#DappCon22 : Day 1 - Prediction Markets Are Dead Long Live Specialized PM, Rossen Yordanov

Transcript

thank you for being on stage thank you for being here tonight with us today uh but please tell us what happened to the prediction marketplaces and how this is better than what was before welcome Round of Applause thank you so much [Applause] hi guys thank you for coming here my name is rossen and uh for uh a year and about four months I've Been Working Day and Night on the least hyped area of crypto and blockchain which is prediction markets thank you for the great intro so I formed a very strong opinion about the past the present and future of

prediction markets and this is what today's talk is about um so let me just first start with uh well prediction markets are dead as you mentioned this was the first project that noses worked on and that was in 2016 2017 and after that NADA erect uh we'll go through a little bit of history but I'm not the first person to basically Proclaim prediction markets debt uh critical thinker popular crypto personality hatsu of The Uncommon Core podcast actually proclaimed prediction markets to be doomed in 2020 already and the article was called the problem with prediction markets so let's start with

does anybody here disagree that prediction markets are dead I think we'll have a separate conversation with you sir later very good well if where I come from they say uh in argument the truth is born so maybe we can speak after um uh can I ask is that my clicker beautiful so but you look like a smart guy maybe I run away we'll see um so let's establish the the the essence of the topic prediction markets what are they prediction markets also known as betting markets information markets etc etc these are open markets where specific outcomes can be predicted using Financial incentives and another

term that I'll use today is specialized prediction markets these are again prediction markets but these are optimized to serve a specific area of outcome outcomes and events this can be sports or politics or any other specific area let's say so looking at a brief history of modern day prediction markets um 1988 University of Iowa does an experiment around the U.S election then 1990 at project Xanadu Robin Hansen who is now called the father of prediction markets um does the first corporate prediction market for the employees of the project and they were basically betting on the cold fusion controversy then in the year of 2000 something that you might not

expect to see here but is very relevant Betfair a becomes one of the first online betting operators in the world and they're the first betting exchange now betting exchange might be unfamiliar to many of you here but a betting exchange is actually a custodial prediction Market focused on sports 2008 the term futarchy becomes New York Times magazine buzzword of the Year Phil turkey is a new form of government which is proposed by Robin Hansen uh in which prediction markets are used to determine what is the best decision uh for the public 2014 things are getting interested decentralized prediction markets are one of the few core Primitives that are very

very broadly covered in the ethereum white paper 2015 and 2016. auger which is a prediction Market platform becomes the first Ico on ethereum the noses which many of you here would know about becomes the first project that is backed by now ethereum King maker consensus years past however and there's little traction still poly Market comes to the scene in 2019 they focus on politics 2021 auger still delivers another project but still basically no traction over turbo and 2022 a project that I'm sure many people here would not be aware of as uh I was not you know like for a long time even though I was working already on this they claimed they're the largest blockchain prediction Market SX

we'll see what the state is now so that's how things look ogre are essentially defunct nurses have moved on for many many years to other projects and they found amazing successes as all of us know but prediction markets are not one of these few turkeys nowhere to be seen it's not used poly Market which is arguably one of the or probably the biggest prediction market and most successful prediction Market in the world today has normally under 20 markets with more than ten thousand dollars in liquidity S6 as we just mentioned the self-proclaimed biggest have under 400 monthly active users and Betfair the prediction Market approach to a bet online betting service although being

the biggest such service in the world actually represents well under one percent of the betting volumes and users globally that's a funny case because it's played out for many many years since the year of 2000 and the online betting Market has been won completely by a different approach now the prediction Market approach the approach is so-called bookmaker or Sportsbook there the house is playing versus The Players it is not a platform where the players are playing versus other players and the experience that the house is providing is much richer better bigger and that's been the successful case for online betting not the prediction Market approach so what are the problems with prediction markets first if we're

speaking about General prediction markets so imagine people betting on basically any topic out there um there's no natural liquidity people have not shown any inclination uh to bet really on events apart from Sports finance and politics mostly and that seems unlikely to change beyond the liquidity however if we speak about more specialized cases right for example prediction markets for sports or Finance you would see that the user experience is a big problem prediction markets do not provide in their current state an experience that is Rich enough in order to capture the attention and basically the usership of people the competition there is very strong and entrenched and it is really optimized to offer the best user experiences

so is there light at the end of the tunnel yes and no actually we created the slide like this at the end of 2021 so a couple of months after we started working on our project azuro and there was 20 plus new projects that were working on prediction markets most of them on Specialized prediction markets either for sports or for finance so there was something there there was a lot of projects some of them well funded that were thinking okay well friction markets are not so dead after all or maybe something can happen here but um but there was a problem all of the projects apart from ours were working on peer-to-peer prediction markets markets where you basically have

market makers versus players peer-to-peer exchanges um why is that a problem we argued that the problem with prediction markets in particularly specialized prediction markets that need to go after areas where there's a lot of liquidity but also there needs to be a lot of events and a lot of options for people to interact with these betting markets is the peer-to-peer design the peer-to-peer design you have a market maker on one side and you have the players or Bettors or Etc the users that are interacting with these markets on the other side the market maker is expected to provide their own liquidity so risk their own money to determine the odds the initial odds of the market so

they need to price these markets understand these markets and they need to manually create the markets so all of these things together basically create a very heavy burden for market makers which on the other hand side is matched with a very little reward as a result there's almost no markets and events market makers do not do this and especially in areas if for example we're speaking about sports as we will speak here more the Necessities for thousands of betting markets at any point in time there's so many sports leagues events teams Etc it's a lot and when there's basically none then nobody's interacting with these markets the other part of the problem with the peer-to-peer design is the user experience with the peer-to-peer design

of prediction markets you have one pool that is correspondent to one outcome in a betting market so let's say this thing is going to happen that's one pool this thing is not going to happen then another thing is going to happen Etc pool pool separate pool you have outcome specific liquidity and with that approach you have many problems which are connected to the user proposition and the user experience that people can get I'll give you one example from the sports betting world the most popular type of bet in sports is called an accumulator or a parlay in the US this is when one person takes bets in multiple betting markets and if they guess everything right

uh the outsize return of the payout you know like is what makes people crazy about these types of bets but with the peer-to-peer design that's impossible or nearly impossible with the peer-to-peer Design This is one example of the problems but there's many other examples which are basically uh in the same direction it is inhibitive of creating a rich overlay of experiences on top of the betting markets and this leads to bad user proposition and bad user experience overall so if peer-to-peer is the problem then can anybody guess what the solution might be that's that's that's already been done

like that's that's that's already done like if if we if we want to do something new okay the solution is appeared to pull um it's actually very similar and inspired by what happened in defy uh in early 2020. so you had uh some of you maybe remember Heath lent uh which is now Ave uh was it land and it was peer-to-peer Landing protocol but nobody really cared about eatland um there was a decentralized order book exchanges but then the amm came and uh we got defy and uni Swap and and so forth we argue that actually the same thing is going to happen to uh prediction markets and particularly prediction markets that are specializing

in tough areas like sports or or Finance or Politics the idea goes as follows the market Maker's role is actually broken down into three separate roles liquidity providers provide money to a common liquidity pool they do not need to understand betting markets they do not need to care about betting markets they don't need to create betting markets then on the other hand side we have data providers data providers only price the markets they specialize in pricing so these are these are more sophisticated and it is that only do pricing of markets then oracles are taking that information about the priced markets and events and they're importing this on chain

after the interaction between the oracles and the data providers you have a feed with priced betting markets this can be Limitless scale here is is of no problem then these priced betting markets meet chunks of the liquidity from the common liquidity pool the common liquidity pool is used to see the initial liquidity into the price into the price betting markets and now at the end of this whole thing you have priced and liquid betting markets from the outset markets that people can actually come and interact with directly and take positions in now so far so good but there was a problem the problem is that

liquidity pool designs as of today could not actually meet the necessity of the specialized prediction Market liquidity pool that that should pull liquidity for for prediction markets a classic liquidity pool design looks some something like this so maybe there's two assets like if it is a pair uh trading pair maybe there's more assets but the idea is that people can add liquidity and then withdraw liquidity at any point in time it is clear what amount of each asset there is in the pool this however is not the case when it comes to the prediction markets uh liquidity pool that we needed why imagine that this is the money that

goes into the pool is just one asset so maybe it's die maybe it's usdt maybe it's ethereum okay so it's one currency and then from there this needs to go into different outcomes which are part of different betting markets and these betting markets are connected to real world uh events that means that the liquidity from the pool goes into Market some of which are resolved and everything is clear and others are not yet resolved because the events have not really happened in the world yet and the markets which are not resolved uh it is unclear how much money they will return to the pool because with this design some markets are going to return more money therefore making a profit for the pool and other markets are going to

return less money and therefore losing money for the pool so allowing liquidity providers to add liquidity and withdraw liquidity with that uncertainty was a big problem the lame solution would be to track the liquidity the money that each liquidity provider provides to the pool and then where this goes with timestamps amounts Flags Etc so that you can actually stop liquidity providers from cheating so it's clear where the recruit has has gone but as you can guess with again Flags amounts and timestamps that solution is really sorry about that it's really lame because with that approach

the size of the data that is required is way bigger than the block size allows it really is not suitable for a blockchain approach so that led us azuro to create a new custom liquidity pool design for specialized prediction markets we call it the liquidity tree it's called the liquidity tree because it utilizes a data structure Which is popular in computer science which is called the segment tree the segment 3 allows four orders of magnitude of efficiency increase in terms of the amount of data that is needed in order for the liquidity pool to operate efficiently and not allow liquidity providers to cheat

well the other participants in the pool now I cannot tell you anything more about that because I do not understand more about it I'm not a technical person so if you want to ask me about this one I will not be able to say much more bottom line is it allows for the prediction markets liquidity pool to operate and to allocate liquidity from the uh from the different LPS to the different markets zooming out a little bit the liquidity pool is here it allows that magical interaction between liquidity providers oracles and data providers this is at the protocol level this creates a feed with priced and liquid betting markets then at the application Level you have front-ends essentially the marketing and the user

experience is outsourced to front-ends which can specialize in different markets or different geographies or different user experiences and then you have the players or Bettors that interact with the environments that the front-ends are creating for them as a result you basically get a decentralized ecosystem for specialized prediction markets azuru is actually live on mainnet knows his chain for a couple of months nearly three months I think now since mid-june there's uh more than 80 000 transactions 2 000 users more than half a million dollars in beds and nearly five percent of the notes chain uh active users recently this is with a very raw better proposition that only has Sports markets

and only markets basically for football and for Dota 2 and CS go so where does this lead us the current state we already see that the same slide that I showed a little bit earlier we already see one two three four five six seven seven of the other projects that we're doing peer-to-peer prediction markets for for sports and finance have already started speaking about peer-to-poo they're moving over so to end on a positive note there is Meaningful Innovation happening in the space of specialized prediction markets I'll bite a little bit rebranded to decentralized betting that uh I'm extremely excited to be a part of uh and

to be part of the project that is leading the church in that regard if you want to follow or participate or maybe even build on top of azuro then follow the link which which has some links underneath that are connected to the talk today Long Live specialized prediction markets thanks so very cool any questions yeah thanks for the talk very interesting so if I understood you right instead of betting against other players

I'm betting against the liquidity pool right yes although it's a little bit more complicated than that the initial markets are created by taking chunks from the common liquidity pool so when you have a market and there's some liquidity behind basically this is going to happen some liquidity behind this is not going to happen etc etc yeah this liquidity initially is provided by the common liquidity pool but then once people start interacting with these markets then the liquidity grows and people can bet on any of the outcomes and therefore the liquidity behind that particular Market starts growing as well so initially people bet against the pool but then after that against each other as well okay so if I take part in the pool I get

a reward or something or yes the economics are pretty cool because um it's a cash positive uh business or protocol the idea is that in the prices of the markets or the odds of the markets there is a margin or a spread okay this is the difference between the probability of the event happening and the actual price that people can take when they bet right out of that difference there is a profit for the pool over time and actually the pool is profitable uh just after 1 600 events so which is pretty it's not much when we're speaking about sports betting which really there is a lot of events so uh the liquidity pool profit goes to the

liquidity providers the front end operators and uh and to the dial and then parts of the fees are going as well to the to the data providers and the oracles okay so so because uh in Game Theory perspective like this betting Market is a zero-sum game basically so somebody's paying for the profits correct these are the people these are the these are the players these are the people who interact with the betting markets so so how do you keep the players motivated like when when they lost ones are they going to stay within the game or so like yeah so there is I think I'll answer this one

like that there is more than two billion people in the world who have placed bets on sporting events 99.9 of these people are losing money over time like if they do it for but they keep doing it so the demand is uh is there that's that's without question and actually betting in a decentralized way brings a lot of positives versus centralized betting which is huge already uh in terms of the transparency and uh the non-custodial nature people here bet directly with a smart contract and everything is visible they cannot get a decline payout nobody can ask them about something and say well I'm not going to pay you your money uh Etc so I forgot to mention that every question is like 10 die so please submit your payment later

who else wants to ask so historically what went wrong with the previous more generic prediction markets and how does that lead to these more uh specialized prediction markets like what were the learnings or what were the backslashes from regulation and why there was so much censorship and how do we come to this day then well I think from my perspective there's objective reasons for prediction markets not really taking too much traction in the past uh and some of the objective reasons were maybe blockchains

were really not so prepared like when ethereum was congested and it was slow and expensive that was not a great you know like environment to actually have a high transaction frequency you know like markets to be to be part of then I think also early teams uh had a great uh great understanding of the of the technical aspects uh but not so much understanding of uh specialization like for example if we're speaking about sports markets or Finance markets and so forth and many projects also early days were not built with the idea of Regulation so they were not positioned and designed in a way that that could withstand uh over the long term but above all we actually believe

that there was two problems one was just general prediction markets like about anything uh there's not much liquidity really people do not participate in markets unless they really care and people care about sports and politics and like really big social topics right but beyond that they don't they don't really provide liquidity to these markets and that was a big uh problem and then now I think the reality is that yeah many projects realize that they need to specialize for a particular area which has natural liquidity but the way to do it we argue is not with a peer-to-peer design thank you Rose and I think this was the last question actually because we are

ready for our next speaker and thank you so much a round of applause for Austin [Applause] when are we going to be able to bet you can already already can okay I'm gonna go and submit submit my stake later