The theory and practice of web3 auctions | Scott Duke Kominers and Michael Zhu

Transcript

[Music] thank you the afternoon block is in on auctions and it's in two parts first time to talk about auction Theory and then Michael is going to talk about you know Unchained auctions in practice and you know the morning session is easy to motivate right like why do we care about business strategy well we're all building businesses and we care about being competitive and we care about maintaining competitive Advantage um auctions you know sort of by contrast feel like a much more Niche topic right um so let me let me give sort of two quick points of motivation just so you can think about why we're actually

studying this right now um first of all solving allocation problems is just one of the most fundamental purposes of markets in existence right like so much of life and and our interactions and like everything we we sort of face in terms of incentives is tied into questions of allocation right like how is it that we allocate resources to the people who value them the most to the protocols they can make the most effective use of them to you know and how do we figure out like how things should be priced how much they are valued by whom and so forth like this is like a fundamental purpose of markets

it's perhaps you know it's arguably the most fundamental purpose of markets right why do we bother enabling exchange and trade in the first place well because we want to get resources to the in assets of different forms to the people who value them and so one reason we're talking about auctions um is that you know many many web 3 businesses in one fashion or another even if it's not like sort of the traditional just like you know some some people like you do anything market and Marketplace business you think like you know Financial trading platform but like also you know the first day in my MBA classes you know usually about online dating applications right like dating is a market you know dating applications or Marketplace

platforms same thing here like you know some web three businesses are literally like running auctions all day right if you're you know sort of you know selling distressed defy assets or something like that um and others it's not that they're running auctions all day but they're solving complex allocation problems that in effect are you know that our auctions under the hood you're trying to figure out like you know which protocols should address you should like do a given computation that's actually like fundamentally like an allocation problem as well um and you're opting to be solving it via something that's an effective procurement auction uh and then last just a footnote you know sort of as we're thinking about why we're here and why we're doing this

um the type of thinking that we're going to work through in in auction theory is utterly like you know representative of like everything we were talking about this morning about like how do you think through an incentive design how do you understand like how the market or individual users are going to respond to changes that you make uh and so there's that too right this is also sort of like a like a brain stretching exercise for thinking about just like how does incentive design in web 3 work and then Michael's component is going to be utterly critical because as all of you know theory is great in theory but when you bring it to practice things change right and this has been true by the way in auction design

forever right like um you know they're really interesting famous examples for example of collusion by the players figuring out how to like send each other messages through the bids right and like auction design like major auction designs like Wireless Spectrum like you know options have been modified because after you know after some round where like people use trailing digits to send each other messages about where they wanted them to stop bidding right like theory has no space for that sort of intuitively right like your your model like your auction doesn't imagine that people are going to creatively game the system around the incentives you set up in the first place using tools that weren't in your model right like you know when we when we model how to run an

auction we don't normally think about the digits people are entering their bids in as as part of the game but as we all know the second you go on chain people can game anything they can see all the code they can think very carefully about what to do and conversely doing auctions on chain introduces new constraints and New Opportunities um which Michael will dig into in a lot of detail so this is like the fundamental purpose of markets it's something a lot of you guys are going to have to do and a lot of like a lot of your teams are gonna have to do in one capacity or another and it's also like you know an example work through of how do you go from Theory to practice in Marketplace design okay what's an auction earlier we did

what's a platform auctions are about allocating property or resources or whatever they allocate according to bids a common misconception is that auctions are always about allocating to the highest bidders this is not necessarily true although many auctions we interact with are also about that and there's a virtuous reason which is the highest bidders are often the people who or entities with the highest value but you do not always want to allocate to the highest bid or an auction is just like you collect bids and you somehow use that to determine the allocation so you know your your first bidder shows up and says I've been 15 and the second bidder says I'm in for 20. and then your third bidder you know 75 um that node is going to win the auction

there's no bidder after 75 if we're allocating the highest bidder at least that's the one who gets it wouldn't know why might you not want to allocate to the highest bidder maybe you care about geographic diversity of your participants or something and and our you know our green bidder is is further you know further away from the others and so you know think about like you know for Network robustness purposes you might care about waiting bids by you know the uh you know by you know some characteristic Vector of diversity constraints alternatively I think about sponsored search right like the ads um you know the ads you see on on search engines and so forth aren't just whoever whoever bid highest but rather they're

they're also ranked by the likelihood that you're actually going to click on them right maybe some Advertiser is going to be at really high just to like put something in front of you that you don't actually want to see uh you know sort of just very spammy you might want to screen for that reason but mostly for today's purposes we're going to be talking about um by the way note sorry I know I'm like I'm so far down the footnote rabbit hole and we haven't even talked about the content yet but let me just quickly finish that thought right um can you tell I'm really excited about auctions I really like auctions um so note that there were two very different reasons that you might not want to allocate to the highest bidder but they're both about

creating the most value possible through your allocation process right one of them is you need a diversity of winning bidders for the security of your system or the breadth of your community or whatever and the other is about you know you want the ads that the user sees to be ads that the user actually like wants to see which drives you know virtuous user Behavior like clicking on them and coming back to the platform and using it over and over again so in some sense you're still trying to achieve the highest value allocation it's just not always the case that the highest value allocation corresponds to the people with the highest willingness to pay and why is that

you know sort of willingness to pay isn't you know we think of it intuitively as economy like the economists often intuitively think of willingness to pay as being a measure of value but it's not really it's like value like an individual's value convolved with um like interactive rather with their ability to pay right somebody who's like much wealthier might have a much higher willingness to pay for something just because money doesn't matter as much to them and it also doesn't account for social externalities right it doesn't account for the overall welfare of the system like that there's a value to having you know your your nodes in many different geographies but so the goal is going to always be to allocate in a way that is value maximizing and often and certainly in the simplified examples we're talking about today we're going to be thinking

about that as the highest bidder but as you're going out and building this in the field you want to think critically about like what you know what drives value in your system and that will shape your auction design down to the footnotes I promise we're actually talking about auctions now cool there are lots of them why do we use them we use them because value Discovery is hard and so you see them in context like that like you know Fine Art antiques classic board games got to give a shout out to board games my dad collects them like this is why like you know for unusual Goods we use auctions because it's really hard to know who values them or even what they might be worth and so if you have an auction people can show up cast their bids and now you learn some

of that information resources same thing Wireless Spectrum oil drilling rights uh financial instruments and then in web 3 we have all of these different value Discovery problems and resource allocation problems we have block space we have token sales both you know sort of initial you know sort of like you know you know sort of initial rounds and later tranches like service procurement right governments run a lot of auctions to procure Services um Dows do you too right you gotta like find the coder who's going to bid you know sort of uh you know you collect bids from coders to like write some piece of your architecture reminder that's another case where like you might not just want to go to the you know highest or in this case probably

lowest bidder who's bidding the lowest price uh to code the thing because the absolute lowest price bidder might also just not be a very good coder or might not like be planning to do a good job and so you have to like think about waiting the bids you know the the cost of providing the service you know of procuring the service rather with the value that you get and economically like what do auctions achieve you know fundamentally they're about price Discovery and and as I've said like you know efficient allocation this is like you know getting the asset to whomever values it the most or more broadly creating the socially optimal allocation uh price Discovery is often a useful purpose unto itself right sometimes a

lot of the value of an auction in fact there's a there's a great book called auctions the social construction of value which argues that one of the purposes of auctions is actually to form a sense of value of what of an object like even understand like what Society values it at and like sorry to pull another nft example that happens sometimes right like these are assets where it's really hard to know what it is and the process of even forming that initial value comes in part from a social engagement and consensus around the initial price Discovery process so part of the point is to you know sort of find price Discovery match supply and demand and sort of discover how much something is valued and then the other of course is this allocation question

how do you get it to the people who value it the most another bonus that is especially going to be true in on chain contexts often is transparency right it's not just that like you found a price that clears the market it's not the case you found the efficient allocation but you did it in a way that people can interpret and understand and often audit and that's valuable too right like if you happen to know that somebody is the highest value user of your server and you just give them the server you know I don't know like you know some third parties watching this like hey how come I didn't get to like check out this server like maybe I wanted to use it like the auction not isn't just about establishing value but it's establishing sort of consensus about value in this

sense it's very much like you know even off-chain auctions are a little bit like on-chain activity so that's the positive case for auctions right that's why that's that's sort of like why we like to use them there's this other question which is like why you have to solve allocation problems in the first place right like you know couldn't couldn't we just do something simpler and um Tim rough garden and I wrote a piece on this for for a16 in crypto um the argument is in effect markets you know it's a you've probably heard this like invisible hand right like markets like to find the you know sort of efficient outcomes they they like to clear they like to match supply and demand and so Market forces are

constantly pressing in that direction and in particular if you don't make it to market clearing you know there's some value that could have been created is not right like sort of almost definitionally like you know there's there's if you don't reach the socially optimal allocation you've missed value if you don't reach the market allocation which again is not always the socially optimal allocation but if you don't reach the market allocation there's somebody who would compete to like you know get the good at some you know at the the prevailing price or something and this is why we get gas Wars and nft mints right if you're intentionally offering a good at below the market clearing price people are going to compete for the opportunity to purchase and at least on ethereum the way you do

that is by bidding higher and higher gas prices it sort of just converts into price Discovery and gas rather than price Discovery in the initial sale and similarly you know if you you know another mechanism people use is they have an allow list right they have you like you know do some activity in some Community Building or whatever uh but again now people compete on that activity footnote sometimes you can direct that activity in a way that's virtuous right you'd much rather have people competing to like you know write cool content for your your Discord then competing to like you know sort of just bidding higher gas which is essentially just like sort of like melted Into The Ether pun not originally intended wow

that was totally not supposed to be an ethereum pun but it worked um oh come on no one's laughing at My Accidental ethereum pun okay I got a couple smiles we'll go with a couple smiles it's the afternoon um and so like you know the other reason to you know sort of be thinking about you know mechanism design for allocation is that in some sense mechanism design is necessary right like the market is going to try and get you to you know a market clearing outcome and the question is whether you help it there and sort of control the mechanism by which it happens or let the market sort of just go and do it which can happen to all sorts of strange ways all right we already talked about this there can still definitely be reasons for selling it below market

clearing prices or similarly not allocating the good to the highest bidder in the auction uh Muhammad Akbar and coder towards Jack and I have a have a paper about this and there are many others um happy to talk about this online again I want to emphasize we're sort of this is sort of like you know day one of auction Theory this is again something one could teach entire classes on and and certainly one of the things many of you will be thinking about when you build auctions or other allocation mechanisms is how do you balance you know sort of participation in your ecosystem right like so that you don't have a single dominant winner but rather that you have like a broad base um cool all right

so auction design goals first of all but the two we already highlighted efficiency and revenue right like you know you you want you often want the out you you almost certainly want your auction to find the socially optimal allocation and you would ideally like you know sort of collect Revenue while doing it we already talked about transparency we've actually already talked about I hadn't originally planned on putting the earlier part of the talk but we've also already talked about this like fairness equity and like winter diversity point that like sometimes you care to design the auction in a way that you know sort of produces a more Equitable or uh or diverse outcome uh another thing that comes up a lot is bitter privacy so and and this will you know Michael

will talk a lot about this right in on chain context some types of auctions are harder to run in a private manner um also the type of auction you run affects what people learn right so if you run an auction like in a we'll see in a second there's there's sort of an ascending auction Paradigm and a descending auction Paradigm in an ascending auction people like announce bids you know so this is like what you've often see in like art auctions and things like people like announce bids and it keeps climbing until the last bitter you know sort of like you know second last bidder drops out in a descending auction the price just ticks down until somebody like puts up their hand and says I want it at that price and note that these two auctions reveal totally different types of information

right so the first auction you learn something about everybody's values because you learn when everyone drops out right if somebody drops out at like 70 you know you know then then you learn that their value is seventh year in that range and lower than somebody drops out at 90 drops at 110 whereas in this descending auction all you'll learn is a price at which one person is willing to buy and so even just the fundamental design you choose before you think about like how it's implemented how it's secured even the design affects like some degree of like how much information is shared through the auction and how much privacy the bidders maintain so those are all so this this first auction design goal slide these are all properties of the auction itself

you know how much revenue does it generate you know as a function of the bids how transparent is it and so forth second set of design goals are about what I think of as the metagame so and these are very very critical in web3 um so first of all there's like are people gonna collude or like print many copies of themselves or might the auctioneer print a copy of themselves and and bid in the auction so collusion proofness and civil resistance like if you know for example if your your allocation scheme involves uh you know if your allocation scheme

involves giving out tokens to the first hundred people who sign up for something now suddenly you're gonna be you're lots of people have Bots with like thousands of attempts to sign up or something like that right um you know that's your civil problem uh civil copying themselves many times collusion proofness by contrast some auctions you know you know bidders can benefit by you know withholding bids right like you know a group of bidders can get together and sort of agree to keep their bids low and and share the Surplus among themselves after the fact um credibility um beautiful paper by Muhammad akbarpur and Shang Woo Lee in econometrica and Michael will talk a bunch about this you also want to make sure that the auctioneer is not going to lie about how the auction was run right

are the rules of the auction sort of secure uh and and going on chain can help with this a lot and then lastly and this is I think you know it ties very nicely into the the discussion from this morning uh what we call ex-anti incentives upfront incentives um remember resource allocation doesn't just happen in a vacuum right like you know most of you are not going to be like allocating a token or something and just then that's like forgetting about it right your your research or like you know your resource allocation problem is either ongoing over time or it's a one-time thing but it's about shaping an entire ecosystem right it's about determining who your ecosystem participants are and so forth and so you want whatever that allocation process is

to incentivize people to take the types of actions that you want them to take right you want the eventual outcome of the auction and excuse me likely also even the process of the auction to reinforce you know complementary Investments you know by users um participation by the right set of bidders and all of these are like these are other sort of things you can design for as well okay so last part of the theory discussion is going to be just a little bit of like you know I don't know a couple reps with some standard auction design patterns uh let's actually start in the middle of this slide because I already mentioned ascending versus descending auctions so as I said there's a paradigm you're

probably you've probably seen both of these quite frequently actually so like there's a paradigm of you know bidding up you know where where people like you know sort of bitters like you know sort of you know submit bids over time um or you know another way to implement the same thing is basically have everyone say okay I'm in and then just the price sort of like ticks up and then people drop out as they're as they're now out right as it passed their like maximum willingness to bid that's an ascending auction uh by contrast a descending auction oh ascending by the way also often called English auctions and descending auctions are also often called Dutch auctions confusingly the thing that is called there there are two different things in commonly referred to as Dutch options in crypto that we're

going to use on that we're going to talk about on the next slide this is why I always Advocate you know naming auctions by describing auctions by how they actually run um because that's easier than like trying to guess which of like five different versions of the descending auction Paradigm or you're talking about but um a descending auction the price starts high and then descends until like somebody says you know I'm in or for multiple units right it keeps going until you know sort of the last person says I'm in open versus sealed bid so this is the question is you know sort of are bids public as they happen or are they submitted privately and like only revealed at the end or possibly not at all and you know for example like you know with a seal bit auction you can maybe

even hide how many participants there are who the participants are and so forth with an open auction at least if it's happening in a single room everyone can see how many participants there are because they're in the room uh it's happening on chain maybe they still don't know quite how many they are especially if people like you know might submit bids from different addresses but at least they see more about the level of activity in the auction they know how many bidders you know so how many bids have been submitted and like how quickly they're ascending um or or you know in uh or in the case of descending Yours by contrast actually in the case of descending sealed bid you actually like might know more about how many bidders there are you know sort of a um you know like a first price sealed

bid auction is telling you more you know hi this is just everyone submits a bid the highest bidder wins and pays their bid versus uh you know a descending auction where the price just keeps ticking down until somebody says it's me there could have been a thousand people in the room there could have been 20 000 people in the room there could have been five people in the room you only see that one bid and so again very similar architectures actually have different privacy properties and different amounts of information Revelation um I haven't claimed that they have different allocations because at least in the context we've been talking about so far the allocations are very similar single object ascending descending should get you to the you know open versus sealed bid all of these in an ideal setting will get you to the out

the good going to the you know bidder who values it the most uh bitter who bids the highest rather um and then lastly and this is um this is for you know relevant in multi-unit auctions there's this question of clearing price versus Pay As bid um and so the question here is do all the winning bidders pay the same price whatever it is or do they pay the amounts they specifically bid right so if I'm selling you know 10 tokens and I have a you know I have an auction where everyone submits you know sealed bid

10 token auction first price we'll say sorry no no first price it was absorbed into this I have a sealed bid auction with 10 tokens one option is I give the tokens to the 10 highest bidders and each of them just pays whatever they bid another option is I give the tokens to the 10 highest bidders and all of them pay the 11th highest bid and why is that why would they pay the 11th highest bid well that's the bid just below the threshold rate if any one of them weren't in the bidding you know or or a bit you know bid below that then the 11th bidder would have won and they wouldn't have and so it's the cutoff at which if you'd lowered your bid a little bit below that you suddenly are now no longer a winning

bidder these are very different right in in one state of the world like you know you you collect 10 times whatever the 11th highest bid was and then the other state you know you just save the world you collect strictly more than that right because all 10 of those higher bids were higher than the 11th and there could be a lot of dispersion right maybe somebody bit a massive amount okay this is our like you know these are sort of like again and again like you know for larger more complex auctions you often mix and match these things there are a lot of um a lot of procurement auctions sort of have like a sealed bid round followed by an open

round or vice versa where they're sort of like uh you know like initial proposal and then you know sort of some sort of you know sort of you know direct bidding competition among the top proposals um that's also a good way to deal with bitter diversity as a side note um but so these are these are like you know the standard like your standard three by three design patterns for for auction structure I have a footnote on this slide professors love footnotes fun fact um that there are lots of other design levers we haven't talked about here so like there are Reserve prices sort of like lower bounds on the bit you know on the the winning bids bidding language matters right if you're going to have a bid with very complex Goods then maybe you want to like give a structured

bidding language that people can bid you also might want to sell packages together Wireless spectrum is often sold in like chunks that are you know like you know if you're building a telecom Network it's not useful to only have your Telecom Network work on the east coast of the United States right and so like spectrum is sold in like you know some large National packages but also you know some bidders are local right you might have like a local television station for which you only need a local you know package and so there are auction designs that aggregate small like allow sort of joint bidding on large groups like big bundles and small bundles at once and sort of like resolve the the differences in bidding between the two um so lots of other design opportunities here for for more complex problems

um so I'm gonna give you two different auctions I'm going to describe the rules they're right here and then what we're going to ask is like what's the difference like how does it you know how does bidding Behavior differ and how does like how do the outcomes differ so first one is the descending pay is bid these are these two different sort of things that are called Dutch auctions in crypto land that that we were talking about so this is I'm selling 10 units the price starts at some high number in declines and whenever somebody wants they can like you know claim a unit at the currently posted price so it keeps going until all 10 units are sold and everyone pays whatever price they jumped in at and let's assume that people want to buy at most one of the units so so it's it's

you know a license to you know it's a membership uh nft in some Community or something you only need one of them um the other one is a descending clearing price auction so they're selling 10 units the price starts high in declines whatever someone wants they can claim a unit and at the end all the winners pay the price at which the 10th unit was claimed so it goes on until like the last unit is claimed and then everyone pays you know and then those those 10 bidders the 10 highest bidders all get the unit but they pay the 10th price so two questions we're framing here it's like what's the difference I guess it's really two and a half what's the difference in terms of bidding Behavior so like how would you bids you're like think strategically like I'm going into

this auction what am I doing in these two different contexts and then two once you've thought about how the bidders are going to bid it's like what's going to happen do we get the goods to the the bidders that value them the most do we think like how fair is what happened at the end of the day uh and you know how happy are we with the outcome I haven't got the question it's like a many part question but I think it'll be fun to think about all right go for it think to yourself for 20 seconds turn to not your co-founder and discuss for two or three minutes and then we'll come back bidding Behavior how are people going to bid in descending pay as bid versus descending clearing price for 10 units um I'm gonna try

based on our conversations uh depicts you know what we talked about so on the example on the right pretty much the price Setter is like the one who's the cheapest wants to give the least um and I think like generally if you're like a buyer a consumer um on that side of the market I think it's more Equitable for you because um we will assume like there's a variation among the bids right there's like a high bid a little bit if that assumption is broken um and there is no Cheapskate like everyone is a whale then like there there isn't really enough Discovery for

the market to find like a low price right um that's the caveat on the left it's um the price Discovery is is more active but I'd say like the seller has a little bit more of an advantage here um if you're like a cheap buyer um you might not even get get that opportunity to bid whereas opposed like you know in the other example like you can bid a lower price um and then you know kind of end up spreading that price across everybody so it does sound more Equitable for the buyer on the right and then seller on the left generally but yeah okay I love this Equity point so

the fir you know the uh the the descending thank you so much by the way um the uh uh also we have you can't sell it you're particularly telling we also have matching hair which I'm super stoked about um the uh so yeah so this Equity point is very much real so in the descending pay is Bid auction right let's let's imagine we've got 10 units for sale we've got five whales in the market and they've been some really really high price well you know he even as the price ticks down like at the end of the day like you have five people who paid way much more than everybody else in the descending clearing price auction

if they bid some really really high price then you know the price ticks down at least at the end everyone pays the same thing that somehow feels very Equitable right like even if you were someone who was willing you know really really high bid um you know you got to pay the same amount as everybody else you know the price realized lower and like you didn't lose out from bidding High and Meanwhile by the way right on this other auction like you know you can lose out by bidding high and who is it who loses out it's like the ones who most value the thing right it's like the people are most excited about the community are the ones who sort of like most in in equilibrium you know sort of overpay in in some sense which like you know maybe is adoptable it's it has economic you know rationale

to it right it's like they're the people who care the most about getting it they will have the least risk of whether they managed to get it but also yeah you know it sounds a little it feels a little quick all right what else have we noticed other thoughts like what other sort of observations did you guys make about these auctions hi I'm natia um on the right side auction there's more incentive for a bidder to express their true value preference um um so because on the left side you want to be the 10th bidder um whereas on the right side it doesn't matter as long as you're in the first ten absolutely nailed it so so combining so these two comments combine perfectly right so like you know what's going on you know like what's what's going on

with this Equity point it's like in the descending clearing price auction because there's this equity and because you're protected right if you've bid way too much sort of by accident you know you just happen to be the guy you're the person with a much higher value than everybody else you're protected right like the auction realizes a price that's much lower whereas in descending pay is Bid you are explicitly trying to be that 10th bidder this by the way is why a lot of those like descending pay as bid auctions actually have a gas War at the very end right like everyone is sort of like waiting for each other to like jump in and they're waiting to try and be the person who like bids at the final realized price and then there's this you know gas war that actually spikes the effective price much higher than it was

you know 20 minutes earlier or something that's like not an accident right people are like explicitly competing to be the lowest the lowest bidder whereas in a descending clearing price auction unless you are the lowest bidder um you know the lowest bidder has a little bit of impact they set the price but like most of the bidders and if you think there are a lot of bitters with like a lot of you know sort of valuation dispersion you don't really ever expect to be the lowest bidder going in most of the bidders are not setting price they're just expressing their value and exactly like you know they're they can express their true value because they know that like you know any price below if if they bid at their true value any price below that they're going to be happy with cool any other observations in terms of

uh Game Theory there's this trade-off between price and guarantee of getting a unit and it's possible so that you know some bidder can get multiple units and then as you noticed uh like for the first auction uh there's an incentive for the bidding to speed up towards the end and so like if you copied a 10th person maybe you want to be the ninth person so there might be some example where you know people are incentivized a bit sooner and shooter and so you know the ending of the the bidding can happen very quickly absolutely and you know that implies that the descending pay is Bid auction is just much harder for a bidders to figure out how to bid in right like the equilibrium and it's also a little more random right like you can

sort of like coordinate on the wrong equilibrium like you're sort of just guessing when you know if you're trying to be not the tent you know that if you you want to be the 10th but oh maybe maybe I'm only gonna be able to be the ninth you're still sort of like guessing where everybody else is much harder to like you know and much more sort of stochastic in terms of the outcome totally and if there's not that many bitters there's an incentive to collude uh like for the first absolutely and then for the second one like if there's not like 10 bidders then you can just wait until the price is much lower and you have like a much lower clearing price for everybody which is could be unfair but of course you can say it's more Equitable since everyone's equal absolutely yeah there's like a lot of

both of these are actually like very subject to you know active bid you know bid reduction but the the pay is Bid much more right like if everyone can sort of you know in both cases you're if you can somehow coordinate to get the 10th bid be very very low you're much happier um but like descending pay like in if you only coordinate like 50 you know five bids that doesn't do nearly as much into sending clearing prices a dozen pay as bid especially if people like are learning over time where the other bidders are uh or appear to be yeah okay sweet so so we've got a lot of argument in favor of this descending clearing price auction and I will I will tell you you know cards on the table I am a descending clearing price auction fan uh so I strongly strongly you know

sort of recommend them in a lot of contexts but let's just why do that like why is this descending pay as bid format exist at all right like it's actually like a much more common auction in in many crypto contexts why is it why do we do it why does anyone do it um the the bid amount is more predictable like you you get what you expect to pay yeah if you're doing things based on that bid like in the context of of ethereum you might have a trade on coinbase um based on the price you're paying uh so you can now execute that other trade

uh knowing what you're paying on the other side whereas with the other one you have to wait until the like the auction concludes or something and totally possibly get yeah the larger Bill than you expected depending on the conditions yeah so so descending clearing price is never going to end up charging you more than you than than your bid but there's still like you know in a pay as bid format you know like you know with certainty what you are paying in the event that you win and that is actually often very useful and it's easier for a lot of bidders to understand too right like we found in like sort of like large-scale auction experiments that like bidders get confused with clearing price you know the unsophisticated you know sort of

inexperienced bidders not unsophisticated in the sense of like they're not like intelligent or something but just they're not like sophisticated with respect to auctions they don't have a lot of experience and and sort of training and how to bid they often get confused in clearing price auctions or second price auctions in different forms pay as bid is very intuitive uh and and also it's a little bit harder for the auctioneer to manipulate is anyone here ever like tried to implement oh you've got it I was just gonna say this is far better for the seller and the seller often designs the auction so you know yeah so so one so why hang on why is it better for the seller uh because the prices are all like differentiated they're higher uh potentially at what you know the highest someone would be willing to pay

for something whereas you know to sending clearing price often they're not paying at the willingness to pay they're paying at somebody else's 10th person is willing to pay and so there's all this loss uh you know sales price it's profit yeah so again one has to think about in equilibrium right like you know because there is this like you know sort of funny like equilibrium coordination problem but like yeah at least in principle if you have some of these whales people who really want the good are just going to bid the high price no matter what descending pays bid auctions can result in much higher Revenue in some circumstances right like you know I've certainly been in auctions where there were like five bids in the first 20 seconds and then all the others came in four hours later and like you know

the auction format again it's not quite Apples to Apples because you wouldn't necessarily see the same set of bids but the auction format is the difference between you know sort of those top five bids coming in at like you know I don't know two eth or whatever and like you know maybe like 0.05 yeah go for it yeah I just wanted to try to get like an example of descending clearing price um just just to kind of etch out that example you're talking about because it seems like the Surplus on passbid goes to the seller right like the majority of the Surplus but in the rights we're kind of like saying like all right the seller doesn't need you know all that Surplus like keep it yourself um yeah maybe you spend it on something cooler or I don't know but like what

what is a real life I I just have trouble like thinking of like has it been used does it is it the best method for certain things absolutely great question um okay first of all so the Surplus comparison is not totally straightforward again because bidding you know bidding adjusts to the auction format right so bidding Behavior might like leave like like lead to like different sort of total Surplus flowing to the to the auctioneer in both cases but but I agree basically that like this is you know you know like like yes right like you know we've been talking a lot and and the intuition is that descending pay is Bid is like often gonna like result in a lot more Surplus going to the auctioneer especially if there's like a lot of stratification in people's values

um you know when might you want to use descending clearing price um first of all like you might want to use it simply for the fairness property um right and and sort of the the true price Discovery right like a descending pay is Bid like because of all these funny Dynamics you don't actually really quite discover the price right you sort of discover some you know iteration of people's like various bidding um descending clearing price like uh you know for example um tally Labs uh used a descending clearing price auction for the you know for the Jenkins the valet board and dangerous public sale and you know the view was twofold right like again I I don't sorry

I don't want to speak for the team but but my interpretation of their sort of public explanations of like why they thought this was a good idea was one of them was that just they thought it was fair right like it's incumbent on the auctioneer to create something of value that generates Surplus and like you know you you don't want to like take your you know your highest bidders who are likely the people who are like the most dedicated to your community the most excited and extract more from them right so that was part of the argument and the other argument was that it was simpler on two Dimensions right one of them was it simplifies the bidding process right like people could just show up and bid whatever their value is and then forget about the auction for the rest of the day they don't have to sit there for

like six hours watching the clock price descend and secondarily you don't have this like coordination problem at the end with a with a gas war or the potential gas War where everyone just like wants to jump in sort of at the same moment to be the 10th highest bidder um and so even in a context where a lot of sales are descending pay as bid they chose a descending clearing price because of what they thought of as virtuous properties for descending clearing price um at high level like descending clearing price auctions are often really good when goods are homogeneous right when the all the goods are sort of the same at the end but there might be dispersion in value um there's a like strong fairness argument that you should give everyone the same price if they're buying the

same good in the end um so like you know random allocation mints are a good example of that or or sales you know sales of the same object um I've actually personally been a fan of hybrid formats where like you know if you're going to do something that's pay as bid you actually like tie the the quality of the outcome unit to like the price paid or something like that but it's all much harder to implement um okay sorry I'm nine first of all hopefully that was a helpful answer I'm happy to talk more about this offline uh as you guys can have probably observed I'm totally obsessed with these things they're they're fun one other point which interestingly like you know possibly like sort of like you know I did this mistake of that puzzle of like framing you know of you know I kept us

too far in the frame but another reason descending pays bid is popular is just because it's much easier to implement on chain right descending clearing price auctions you have to figure out a way to do the refund right if people you know the only way you can like certify their bid is if they lock that amount of eth or whatever in the contract and then you have to figure out a way to give it back similarly like you know in in major Spectrum auctions right are they gonna like escrow like you know Millions hundreds of millions of dollars or whatever you know to bid in this auction or or as pay as bid just like you know you you bid and you're in effect done um and so another reason that you might find yourself in a pay as bid format is for design and implementation reasons

which is you know where where we're going with Michael next but okay so uh quick quick uh you know wrap-up slide you know auction design is about incentives and it shapes behavior and outcomes uh and this means you have to think really carefully about how they work in theory right we just like we literally probably could have gone on for another like half hour just on this question of how these two different auction designs affect like behavior and outcomes and incentives right like you know that's the level of Game Theory one has to go through to really like rock how these things are going to affect you know what happens and they can make or break your ecosystem right a bad allocation mechanism means the wrong people are getting you know access to

resources which means the wrong people are like you know incentivized to invest and like you know the whole thing collapses but all of this is in sort of reflection or refraction of how they work in practice and and as we'll see in Michael's uh you know component next um there's a lot of differences and constraints and also opportunities that on-chain auctions in particular offer QED [Applause] yeah so you guys thought this week is going to be a about business but we thought we'd uh throw in some engineering content and as well just to keep everyone on their toes so anyways uh like Scott said I'm going to be talking about auctions in practice

looking specifically at what happens when you try to implement an auction uh on chain so first things first uh this is my working definition of an on-chain auction uh so this is an auction that's got at least uh part of its components implemented as smart contracts on a blockchain and in particular this definition allows for certain parts of the system to be off chain and we'll see some examples of this later right so now that Scott spread the good word about auctions and all that they can do you might all be eager to implement an auction of your own

uh however it isn't quite as simple as just like taking your favorite auction Theory paper and translating it into solidity there are certain aspects of on-chain programming that make auctions uniquely interesting as well as uniquely challenging to implement and that's what I'll be talking about here today we're going to start with the good news what are the advantages of on-chain auctions compared to their offshade counterparts and then we'll move on to the bad news so what uh challenges specific to the on-chain setting uh might you encounter when you when you write an auction

uh and finally we're gonna explore some common design patterns that you might encounter in in the wild so why on chain auctions I think there are basically two sides to this question uh first what value do auctions bring to blockchain and this is kind of what Scott talked about in his portion uh in in great detail but to summarize auctions are tools for resource allocation and price discovery and as it turns out there's a there's a lot of resources on-chain resources that need allocating as well as on-chain prices that need discovering so we've seen that auctions have become an

extremely important primitive in the on-chain ecosystem the flip side is uh what value do blockchains bring to auctions and I'm sure a lot of you have heard or asked questions of this form yourselves like what value does blockchain bring to X but I think the answer in the case of auctions is actually uniquely clear and incontrovertible and the answer is reliable broadcast and credibility and do you understand what these two terms mean we're going to work through an example here so one of the classic auction formats from from theory is the sealed bid

second price auction this is also known as the victory auction so in this format uh bidders privately submit their bids to the auctioneer you can think of these as being in a sealed envelope so that no one else can can see another bidder's bid and so here in our example we have three bidders bidding again 15 75 and 20 respectively and they're sending them privately to the auctioneer and if the auction proceeds as intended then we then the auctioneer is going to read all those bids and report that this second bidder is going to win the auction

um since they had the highest bid but instead of paying 75 they're going to pay 20 which is the second highest bid um so this might seem a little counterintuitive at first but it turns out there are good reasons for doing this specifically related to aligning the bidders incentives and we'll talk a little bit more about what that means in a bit but of course all this assumes that the auctioneer is behaving honestly so what happens when the auctioneer isn't honest here we have the the same setup we have the same bidders uh same bids but the auctioneer is going to be malicious so once again the bids are submitted privately from bidders to the

auctioneer and now uh once the auctioneer has received all the bids they're going to report again that the second uh bidder won the auction so far so good but instead of reporting the second highest bid as twenty dollars the auctioneer is going to say that actually they received this bid of 74.99 so this means that the winner pays 54.99 more than than they should have uh worse yet like the bidders can't uh tell that the auctioneer is lying without talking to one another and this might not be practical if a if you have a lot of bidders or B you might not know

who all the bidders are or how to contact them uh so yeah obviously this is quite a big problem uh luckily for us this is where this principle of this idea of secure broadcast comes in so let's say that instead of submitting your your bids to the auctioneer bids are instead broadcast on the blockchain um one caveat here is I'm gonna in this presentation I'm gonna gloss over the fact that it's still a sealed bit auction so um bids are gonna have to remain private for the duration of the bidding period and only later revealed publicly it turns out this is possible to do on the blockchain but it's going to require

some cryptography and we're not going to get into the details here um so here the blockchain is serving as this sort of public bulletin board where message uh can't be corrupted or censored which is a really important property this allows bidders to know what everyone else bid once the the auction is over and therefore tell if the auctioneer has been lying about the result by just looking at the current state of the blockchain right so that's already an improvement over our situation before uh where bidders don't have an easy way to detect whether the auctioneer has lied but as you may have guessed already we

can do better by replacing the auctioneer with a smart contract so like Scott was talking about earlier today with um uh contracts as being credible commitments with the rules of the auction written in public open source code you don't really have to worry about the auctioneer and whether or not they're honest you although you do have to worry about bugs and we'll talk about that a little later as well um so pausing here for a bit does any anyone have any questions about this example I have a question about uh how this is solved like in uh not in the blockchain word because sealed second price auctions are a thing

um is does people just trust for example Google as an ad exchange or are they other mechanisms like what are we competing with in terms of uh credibility uh yeah so I think in practice for the most part uh in in like a web 2 setting yeah you're exactly right so um the Bears would just have to trust uh whatever ad platform to be uh honest okay so taking a step back for a bit what we just saw in our second price auction example is actually quite remarkable so um in that credible auctions paper that that Scott alluded to earlier Akbar poor and Lee approved that under the

traditional model it's actually impossible for any auction to have all three of these properties um a credible auction which we've alluded to is one where the auctioneer is incentivized to behave honestly a truthful auction is one where bidders are incentivized to bid their true value for Whatever item is being auctioned so for example if I'm bidding on like a Pokemon card that I value at a hundred dollars all have um no reason to bid anything other than exactly 100 no more no less and finally a static auction is one where the amount of communication between bidders and Auctioneer is

bounded and I think the easiest way to understand what static means is to uh to look at the example of an auction that isn't static so the classic ascending price or English auction is not static because you might have two bidders locked in a bidding war you know uh repeatedly bidding outbidding one another with no end in sight and so that communication is therefore unbounded and not static so under the traditional model the the sealed the second price auction that we saw is static and truthful but not credible because as we saw the

auctioneer might have some incentive to why but by implementing it as a smart contract it that same auction can be static uh truthful and credible one way to think about how that's sort of possible is that there's kind of a division of labor between mechanism design and blockchain going on um so by providing trustless smart contracts and secure broadcasts blockchains make it easier for the auction to be credible and that sort of gives mechanism design some slack to focus on the other two properties static and truthful um so now given the Synergy between

auctions and blockchain you might be wondering like you know why don't we just Implement all options on the blockchain as you all know there are certain challenges you run into when building on chain that you wouldn't run into otherwise and um I have what I see as the four main ones here although this list is by no means supposed to be comprehensive so first gas this is the question of whether it's efficient for both Sellers and bidders to participate in the auction and of course this is uh largely impacted by the chain that you choose to deploy on for example if you're on you deploy to a chain where transaction fees are relatively cheap then gas might not be as much of a concern for you

uh second Mev so this is uh depending on how your auction is implemented value might leak from the auction participants Sellers and bidders to other parties so think about parties that are in the quote unquote Mev supply chain uh Searchers builders proposers as an application developer it's kind of part of your job to make sure that to minimize that as much as possible uh privacy so as it turns out it can be tricky to implement privacy on a public blockchain um and I'll expand expand on this a little bit more in a later slide finally smart contract security uh this

goes without saying but any smart contract that handles users assets introduces a security risk and needs to be handled with the appropriate caution so Scott talked about how design choices can impact the incentives and outcomes of an auction and when you're working on chain you also have to work under these on-chain specific constraints which are going to influence those design choices which in turn influence the incentives and outcomes so just going through those uh those four constraints I had up earlier um if gas costs are are a huge concern if you're on a you know on a chain with high gas fees

uh that might lead you to off-chain bids so instead of uh bids which are on chain transactions you might have off-chain signed messages as the bids which are free for bidders with regards to Mev uh like Scott mentioned earlier you might tend towards using a clearing price descending auction rather than pay as bid and the fact that privacy is hard historically has led to most auctions on chain being open bid formats and finally security concerns might lead you towards simpler designs which have smaller surface area for potential bugs and are easier to audit it's also interesting to note that this

this sort of pattern of on-chain constraints informing design choices and forming incentives and outcomes isn't just unique to auctions we've also seen this with like for example the rise of amm's on chain those were sort of born out of on-chain constraints people realized that on-chain order books were maybe not as effective as they were in traditional finance and that led to the rise of of amms like uniswab right so if you did a survey of all on-chain auction implementations being used in the wild today uh you'd definitely find a quite a bit of diversity but I think most of those implementations would fall into one of

these three broad categories so first we have English or ascending price auctions with on-chain bids and this is what I would say is like the most intuitive implementation of an on-chain auction um so basically to place a bid you submitted on-chain transaction with the uh collateral for that bid and just like in a Christie's or Sotheby's auction the the price gets ratcheted up and up quickly you would have some sort of minimum bidding increment and typically the auction would end if certain uh if some amount of time has passed without any bids this is the type of auction that's used

by maker for liquidations as well as Zora for selling nfts the second format is English auctions with off-chain bids so instead of submitting a transaction to place your bid uh you bidders would sign an option message indicating their bid uh with the seller ultimately submitting an on-chain transaction to basically accept the winning bid and that's when assets exchange hands if you see an ascending price auction on openc this is how it's implemented um the main benefit of this approach is again that bidding is free but of course the downside is that this requires some off-chain infrastructure which introduces some degree of centralization

risk so you know bids could be censored and for whatever reason the seller might choose a bid that's not the highest bid uh things like that third we have the Dutch or descending price auction yet again so to quickly review price starts high and decays as a function of time Dutch options are nice because they tend to minimize the number of on-chain transactions needed without requiring that off-chain infrastructure for example in the single item case the first bid that comes in wins and immediately ends the auction right so whereas if that were implemented as an

on-chain English auction you would potentially have many rounds of bidding each of those bids being a separate on-chain transaction costing gas for those bidders Etc an example of an on-chain Dutch auction is the um is when an ens domain name expires a Dutch auction is automatically kicked off with the price starting at I believe a hundred thousand dollars and decaying exponentially from there the winner of that auction wins the uh gets the ens name and the proceeds of the auction go to the ens Dow um right so I would say the main downside of the Dutch auction is that if you want to bid at a specific price you

have to be online at a specific time and to submit the transaction and that might be like 4 am in your local time zone uh and so that could be inconvenient unless you have a bot set up to do that on your behalf and that also requires some extra steps uh yeah and as as Scott said uh with Dutch auctions there there are some that are implemented as pay as bid and others implemented as clearing price all right so with these sort of blueprints in mind uh I have a little Pop Quiz so whenever you're writing a smart contract you should be thinking adversarially because if you're not the one thinking about what could go wrong

what could be attacked inevitably someone else is going to be doing the thinking for you so I've got uh this hypothetical auction implementation here so your job is to figure out how Mev would change the intended Dynamics of this auction so we have this English auction on-chain bids and the auction ends after 24 hours and that's a hard deadline which is not extended even if a bid comes in at the very last minute um right so what might go wrong here and again uh 30 seconds to think about it for yourself and then uh find a neighbor to discuss with all right anyone have an answer

uh if you want your uh transaction to go through uh you can buy up all of the remaining block space between when you place your bid and when the auction ends yeah that's true that would be potentially expensive to to pull off that attack but that yeah that is that is very true anyone else have any ideas about um the the effect of Mev On the Origin Dynamics uh it was incentive to wait till the very end and then snipe at the end of the auction yeah yeah exactly um right so so like both of you uh notice there's the matter of like uh controlling block space or influencing

um transaction ordering being very important to the outcome of this auction uh the answer that I put down I just said last block chaos basically a lot of these more sophisticated bidders might want to wait around until the very last block to decide whether they're going to bid or not um depending on what the price is going into that block uh granted you know that might be that strategy might be authoritative if someone manages to you know buy up uh some of those blocks at the end then with all these people trying to squeeze into the same block you know gas prices are going to go up you might have bidders trying to like dos each other even and the the block builder for that

very last block is going to have an outsized role in the outcome of the auction which we we don't want again that that's leaking um value from the auction participants to the Mev supply chain um and then again to return to the Dutch auction again in in the single item Dutch auction case uh pay his bid and clearing price is is actually the same if you think about it right because there's only one winner and that person is going to pay what they bid but they're also the you know setting the clearing price um and you know for in the interest of time we're not going to go through uh I'll just you know we've already talked

about this a bit um kind of similar to the last scenario there's this game of chicken um people are going to try to take advantage of that Gap in time between when the bid is broadcast in the mempool and when it's included in the chain so you know everyone's going to see that first bid in the mempool and have some time to react um right so if you see that settle what that settled price is going to be and are willing to pay a little bit more what you could try to do is front run the transaction so that way you're paying the same nominal bid price the the price that the contract is aware of but you just pay a little extra for gas so if you have multiple people behaving this way

that's going to trigger a gas War with people you know broadcasting bids at incrementally higher gas prices and yeah like Scott said basically what was intended to be a descending price auction becomes an Us ending price auction uh for transaction ordering where uh the auction proceeds are being captured by the Mev supply chain since it's kind of happening out of band instead of in in the smart contract so yeah again leaking value from the seller of the auction now before we end I wanted to highlight two areas of on-chain auction design that I think are currently underexplored

so first is privacy sealed bid auctions are well studied in both in auction Theory and actually cryptography however they can be kind of tricky to implement on chain so it turns out it's hard to hide the value of that's being used to collateralize it it's easy enough to hide to have encrypted data in in the transaction but for most on-chain assets it's uh you can't really encrypt or hide the the value that's being transferred that said it's not entirely impossible and I've written three solidity implementations of sealed at auctions uh each one using a slightly a completely

different method of hiding those bids and those write-ups are on the a16z crypto website if in case you're interested the second area I wanted to highlight is cross chain auctions this I Loosely Define as any auction where there are on-chain components on two or more blockchains and for example you might have an auction item living on one chain but bitters funds disperse across multiple others and if you believe in you know our glorious multi-chain future then this is going to become more and more common alternatively you might have uh an auction where bids are placed on one chain say an L2 that has cheap gas fees

or some sort of privacy features but the assets live on some other chain maybe one that has you know stronger security guarantees um and I'm sure you can think of other plausible blueprints for uh cross chain auctions and I think as cross chain infrastructure continues to mature this will be an interesting space to both uh watch and and build in yourselves uh that's yeah that's it thanks for listening if you have any questions please find a mic uh I think we're going to bring Scott back on on stage um here we are we got the full band hi thank you for your uh for your talk

from both of you and I gotta save up your enthusiasm is very infectious uh when it comes to talk about auctions so uh Jay from Ethan ID I was curious uh so our our project actually allocates a namespace of what might be the um the latest uh uh best practices and in literature when it comes to auctioning off domain names it's a great question yeah I guess Scott if you want to take this from the auction Theory perspective and I can try to answer from yeah so the couple interesting observations about domain names one is that the markets for most of them are very thin right there's sort of a

small number of and you should tell me if I'm getting any of this wrong because you actually know the domain name business and I'm just you know extrapolating based on how I think this Market works so correct me across the board if any of this sounds incorrect but my instinct is that domain names the market for almost everything is super thin and then there's a small number of domain names for which demand is really really high it's like very competitive to own I don't know like you know x dot XYZ or something like you know x dot YZ maybe um and then secondarily there's this issue that you like worry about domains you you sort of worry about domains going

um you know sort of going sold and then underutilized or something of the sort right like you know you sort of don't want you you know a bad state of the world is if people are basically just buying domains to spend this by the way is not limited domains it's true for many token categories right if you are just buying it to speculate on future users wanting to buy it later and then they don't develop anything on those domains and so the domains don't actually like realize their true value are those both is that first of all is that like just a good like ridiculously simplified description of the market oh that's a that's absolutely accurate like the the real goal is to like assign the domains to the right people to maximize the utility of the platform awesome okay so one thing I would think

about doing is have the allocation feature some sort of clawback so some mechanism by which um and again you know I don't know the exact like sort of smart Contracting implementation you're using but you could imagine this living in a contract that basically if the domain is not utilized you have to figure out how you verify that it's not utilized maybe this is like a massive Community you know like a massive governance Vote or something or maybe it's just that like you know no access is over some time period or something like that like if it's not utilized then maybe like you know it disappears or they have to pay a maintenance fee ongoing to like maintain it or something of the sort um it's also a context where you might want bidding we talked about the sort of like bitter diversity type questions at the

beginning you might want bidding in units that are not only cash right or or cryptocurrency or whatever like the you might have bidding in like you know proposals for what they're actually going to do with the domain and like do some amount of you know sort of personalized pricing as a function of like who the bidder is and how valuable their value is their proposition for the ecosystem is or something of the sort you can't do that at the scale of like you know millions of domains necessarily but you can certain you could do that at the scale of hundreds or even thousands of domains um you know and and particularly for the ones that have a lot of competition like that's you know you know it might be that like if bidding for a given domain is sufficiently competitive it triggers one of like a a second stage where

people actually have to compete on proposed quality or something like that yeah and then the other thing I guess and this is harder actually Michael would be super interested if you have any instincts for how you might do this is you might want you might want some mechanism to try and it's very hard on chain because like you you have like you know people who can civil themselves easily but you might want some mechanism for trying to limit the number of domains that any individual entity can acquire um you know and in like you know physical world auctions like this is often done like through a registration process but I'm not sure how you would do it in an Unchained manner without kycing or something like funny thoughts yeah I think civil would be hard to uh overcome it and

in the launching setting like that I'd be curious Scott to get your your thoughts on like with regards to the Callback idea something like a harburger tax on on yeah possibly yeah um I don't know if you're familiar with I'm actually not uh that concept it's been thrown around in in some like recent conferences um but basically and Scott correct me if I'm wrong but if I'm if this description is incorrect but like basically the the item that you hold is sort of always for sale but you just determine what that for sale price is and you pay an ongoing tax as a function of of that sale price um right so if you really want to hold

on to that domain name you'll set a high sale price but you would have to pay a higher ongoing interest uh interest rate yeah granted and the challenge with Herbert the exact you know perfect description um totally nailed it the challenge with Herberger taxes of course is that they're very sensitive to this problem that like some people might just have like more ability to pay than others right if you've like managed to allocate the domains to a bunch of users who are really high value creation but don't have a lot of cash on hand to like pay these large taxes they're sort of like forced to set low prices then you worry about people Sweeping in and like just stealing the item from them or even just like pranking them right like you know some you know some competitors comes in like pranks them buys the domain for an

hour and then sells it back later like so so you have to be careful about the implementation but there are versions of this where basically that are that are similar that are the sort of like a you know a harburger tax with guard rails where in effect you know you're you're paying an ongoing fee that's proportional to your realized demand for the for the asset um and you know maybe there are like you know maybe that fee happened you know every time you pay the fee you're sort of like you know you're you have you're you're locked and guaranteed to have the thing for a year or something right something that's basically like a higher you know or it can vary as a function of the community's estimate of how much value it was actually accruing to that domain um something that like links the fee you

know sort of like an ongoing maintenance fee with this idea of like how much the realized value there is um is helpful um oh one footnote on this sorry I know we're giving a really long answer but this is a really fascinating question one footnote is of course anything that involves having the thing be like sort of constantly up for sale right another like intuitive way to like just make sure the thing always ends up where it's most valuable is you just have like a very liquid Market in these valuable domains right sort of have them like you know sort of make it really easy for people to sell and transfer them but you have to like be careful when you set up things like that because that of course also makes it easier for speculators to unload them later right and so like you know

the answer is not necessarily maximum liquidity it's some mixture of liquidity plus maintenance fees with if you can some degree of consensus building around who the value you know the the value creating holders actually are and somehow discounting to them or raising the price to those who are not uh thank you for a very detailed answer a lot of things to think about hey Scott my name is Marat I'm the founder of primav primav is a block Builder communication Network we're looking at kind of revealing more information about the block auction process Without Really um you know adding a new auction but obviously what we're doing is going to affect the auction so my question is what if let's say uh in a you have a 10 second auction normally and you divide

that auction into 10 Slots of one seconds what what should we expect the general Behavior to be based on it maybe any literature or you know experience that you can derive from great okay tell me a little bit more about the context so ethereum blocks ethereum blocks to be precise it's 12 seconds right so we're in the process of revealing metadata about the blocks through that 12 seconds to participants of of you know transactions mainly Searchers and uh you know we're going to be running this experiment over this month month but we're curious to see if you know you have any pointers and thinking about oh goodness you're having me lock in a prediction then you're going to run the experiment and we're going to learn whether we got it right I

love it I want to hold you to it but I'm curious about you no no I'd be super interested right like okay and so they and the 10 second to one second thing so it's the same bitters or are they being split across the seconds in some in some meaningful way same better 12 to you know one same bitters okay and the new optionality here is that you can basically decide which of the one second components you're in uh no it's it's rather that um you know you get or you're just speeding it up you're making the whole thing just faster yes okay so the thing that the thing I'd be thinking about here there's a set of bidders for whom nothing is changing sure right if you're

in the business of submitting a transaction in every block like nothing has happened right the blocks are now like you know 12 times faster or something but like you know you're you're still submitting those same transactions with some some funny caveats I'll describe in a moment what's different I think is about the arrival of transactions and how that interacts with fitter's incentives right so if bidders arrive so I don't know what the natural frequency of like transaction arrives I come in that's the 10th second of majority of them come into the 10th second of the 12 second slot yeah yeah okay so so you might see them all starting to come in at the uh you

know you know five sixths Mark of the one second slot right like it's possible that you'll just see targeting of exactly the same style um you know those those are the videos for whom I think nothing has changed right so like the ones who are like you know the ones who are trying to basically like close out the block are probably you know to the extent that their technology will allow them to are probably gonna try and close out the Block in this in more or less the same way um I'm thinking of sort of like what I think of as like retail transactions so transactions that like just arrive at random times like if this is a meaningful difference with the frequency with which those transactions arrive relative to all the other stuff that's just sitting in the mem pool like then you might see some significant

changes right so like if this reduces so for example one thing that could happen is like suppose retail transactions really like this is a ridiculously oversimplified model so it's totally wrong but like suppose retail transactions only allow arrive one every 12 seconds and like everything else just flow transactions of like various D5 protocols like do it you know sort of doing stuff various like you know protocols like you know try to snipe around them like but like you know these retail transactions used to only show up one every 12 seconds they're still only showing up one every 12 seconds now like 11 of your 12 blocks have like nothing in them right you know no retail in them right they're

all just these sort of Maintenance transactions and so like the average block competition maybe goes down right you can imagine that like price Falls you've raised throughput but it's not just falling because you've raised throughput but rather also because the types of transactions that like trigger sort of everything else are like less frequent somehow right there's like more time for like basic maintenance and then like every so often you know you you have this like drop of a retail transaction and maybe now some people are trying to pick off their price quotes and whatever um but honestly my guess is it's my my first order instinct is that it may not change very much other than just clearing more transactions and like and you might get like more of an you know more of a price

drop than than just from like the increase in throughput because you sort of like add heterogeneity to the transactions you could also Imagine a net price increase I guess [Music] I don't know let me take let me take this one offline yeah um I'll I'll send you send me just like a one paragraph description of like what precisely the the change is and I'll try and sort of like lock in a you know a better a better more like precise prediction but like my first knee-jerk instinct is that you might actually see a massive change except in as much as like 12 seconds versus one second is a difference in the

number of sort of like random transactions that show up in a given block okay um well I'm it makes me happy that it kind of had you think a little bit so I'll send you something no I love it this is like so much fun holy cow thank you good thank you guys [Music]