Monetizing Crazy Times with Prediction Markets - Anne-Grace Kleczewski

Transcript

Good afternoon, everyone. So, as mentioned, I'll be presenting a topic about prediction markets. And funny enough, I'm a lawyer in Switzerland, but over the last two years, I show up in Prague for a completely non-legal topics. So, I always have this legal lens of let's see what may go wrong and what are all the caveats somewhere and then I present something that is not legal at all. This time it's going to be prediction markets cuz it's the biggest hype. Everyone is talking about Polymarket, Kalshi. Can we bet on Trump whether he will say a specific words tomorrow in his speech or not? Where will the next missile land during some wars? This is the trend now.

Now, prediction markets also presented as a new source of truth. We're going to have crowd intelligence that's gets out of this. Now, is it really a source of truth or is it just some bluff? That's the topic today. Well, first of all, let's remind ourselves that it's the old idea in a new stack. Why an old idea? Because prediction markets are not new. Polymarket and Kalshi, this is just a new implementation of something that existed before. And the real question is are those new players adding something to the landscape? Are they solving something?

And that's the only legal reference I'll make, it depends, actually. So, some of them are improving what we had before, while others are just rebuilding the same thing on web free rails. Now, crazy times. Why is it important? Because as you'll see with the examples I'll mention afterwards, prediction markets were always tied to things that stir our feelings. It has to be important enough that we want to bet on it. And it was always about politics, elections, sports, and wars, unfortunately. And these last months we have like so many events, like even the last years. This is kind of what feeds also the markets. If there's nothing interesting happening, what should I vote on? Just like the the next football match.

So, the old wave. Well, I was actually surprised myself when I was researching. There's something that goes all the way back to 1868 and was already referred as a prediction market in the US regarding elections. And then when you look at the other dates regarding like the Hollywood movie exchange or like the IO electric markets, well, we have 88 96 2014. So, really not so fresh, not so new. It was just out there. Now, what about the old wave? There are four things. The old model worked. I mean, there was traction, user people use it, but it was not as prominent as now. Like Polymarket is everywhere on the news, same for Cashi.

How come that we haven't mentioned these examples instead? Well, first of all, regulatory uncertainty. All of these examples were not big enough to really stir like the debate in terms of should we regulate it differently? Uh some of these players as the IO example, they were operating on a nonprofit basis cuz it was mostly academic research. Um not all of some of them had caps, for example, meaning that the regulator just like issued non-action letters or like specific frames that work for this specific project. And then it was just living like that. There was also a custodial risk. I mean, all of these platforms if money was involved, well, they were holding the funds. There wasn't a web free infrastructure that

allows all the non-custodial solutions. There's also resolution opacity in a lot of cases. Like how does it work behind it? Like who is telling the truth? Who is deciding on the weather what is was the right outcome? And how are things allocated? You had to trust the platform. And limited access. But this is related to regulatory uncertainty. Well, because of this like it was probably just access some of them in the US, some It's like very local. Now, the new wave, the new names. I mean, we've all heard about Polymarket, but you have Limitless as well, Augur, uh, Eximchain, that also is like a new rising player. You have Cindicator I added it here.

Well, the new wave is actually very, uh, eclectic. Not all of them are the same. I mean, all of them are referred to themselves as prediction markets. Some of them say conviction market as Eximchain but it's the same principle. Now, old idea, new stack. So, what changes with those markets? Well, before I dive into the details, the biggest thing that changes is the scale. The previous example, some of them like the Hollywood Exchange it was like about fake money. Others were just very local American politics. Others were also like focused on specific things that like sports for example. Here we have platforms that aim at being as broad as possible. The slogan is bet on anything you want, basically.

And what else is being improved? Are we reinventing the wheel or just doing something completely new? Well, they're kind of improving the past with some design choices. Well, first of all, custody. We shift from something that just you need to trust a custodian to smart contract base market creation. Some of these examples have permissionless market creation. I can decide tomorrow to create a market. I could have created a market about how many people would show up in this room. This was not possible before. You had to, let's say, kind of pray that the market you're dreaming about is important enough that someone created it. Uh, market resolution as well. We're moving towards more like decentralized or at least attempts at decentralize and more

transparent manners of resolving markets. Settlement, which is on chain. You can see what happened there. You don't need to trust uh, a centralized uh, entity. And then the access to go, especially for the permissionless decentralized platforms is to have everyone being able to participate and not just people from specific jurisdictions. Now, this new stack and these new benefits also come with their own wave of risks because obviously custody, if custody is with smart contracts, you add a technical risk because smart contracts could be drained. Market creation, if you have something that's permissionless, well, people could create markets on very shady things. And this is something that was reminded in

general, you're not supposed especially by the CFTC in the US, you're not supposed to create unregulated setups, markets on illegal and moral things. But on paper, if it's permissionless, I could create something as let's say negative if will this person get killed tomorrow, and nothing would prevent me from that. Same for market resolution. Well, having something that's decentralized and transparent doesn't solve another problem that I will mention more in detail is is the data I'm relying on reliable and truthful in the first place. And I'll dive more into that, but the fact that I don't know, a jury for example or several oracles are being used doesn't solve that issue fully.

Um settlement on chain, the same. Technical issues could arise, and in terms of access, I will see as we'll see afterwards, it wasn't really solved. So, waves within a wave. Not all of these market, I said there's like very um diverse landscape, all of them are prediction markets, but not all of them have the same approach. And some of them are simply centralized and regulated, that's the Kalshi example. They're kind of just reinventing the wheel because it's the same thing that we had before, but at the larger scale. Some of them are using decentralization just in doses, not all of it, but part of it. So, they are improving the wheel.

I would say Polymarket is kind of a good example of that with their resolution system, but some other things are still like quite centralized. Um and then you have decentralization at the core, offering a completely new wheel design. And this is, I would say, mostly axle markets, where all the layers are supposed to be decentralized, and they build everything on top of their own layer one. Um, the new wave with old flaws. Well, this is the main thing, and for whatever reason this does not appear. Yes.

So, one of the things that was wrong with the first wave, the old wave that did not gain as much traction, was regulatory uncertainty and limited access. And when we see, well, for example, Kalshi, after all I said they're just reinventing the wheel, but same thing. I tried to connect. I'm in the wrong jurisdiction. I cannot. I would need to be in the US, because, and I added the whole list, all of these countries, and you'll see all of the European countries included, cannot access it. It's US-based, just accessible there. Uh, this comes from I should have remembered, but I don't remember which one anymore. I think it's Limitless.

Same, a whole list of exclusion. And here it's the opposite, because they're not regulated in the US, they don't want to touch that jurisdiction. You cannot access it from the US, and then there's a whole list of countries being, um, there because of sanctions. So, we haven't really solved that thing, except if you say you are completely redesigning the way prediction markets work, and you're fully decentralized. And this is the same thing as we're doing with DeFi. We are decentralized, we're not regulated. Narrative works only if you are fully decentralized, Polymarket kind of realized that there is there are limits to that, because they got fined in the US because of how they operate. And I see increasingly projects who want to serve that wave, but saying, "But, hey, these

players are out there. They're not regulated, they're offshore, they do this, they do that." And then, what happens? Well, the minimal level, you're going to get blacklisted in a country, and your website will not be accessible from there. This is not too bad. Worst thing that could to is actually that someone considers they have jurisdiction over you and files for criminal charges because a lot of countries have gambling regulations in place or financial regulations depending on which one is applicable. Um, and this depends on jurisdictions, but it's not the topic today. And then it may get worse.

Now, the prediction Motley Crew. Why do I call it that way? Because as I said, all build prediction markets, but not all of them are built the same. On top of like how they approach the whole thing, well, it's more than just centralized versus decentralized. This would be too simplistic. There's like a whole lot of questions that they are not tackled the same way depending on the platform. First, well, the permissionlessness. Well, this is very important in terms of scale moderation risk. If like it's permissionless, do I know whether these markets were defined in the correct manner? Is there a moderation policy that makes sure that this market will be solved correctly because it was created correctly?

Is the custody centralized or smart contract base? Not all of them are smart contract base. Some of them just utilize fiat. Um, is settlement on chain or platform controlled? Also very important and I'll come back to it. Like, what is the model? Is it an order book model or is it an AMM? It influences a few factors as well in how you interact with them. Is the platform regulated? That's the information that is kind of hard to see on most of them. If it's regulated, you'll see it. It's in plain sight. For all the other ones, what happens is like just guess where they are established. Um, is there a centralized monitoring or moderation? Like, is it truly just a decentralized wild west or is there something in place to ensure that the bad seeds are taken away?

Um, um, how can users pay? This is also important. How are fees charged? What happens there? Is it like which share goes away? Often the documentation of these projects hands on this information, but sometimes you really need to dig deeper to find it. And finally, how are markets resolved? And this is the most important because it will directly influence whether this thing you hope to earn at the end if you're right on your prediction, goes to you or not. Dangerous ways. Well, there's an interesting vulnerability in all prediction markets, and this is the the biggest flaw regardless of the design.

It's basically the data that will solve for resolution. So, the resolution requires the data confirming whether the event occurred or not. Like, is it true, false, what was the price, what's the weather, and so on. Well, the data can be influenced. So, first of all, not all events are clear. And this happened in a few cases. Uh for example, will Zelinsky wear a suit? Well, was he wearing something that is actually defined as a suit or not? It raised like a whole debate just knowing was this a suit? Same for something else. Like, there are some concepts that are just complex definition, and then you try to create a yes or no market out of it. How does that work? Data may be also preliminary or evolve.

What are the numbers in terms of how many people watched a specific event? Well, you may have preliminary data, and a few hours later the final data. If it's resolved based on the first data, what happens? Um data may be also tampered by third parties that actually just control the data feeds and this kind of things. But, also events themselves may be tampered, and maybe you've heard the example of the guy with the the hair How do you call this again? The thing for the hair to dry the hair the hair dryer, thank you. Who used a hair dryer to influence the temperature on a sensor, so it goes up. And he won a lot of money out of it. I mean, if I can earn a few thousand just with a hair dryer, why not? So, this is also something that's possible. And all of

these elements basically influence the risk factor of markets. So, regardless the platform, when you access a market, you need to have this in in mind and say, "Okay, I'm maybe putting a thousand bucks there, but the event is quite difficult to define. It may raise us give rise to some contentious discussions, so potentially it's going to be resolved in a manner that doesn't align with how I understood it. Or the event depends on the decision of one person, so it's kind of hard to know what happens there. Same for like football matches. We all know how like some teams may I mean now less because there's a lot of control, but someone could be just set up and say, "Oh, I I take the money. I'm going to lose the the matches because it allows someone to win." So, are there solutions to that? Well, obviously if

there's like a good dispute resolution system or if like there's a way of contesting the results because something was solved too early or based on erroneous data, that's already a good start. I mean, if there were cool off periods to avoid issues when there's like preliminary and like final data, that would also be nice, but I haven't seen examples implementing them. I mean, if you know some of them, please let me know at the end. Also shifting away from the betting on anything and or accepting the risk could be a solution. What do I mean by shifting away from betting on anything? I mean, some platforms as Limitless, for example, have like specific topics that are put forward. If you just need to bet on whether the price of Bitcoin will hit a specific threshold tomorrow,

there's some risk of manipulation, but it's way smaller than when you have a market that's defined in broad terms. So, this may be a a way of just rendering those markets a bit more easy to comprehend and like less prone to disputes or just accepting the risk. We say we bet on any everything, but if something is resolved not according to my understanding, I'm fine with it. This would be the most reasonable approach, but as we've seen in a lot of things of like being non-custodial, using DeFi, and so on, people like to blame someone, so obviously I don't think this would be the the way of going forward. Now, dangerous waves, what else happens there?

Well, there are also some key design elements that actually influence the risk beyond just how markets are presented. First of all, whether it's custodial or not, and how things are being there like I've mentioned it already before smart contract risk or if you have like a regulated intermediary and so on. But you also have just a problem of resolution and this is like the major thing. It's kind of opaque how it's going to be solved and it's not always as straightforward and it's not like a big tsunami coming and you'll be like, "Oh damn, I know this resolution will be completely messed up." No. You have like things in place. You have extra market has a system called Modra for example of a jury and like AI agents also intervening first. So basically small markets AI agents decide on the

outcome. If there's a problem it's escalated to a jury of people and so on. So they have like a whole thing that's like AI agents and then decentralized setup. Straightforward, right? Same Polymarket has also like a system if it doesn't work out you escalate it but then yeah, if it escalates and it's being then overruled just by a centralized entity does it really work? And then what were also the things at stake there? Like why was it overruled? Is it just because people complained or was it because there was an actual system? And there were a few cases where basically people realized that they want to complain about the markets and there's no clear procedure.

Or there is a clear procedure but it can be overruled by someone. So this is like a whole thing that basically influences what money you get in the end. Now surfing the waves uh just quickly here um well beyond this first aspect as I mentioned how is the market defined, how is the market solved all of this is also about liquidity and market makers being involved and just the model influences a lot for them. Like market um so AMMs or order books have very different characteristics and I've heard someone say something very interesting by saying basically AMM based prediction markets don't make sense for market makers. Well, someone else told me that actually it could make sense to a certain extent but

there's like a lot of differences. So saying "Oh be a market maker on a prediction market." Well, just tell me what kind and take all of these elements into account. Now, all in all, I will not if you're interested in the market maker part, just let me know, but I I think it let's get to the most interesting part. So, basically, um do prediction markets actually tell the truth? Well, they don't really tell us what will happen. It's not like reading in a crystal ball, but what they can say is what a specific group of financial incentive wise participants actually believe will happen under a specific design. The design directly influences that. If the wording is clear, it's easier to trust, but I

highlighted three different points. The objective objective resolution it's easier to enforce. So, if the resolution is kind of like the system is weird, the wording is weird, like how do you work out there? We will not get the truth in the end. We'll just get like something that's more like of a bluff outcome. The credible oracle, very important as well. What oracles are being used? Is this one? Is there a backup one? Is it several one? And data is being combined. How does that work? It also influences whether we are more attending towards the truth or just like whatever outcome that's basically convenient to get some quick money.

And also, we tend to forget about this biased participants. Like, what about the fact that people who vote are actually simply biased and they're not voting on what is the truth? They just vote on what they believe will be the truth. And I mean, funny enough, when I check the numbers, and I don't know to what ex- tent we can rely on these numbers because most of these markets have are permissionless, not fully regulated, whatever, but a lot of these users are in the US. Then second comes Israel, and then you have uh people in some other countries. Let's guess why. US, a lot of things are happening. Trump is creating a lot of traction. You can be easily biased there. Israel, the same because of the whole situation in the Middle East. So, to what extent are people voting there

objective or biased? And we've heard about some insider trading issues in both of these jurisdictions. So, when these people vote, they just want to make some a quick buck. They don't want to tell us the truth on what is their actual conviction. So, because they don't tell us what will happen, well, this is kind of what we end up with like people just being betting there and being on what will be the best outcome for them financially. And like for news, at the end of the day, the question is always what is the source of information? Same for prediction. When we say those predictions are influencing um the truth and kind of giving us like some insights, crowd intelligence, okay.

Just let's wait wait for a second and think about all of these design aspects. Okay, so the source is a prediction market where things are designed that way and specifically a market that was set up in the way that is very well designed, easy to solve. Okay, maybe it's a source of truth. If not, then potentially it's like the equivalent of like a shitty newspaper that is kind of interesting, but you need to double-check the information. Um so, greatly designed beliefs certainly contribute to crowd intelligence, but poorly designed ones are merely depicting attempts at making some quick money. Now, if this was of interest to you, let's keep in touch.

Uh beyond talking about prediction markets, I'm also advising them and advising other projects in DeFi. Um and it was a pleasure to present to you. If you have any questions and I have time for questions, I don't know, maybe uh let me know. >> Yeah. Thank you for the presentation. Any what um like questions about pre- prediction market for Anna? Yeah. If not, yeah, you get the Uh we will be moving for uh