all agree. But this is still value. So again, this is a trivial example, but just to set the stage of the type of argument I'm going to be doing. And critically, this is all intuitions. There's no proofs here. I'm just trying to hone in a concept and it's up to all of us what we decide to call MEV. It's a matter of choice. There's no proofs here. Okay, so first non-trivial example. Is MEV equal to a set of value on the blockchain? So if you think of the definition, the original definition in the Flashwatch 2.0 paper, there was value that can be extracted by miners, by censoring, reordering, transaction. So there's a lot of block-changing concepts there. Transactions, miners, etc. So it might well be that what we're calling MEV is nothing else than value on the blockchain. And sure enough, there's things that are value on the blockchain that are also MEV. Arbs are one of them. We're using Arbs like a canonical example of MEV. Arbs like talking about arbitrage on a blockchain, for instance. But are there things that are value on a blockchain that are not MEV? Well, here one could argue that, okay, if I do a simple ETH transfer, then the miner has no way to extract that. So we can call that value on the blockchain that it's not MEV. Although put a pin on this idea because we're going to come back to it because it's not as simple as it looked. Even a simple ETH transfer can get trickier. But in principle, we can say, okay, there's some value on the blockchain that's actually not MEV. How about the other side, right? And let me also say that mostly for the question of this is not MEV, we're going to be thinking of the right side of these slides, like the inclusion of the pink set in the white set. But we're going to find interesting things along the way on the left-hand side too, right? So we're going to also comment a bit on those. So in this case, is there MEV that's not value on the blockchain? This is again a matter of choice, right? But I would argue that we are striving for a more abstract general concept than value that's strictly on the blockchain. I don't see anything special about blockchains in terms of value. Blockchains are certainly special in terms of coordinating devices, but I don't see them as providing value, a characteristic that's so special, right? So in general, we would like to think of MEV as something more general. And here the canonical example you can think of is a centralized exchange. In a centralized exchange, you have all the orders coming and the operator can decide to extract value from users. Like Robin Hood does, that's the Robin Hood business model, right? So we might want to call that MEV in a matter of definition. You might not agree. Feel free to come and chat if you don't. But in principle, I would argue that perhaps we want that. One subtle point here is that we might need this if we start quantifying a centralized exchange versus decentralized exchange arts, right? So if we have value flowing in and out of the blockchain, then maybe when we start quantifying this kind of arbitrage and so on, we might need to call that MEV if we want the whole calculation to be sound, right? Okay, so value in the blockchain doesn't seem like it's the best candidate for our concept of MEV, so let's move on to the next one. Is MEV the same as front running? Okay, we start with the intersection of these sets. Trivially, ARBs, again, as we saw, I'm going to use always ARBs here. The first leg of a sandwich, for instance, you can think of it as an ARB. This is clearly MEV. This is clearly front running. We start with the left side here. Also, obvious point, there's clearly MEV that's not front running. You can think of back running, trivially, reorgs, other things. So here the tricky part is on the right side. Are all things that are front running MEV? Okay, and thinking from running here is you learn some information from other parties and then you are faster than the other in catching or capturing that value. Is that MEV? Again, the rugby example in the first slide, right? So here you're going to take two paths. You can say, okay, let's look at the rugby example. Okay, maybe I don't want to call that MEV, even if it has the characteristic of being faster than the other person. And so there has to be something else for us to call being early. Because being early, again, doesn't have that much to do with the structure of the value itself. So here, okay, what's the missing part here? Perhaps we can think of a coordinator, a coordinator that is doing the allocations of value, right? And in the rugby there's no really meaningful coordinator. You can think of gravity as a coordinator, but it's a coordinator that cannot extract. It's like a damn coordinator. So here if you have a smarter coordinator, what happens is that the game theory gets more interesting, right? So the MEV concept becomes much richer. So here, okay, this is one path you can go down, and you can say, okay, not all front running is MEV, because for it to be MEV, we need some extra player. So this is perfectly fine until you start thinking, and, okay, you can think of the example of the arbitrage chain, right? In arbitrage, you have a first-come, first-serve centralized server that cannot extract, and people are doing arbitrage. So if you go down this path, considering that the coordinator there cannot extract, you would not be calling Arbitrum ARPS as MEV, right? And this is something many people would frown upon, right? And in fact, the avalanche devs at some point were saying, okay, ARPS are not MEV, and the searchers were laughing because they were saying, okay, call it what you will, we're making millions of dollars out of this. So I think there's some consensus that, okay, these type of things are MEV. So again, tricky contentious point is a matter of choice where we want to call that MEV or not. Okay, so this, again, question is the set of things that are front-running included in the set of things that are MEV. And we'll come back to this at the very end. Okay, next set. Is MEV the set of things that are MEV equal to the set of things that are permissionless value? Here's an example in the center, ARPS, right? This is permissionless, anybody can make an ARP, and it's also MEV, canonical example of MEV, so that's fine. How about the left side? Are all things that are MEV permissionless? So when I worked with definitions and formalization of MEV, I argued for the importance of permissionlessness in the definition, because there were many definitions out there that included the notion of a player that can extract the value, but this is very tricky. And it's tricky because you can think of an example where you have an airdrop, right? You have an airdrop that's only entitled to you. And when you are going to claim the airdrop, the miner cannot do anything, right? So there's no way that miner can steal that value from you. So I would say, okay, permissionless looks like it's important in the concept of MEV. Although if you think of another twist to that example, say you have an airdrop which you can only claim at a specific block height, and say that the proposer of that block knows about this. So then they can credibly extort you and tell you, look, if you don't give me a share of the airdrop, I won't include your transaction, right? So then, okay, after all, there might be some MEV character to this, this extortion capability of the coordinator, right? And this might look like a very contorted example. Okay, after all, is there any airdrop that is only valid for one block? But if you think a little bit deeper on this and if you use some numbers, you can think of the earlier ether transfer from the earlier slide. And there, if you have a collusion of, say, 10% of the validators, they could in principle extort you for 0.4% of the value. Sheen has run these calculations, right? So 0.4% of the value is a lot, right? It's more or less the same as the fees that you have, you pay like swapping or uniswap or whatever. So actually this example, again, that looks contorted, is not and could very well happen, right? So extortion is a real thing. Okay, that's again for the left side. On the right side, this is not too interesting. We have the example I gave you earlier of the dollar bill. You pick up a dollar bill, that's permissionless value, and that's arguably not MEV. But okay, that's an example that's outside of the blockchain. So perhaps we want to consider permissionless value on the blockchain, how this set of things. The two examples I gave you for the intersection and for the MEV side are on the blockchain, so we know this exists in principle. So how about the other side? Are there things that are permissionless value on the blockchain that are not MEV? This is a question for you, actually. I have an example. So if you can come up with an example, again, of things that are permissionless value on the blockchain that you wouldn't call MEV, please let me know and we can sort out that set inclusion there. So this is one candidate set to consider for characterizing or half characterizing MEV, if you will. Okay, thanks for bearing with me. We've been through a lot of Venn diagrams. This is the last one. Okay, how about value extracted by a monopolistic coordinator? Again, I've argued for the importance of a coordinator because it changes the game theory. It makes it more fun. So here again, critically, the coordinator needs to be able to extract for that game theory to be richer. It has to be monopolistic because also for the extortion I showed you before, if it's not monopolistic then this doesn't work. So in the intersection, we have the same example as before. Arbitrage is an example of this. Monopolistic coordinators come naturally in blockchains like the proposers are that. Okay, so how about the left side? Are things that are MEV and that are not value extracted by a monopolistic coordinator? And here it depends on your earlier choice on the arbitrage example, the front running example. Okay, if you decided to call arbitrage, the front running example, then this is an example of an element here because it's not value extracted by a monopolistic coordinator but you're still choosing to call it MEV. Again, a choice. Finally, let's look at the other side. Is there value extracted by a monopolistic coordinator that is not MEV? Can you find examples of this? I haven't, so again, a challenge to you all. Please come talk to me if you find an example in this set there. And again, we can sort out this question here. Okay, so we've dealt with the concept of MEV. We came up with three sets that look like might always be MEV. This is of course not a full characterization just to sum up value extracted by a monopolistic coordinator, front running, permissionless value in the blockchain. This is quite an unfortunate picture, I would argue, right? Because there's these three disjoint sets and how are they related to each other? This doesn't look like a unified theory of MEV, right? One would ideally like to have a more concise idea of what MEV is, right? And these are only, if at all, partial characterizations. So in the talk that comes next, she is going to share a little bit of that and try to come up with this was more like an outside-in exploration of the idea of MEV. She is going to share inside out a more constructive way to think cohesively about MEV. So again, thank you so much for coming. I appreciate you all being here.