Prediction Market Governance and Alternative Prediction Market Applications - Martin Koeppelmann

[Music] all right so as said the topic of the talk is alternative use cases for prediction markets the real topic should just be prediction markets are awesome in my opinion all right so this is an idear this has been often around is their charm that I can see the slides on the second screen anyway so this is idea that has been often around we have this concept that smart contracts are components so someone builds no identity component someone else and that means every contract can use the identity component every contact contract has the concept of identity someone else general registry every contract gets registry someone builds in reputation system every contract its in reputation system someone does a stable coin every contract gets a stable coin so I want to show the components we are bringing we bring into the ecosystem and we are building a general prediction market platform and there are a few applications I want to present so I want to present how we bring to every contract the ability to incentivize actions I want to show how every contact gets the ability to gather information and have has precise forecasts and I want to show how every contract or every for example Dow has the option to use prediction markets as a governance mechanism for that this is the base contract of a prediction market and it is really simple but that is in my opinion the good thing so it is simple and it it is also that's super important because it should be secure it should be so simple that we can create it but free because this will be a contract the potentially will hold a bunch of money or a bunch of tokens so simplicity and security is simple is super important a good thing you can build tons of stuff around this very simple base contract the idea is very simple you have any kind of collateral token it is stored in the contract and then you have end potential outcomes and for each outcome you create token token will be created and there's Oracle that later will will ping the contract and and set the contract in a way that that one of those outcomes are chosen and then the collateral token can flows to the outcome tokens or in other words the outcome tokens can be exchanged or revealed for the collateral tokens the outcome token of the winning outcome will have the exact same value as as a collateral it's important to note that the Oracle can be any kind of contract so it can be so the notes platform will be there agnostic it can be decentralized Oracle's can be centralized Oracle it can be some combination it can be centralized Oracle was a backstop mechanism as a decentralized oracle and all kind of things but now let's look at some properties this brings so again we can we have the simple function by all outcomes in contract code solidity code you put in 1 or n collateral tokens that you receive n token of each of each outcome and you can do it the other way around so you could take a complete set of outcomes and exchange it to the collateral token and finally as soon as the outcome set is set you can exchange the winning outcome token against the collateral and that brings a few relationships about the values of those tokens for example one outcome token will always be smaller or equal in value than the collateral but enough of that so not lousy now have a look let's have a look how we can use this concept in practice so the canonical example for prediction markets is to buy or sell information we want to have an information for example when will Casper be released so we can set up a prediction market and we have different outcomes let's say first outcome is early next or first quarter next year second quarter next year and so on and then we just create a market for those we discover the prices and the prices are always directly the forecasts to overcome to have good prices it's useful to use something we call an automated market maker so we reserved a bunch of funds and they will continuously create an order book or an arithmetic order book and continuously provide a price and exact price for those outcomes and everyone who who disagrees with this price who disagrees was his forecast his direct incentive to trade against the automated market maker and buys this trade providing the information the problem probably private information he or she has into the market return money if he's right but reveal the information to the market and make it public so again how would this look like in a complete example the event would be when will Casper go life could have three potential outcome tokens and we can just observe the price of those outcome tokens and know directly what is the likelihood an example we already from our life market from already six seven months ago was simple forecast and the difficulty with four different outcomes and you just see a graph how the prices change and you see the property that the price is always together some up to two one in this case to 1e cereza was a collateral token and eventually they will converge one outcome will converge against the price of one because that will be the winning outcome and eventually it will be have the same value as the collateral as easier but there are more applications we can do so another application is create an incentive so we do a market we do a creative market on something we want to happen and then we have again two outcomes it will happen or it will not happen and what we try to do now is we either give tokens of the it will happen outcome to someone who can make it happen or we will sell or auction off those tokens to them and what this means is they now have tokens they can make it happen and they know if they make it happen then those tokens will have a value of 1 so for example if you want to change the legislation the bit license legislation in New York because you think it's not a good idea you could create such an incentive market create a market will it be changed and obviously you should specify it a little bit more in a positive way and then you pile in just a bunch of money and and you hold the no tokens mmm and you try to sell or auction of the yes tokens tell you you take the S tokens and just do a simple auction whoo-hoo bid something for that for those tokens I don't that can be a very small price if someone thinks they can do something to make it happen they should buy those tokens at very little price make it happen and then reveal it against the full collateral if if they buy those tokens and they cannot manage to make it happen well in the end you will you will kind of win the you will win the money you sold it for your money will be you have a little bit more money and you just put it into the next prediction market in fact you can already set up a contract who would repeatedly do this until it is achieved so you put your money really to works you put your money in the contract and it will constantly create an incentive until it is done another one next application you can do this is hedging a hedge or insurance is basically a bit a betters bet insurance same thing so fire insurance is basically a head bed that your house will burn down if it does you win a bunch of money if not you're the amount your premium is lost concrete example and and this has been implemented like like this a lot is this flight insurance market so you predict whether or not your flight will be late you don't want it to be late but well if it is late you at least when the prediction again will my flight be late so and how you do it or how you set it up you just bring for example ten units and the insurance company brings in 90 units together those hundred units by hundred tokens of yes hundred tokens of know you as a person who wants to insure yourself you get all the S tokens the insurance gets all the note tokens if it happens well then this hundred yes tokens will be worse unready sir in in the case of no the insurance company wins but look let's have a look into this insurance example a little bit more in detail there are problems there are challenges so a big problem is that the insurance company would bring up the full collateral and no insurance in the world can upfront pay all potential claims so that's that's not a real option or to put a different it's easy to build trust alert systems if you have unlimited capital we don't have unlimited capital so we have to come up with smarter solutions but what you can do is you take a bunch of those of those no tokens and and create a basket of them so the insurance company usually has a bunch of those no tokens you create a basket of them and then you can zone users as a collateral for for future events so for example turns against earthquake in San Francisco you take the no tokens of other insurances as as a basket for this collateral but what you not should do is use token that might be related or where the likelihood of of this happening might be related so a flood in San Francisco is potentially related to earthquake in San Francisco so that's not a good idea we will go into that in more detail this general concept is if the collateral is somehow related to the next to the event then you need to be more careful for example this prediction market would not make any sense if you predict that the value of easier is zero but you use either as a collateral who would bet on yes so you bet on yes you win a bunch of easier if it happens but they are worth zero so that doesn't make sense but you can set it up in a way that it makes sense so you can set it up that you use those tokens as a collateral for the next event and what this what this is in effect that you do here a prediction that this outcome and that outcome that outcome happen so concrete example event a event B and essentially if you buy tokens of event B you do a prediction that a1 happens and b1 happens and here comes more interesting stuff so here by setting up those markets you can see whether or not there's a relation between event a and event B if there's no relation between the two then the ratio between PA 1 and PA 2 should be the same as the ratio between P 2 a 2 and B well yeah we be P 1 a 2 and P 2 - so simple example will the sunshine tomorrow but the and will the price of ether be above $20 in let's say two month that's potentially or I'm pretty sure that is an independent event if you would do the same thing as Polonius would get Hecht it's probably not an independent event and you would see in this case that the ratio here is is different so pretty sure if Polonius gets Hecht then the likelihood here is smaller that it will be bigger than $20 so again for the predict for the idea of using a basket as a collateral for insurances all the events the event tokens on the basket in the basket should be independent and you have a way to measure the independence with this concept now we can use the next application you can use for prediction markets is to use more or less this concept of doing predictions on the effect of a potential event for governance so the concept there's a general step-by-step concept how you can leverage a predictor market to govern your DL your smart contract first of all you need a continuous market on some measurable objective of your Dao so something like maximize revenue make some eyes a token the value of the token something like that and then you can have the proposal system anyone could make a proposal or maybe a group of people could make a proposal and as soon as someone does a proposal you create those two markets first will the proposed will be implemented yes/no and then second step how likely is this that we achieve our objective and you can directly measure or you you love trading for a bunch of time you potentially set up an automated market maker to get continuous price information and then the contract could automatically implement the decision that will maximize the forecast for for your objective so in the example of Knossos we want to have a fund for noses that will find any things that get built on top of noses so our simple objective for the noses platform we want to maximize the volume on the platform and then there could be the idea well maybe we should fund this sports betting interface and we set up this market and we can observe you can directly see from the prices will it help our objective and if it does we automatically implemented there are a few things we want to measure and observe before we actually implement it so we will be careful in gnosis we will not directly use foo taki with some more human control but eventually if we go through those steps we it would be our goal to completely have it governed by for turkey so we need first to make sure that indeed those forecasts on this objective performance indicator are good and then the next step would be that we can do good forecasts under those conditions and then what about how difficult or how easy is it to manipulate enact actually we have a send of a set of experiments that we will start as soon as our security audit is done and we will create incentives for someone to break the system to manipulate a market and we will see how resilient markets are because if you automatically make decisions based on markets then of course it should not be possible to manipulate it so the ultimate test or the ultimate test before we would use foo turkey for four noses would be something like can we beat alphago so can we do food turkey mechanism that is capable of aggregating all available information in a way that it can beats best current machine learning approach so you could set it up in a way that you always ask the question will we win the game and then what what should be our next move and we have a bunch of moves and we are again observes the market which move will lead to the highest chance of winning the game all right tonight meet us tonight for drinks join our WeChat select widget group or just on on slack and let's continue the conversation about predict markets thank you [Applause] [Music]