Editor's Note: This article comes fromBabbitt Information (ID: bitcoin8btc)Editor's Note: This article comes from
Babbitt Information (ID: bitcoin8btc)
, Author: Vitalik Buterin, Compiler: Free and Easy, released with authorization.
Note: The original author is Vitalik Buterin, the co-founder of Ethereum. In this article, he reviews his experience in participating in the prediction market, and summarizes the four major problems facing the current prediction market. Vitalik also predicts that in the next few years, the prediction market will become an increasingly important application of Ethereum, and the 2020 US presidential election is just the beginning, followed by conditional prediction, decision-making and other applications.
Special thanks to Jeff Coleman, Karl Floersch, and Robin Hanson for critical feedback and comments.futarchyWarning: I express some political views.
Prediction markets, a topic of interest to me for years, allow anyone to place bets on future events, and use the odds of those bets as a trusted neutral source of predicted probabilities for those events, which is a fascinating application . closely related ideas (such as
), has always interested me as an innovative tool that can improve governance and decision-making. Prediction markets are also a fascinating application of blockchains, as Augur, Omen, and most recently PolyMarket (note: all three applications are built on Ethereum) have shown.
During the 2020 U.S. presidential election, it seems that the prediction market has finally entered people's sight. These blockchain-based markets have grown from close to 0 in 2016 to millions of dollars in 2020. As someone who is very interested in seeing Ethereum applications cross the chasm and become widely adopted, this certainly intrigued me. In the beginning, I tended to simply observe rather than personally participate: I am not an expert on US electoral politics, so why should I expect my views to be more correct than those of others already trading? But in my twitter circles, I see a lot of very smart people I respect who think the market is in fact irrational, and I should get involved and bet against them if I can. Eventually, I was convinced.
I decided to do an experiment on a blockchain app I helped create: I bought $2,000 worth of NTRUMP on Augur on August 1st (holding this token gives you 1 Dollar). Little did I know at the time, my position would eventually grow to $308,249 and earn me over $56,803 in profit. And what happened over the next two months proved to be a fascinating case study of social psychology, expertise, arbitrage, and the limits of market efficiency, with a keen interest in any possibility for economic system design people have significant influence.
before the election
My first bet on this election, it's not actually on the blockchain at all. When Kanye announced his run for president last July, a political theorist I usually respect immediately took to Twitter to claim that he believed it would split the vote against Trump and lead to a Trump victory. I remember thinking at the time that he was overconfident in this point, and maybe even the result of an overly internalized heuristic (i.e. if an idea seems smart and different, it is likely to be true). So of course I offered to bet $200 on my own boring pro-Biden views, which he gladly accepted.
The September elections caught my attention again, this time by prediction markets. The market is giving Trump a nearly 50% chance of winning, but I see a lot of very smart people in my Twitter circle pointing out that that number seems too high. This of course leads to the familiar "efficient market debate": If you buy a token at $0.52 that gives you $1 when Trump loses, the odds that Trump actually loses Much higher, why don't people just come in and buy tokens until the price goes up even higher? If no one else is doing it, why do you think you're smarter than anyone else?
Ne0liberal did an excellent job with a Twitter post ahead of Election Day summarizing how accurate prediction markets were at the time. In short, the (non-blockchain) prediction markets that most people use before 2020 have various limitations that make it difficult for people to participate with small amounts of cash. The upshot is that if a very smart individual or professional organization sees the possibility of what they think is wrong, their ability to push prices in the direction they think is right will be very limited.
The most important limitations noted by the article are:
Low limit per person (less than $1,000);
High fees (e.g. PredictIt charges a 5% withdrawal fee);
Here's what I objected to ne0liberal back in September: While bland old world centralized prediction markets may have issues with low limits and high fees, cryptocurrency markets don't! On Augur or Omen, there is no limit to how much someone can buy or sell for an outcome token if they think the price is too low or too high. However, the price performance of the blockchain-based prediction market is in sync with PredictIt. If the market is really overvaluing Trump because high fees and low transaction limits prevent calm traders from outperforming overly optimistic traders, why would a blockchain market without these problems show the same price Woolen cloth?
The main response to this from my Twitter friends is that the blockchain-based market is very niche with very few players, especially few people who know much about politics. This seems plausible, but I'm not too confident in the argument. So at the time I bet $2,000 that Trump would lose, but didn't place any further bets.
election time
Then the election happened, and in the initial panic, Trump first won more seats than we expected, but Biden ended up being the winner. Whether elections themselves validate or refute the efficiency of prediction markets is, as far as I can tell, an open topic to explain. On the one hand, with standard Bayesian applications, I should lower my confidence in prediction markets, at least relative to Nate Silver. The prediction market gave Biden a 60% chance of winning, and Nate Silver gave Biden a 90% chance of winning. Given that Biden did win, it just proves to me that I live in a world where Nate gave the more correct answer.
But on the other hand, you can make a case where the prediction market better estimates the edge cases of victory. The median of Nate's probability distribution is roughly 370 votes for Biden out of 538:
The Trump market doesn't give a probability distribution, but if you had to guess the probability distribution from the statistic "40% chance of Trump winning," you'd probably give Biden about 300 votes. And the actual result was 306, so the net score of prediction markets Vs Nate is ambiguous in my opinion.
after the election
But what I couldn't imagine at the time was that the election itself was just the beginning. Days after the election, Biden was declared the winner by major organizations and even a handful of foreign governments. As expected, Trump launched various legal challenges to the election results, all of which quickly failed. But for over a month in that period, the price of the NTRUMP token remained at 85 cents!
At first, it seemed reasonable to speculate that Trump had a 15 percent chance of overturning the election, after all, he appointed three Supreme Court justices at a time when partisanship was intensifying. However, over the next three weeks, it became increasingly clear that the challenge was failing and Trump's chances of winning seemed to me to be getting slimmer by the day, yet the price of NTRUMP tokens did not change (in fact, It even briefly dropped to around $0.82). On December 11th, just over 5 weeks after the election, the Supreme Court decisively and unanimously rejected Trump's attempt to overturn the vote, and the price of NTRUMP finally rose...to $0.88.
Just in November, I was finally convinced that the market skeptics were right, so I jumped in, betting myself that Trump would lose. The decision wasn't about money, after all, after just two months, I was making a lot of money just holding Dogecoins, but I chose to donate them to GiveDirectly. Rather, I participated in the experiment not just as an observer, but as an active participant, which helped to improve my personal understanding of why others have not bought NTRUMP tokens before me.
Operating procedures
There are two ways to bet that Trump will lose with Catnip:
useFoundryBuy NTRUMP tokens directly on Catnip using DAI;
use
to access the Augur function, which allows you to convert 1 DAI to 1 NTRUMP + 1 YTRUMP + 1 ITRUMP (the "I" here stands for "invalid", more on that later), and then sell YTRUMP on Catnip;
At first, I was only aware of the first option. But then I found out that Balancer provides more liquidity for YTRUMP, so I switched to the second option.
There is another problem: I don't have any DAI, I only have ETH, I have the option to sell my ETH to get DAI, but I don't want to touch my ETH holdings. It would be a shame if I won my $50,000 bet on Trump but lost $500,000 on ETH price movements at the same time. Therefore, I decided to keep my ETH exposure intact by opening a Collateralized Debt Position (CDP, now also called a "Vault") on MakerDAO.
CDPs are the way to generate DAI: users deposit their ETH into a smart contract and are allowed to withdraw no more than 2/3 of the value of the ETH invested in newly generated DAI, when they return the same amount of DAI, plus an additional interest (currently 3.5%), they can withdraw their mortgaged ETH. If the value of your deposited ETH collateral drops below 150% of the value of your borrowed DAI, then anyone can come in and "liquidate" the vault, forcibly selling your collateralized ETH to buy back DAI, and charging you a hefty amount fine. Therefore, having a high collateralization ratio is a good idea in case of sudden price fluctuations. For every $1 I withdraw, there is over $3 of ETH in its CDP.
The picture below is my entire operation process.
I've done this many times, and Catnip's slippage, means I can usually only trade up to about $5,000-$10,000 at a time so that the price doesn't turn against me (when I skip Foundry and buy NTRUMP directly with DAI , the limit is closer to $1,000). After two months, I have accumulated over 367,000 NTRUMP tokens.
Why don't others do it?
Before I jump in, I have four main hypotheses as to why so few people are buying dollars at 85 cents:
Fears that the Augur smart contract will go wrong or that Trump supporters will manipulate the oracle (Augur REP token holders vote on one or more outcomes through their tokens) so that it returns the wrong result;
Cost of funds: To purchase these tokens, you must lock your funds for more than two months, which will prevent you from using these funds or making other profitable transactions during this time;
It's too technically complex for everyone to participate in;
In reality, far fewer people than I would imagine are truly motivated enough to take an odd opportunity, even if it's obvious.
All four points make sense, smart contract attacks are the biggest risk factor, Augur oracles have never been tested in such a contentious environment. The cost of capital is real, and while betting on prediction markets is easier than betting on the stock market (since you know the price will never go above $1), locked capital competes with other lucrative opportunities in the crypto market of). Additionally, transacting in dapps is technically complex, so people naturally have a certain level of intimidation.
And my experience actually getting into this financial field, and observing the evolution of market prices, taught me a lot about these assumptions.
At first, I thought that "fear of smart contract bugs" must be the main worrying factor. But as time went on, I became more and more convinced that this might not be a dominant factor. One way is to compare the prices of YTRUMP and ITRUMP. ITRUMP stands for "Invalid Trump", where "invalid" refers to the outcome of an event that triggers in certain special circumstances: when the description of the event is unclear, when the market resolves, when the outcome of the event is unclear , when the market is immoral (such as the assassination market), and a number of other similar situations. In this market, the price of ITRUMP has remained below $0.02. If someone wanted to make a profit by attacking the market, it would be more profitable for them not to buy YTRUMP at $0.15, but to buy ITRUMP at $0.02. If they buy a large amount of ITRUMP, they can force the "invalid" result to actually trigger, allowing for a 50x return. Therefore, if you are concerned about being attacked, purchasing ITRUMP is by far the most logical choice. However, few choose to do so.
capital cost
Of course, another argument that precludes worrying about smart contract vulnerabilities is that in every crypto application other than prediction markets (e.g. Compound, various yield farming schemes), people show a surprising level of concern about smart contract risk. disregard. If people are willing to put their money into all kinds of risky and untested schemes (even if they only promise 5-8% annual returns), why are they suddenly becoming overly cautious here?
capital cost
The cost of capital (i.e. the inconvenience and opportunity cost of locking up large amounts of capital) is a challenge, and I recognize this more clearly than ever. From the perspective of Augur alone, I need to lock 308,249 DAI for two months to get a profit of $56,803, which is about 175% annualized profit. Compared with private mining activities, this kind of income is also considerable. But it gets a little worse when you consider that I need to operate on MakerDAO. Because in the case where I want to keep my ETH position, I need to obtain DAI through CDP, and using CDP safely requires a mortgage ratio of more than 3 times. So the total amount of funds I actually need to lock up is about $1 million.
So looking at it now, this rate of return does not look very favorable. And, if you factor in the possibility of a possible hack or a truly unprecedented political event, the appeal of being involved in it becomes much less attractive.
But even so, assuming that the locked funds are 3 times, the possibility of the Augur contract being attacked is 3% (I bought ITRUMP to hedge the risk), which can reduce the risk neutralization rate to about 35%, and if you consider the real Perceptions of risk are even lower. This deal is still very attractive, but on the other hand, it seems very understandable now that such numbers, for those participants who experience fluctuations of up and down 100 times in the cryptocurrency market, are very attractive. not enough.
Trump supporters, on the other hand, didn't face these challenges: They canceled my $308,249 bet for only $60,000 in it (I won less than that because of fees). When the probabilities are close to 0 or 1, as is the case here, the game is very unbalanced, favoring those who try to push the probabilities away from extreme values. Not only does this explain what happened with Trump, it's also why various popular niche candidates with no real chance of winning often receive as much as a 5% chance of winning.
technical complexity
At first I tried to buy NTRUMP on Augur, but a technical glitch in the UI prevented me from placing an order directly on Augur (other people I spoke to didn't have this problem...I'm still not sure exactly what happened). Catnip has a much simpler UI and it works really well. However, automated market makers like Balancer (and Uniswap) are best suited for smaller trades, for larger trades the slippage is very high. This is a nice microcosm of the broader "AMM vs order book" debate: AMMs are more convenient, but order books are really more efficient for larger transactions. Uniswap v3 is introducing an AMM design with better capital efficiency, we will see if it improves the situation.
There are other technical complications as well, though luckily they all seem to be easily resolved. There's no reason an interface like Catnip couldn't integrate the "DAI->Foundry->sell YTRUMP" path into one contract so you could buy NTRUMP tokens that way in a single transaction. In fact, the interface can even check the price and liquidity attributes of the "DAI->NTRUMP" path and the "DAI->Foundry->sell YTRUMP" path, and automatically provide you with a better deal. Even withdrawing DAI from MakerDAO CDPs can be included in this path. My conclusion here is optimistic: technical complexity issues are a real barrier to people's reluctance to adopt prediction markets at the moment, but as the technology improves, using them will become much easier.
lack of confidence
Now, we have the final possibility: A lot of people (especially smart people) suffer from excessive humility, so it's easy to conclude that if nobody else is doing anything, then there must be a good reason why The action is not worth taking.
Eliezer Yudkowsky makes this point in the second half of his excellent book Insufficient Equilibrium, arguing that too many people overuse "humble epistemology" and that we should act more on the results of our reasoning, even when they show that The vast majority of people are irrational, lazy or wrong about something. When I first read the chapters, I wasn't convinced, it seemed like Eliezer was just being too arrogant, but after going through this experience, I saw some wisdom in him.
This is not the first time I have witnessed the virtue of trusting my own reasoning. When I first started working on Ethereum, I was initially plagued by fear that the project was doomed for some reason. I deduce that a fully programmable smart contract blockchain is clearly a vast improvement over what has come before, and there must be a lot of people out there who thought of it before me. So I fully expect that once I publish this idea, many very smart cryptographers will tell me why something like Ethereum is simply not possible. However, no one has ever done this.
Of course, not everyone has a fault with excessive modesty. Many people who predicted that Trump would win the election can be said to have been fooled by their own excessive contrarian psychology. Ethereum benefited from my repression of my own humility and fears in my youth, but there are many other projects that could benefit from more intellectual humility and avoid failure.
However, it seems to me more true than ever that, as the famous Yeats quote goes, "The best men lack self-confidence, and the worst men are passionate," and it seems to me that the whole Society spreading the message that the solution is simply to trust society's existing outputs, whether in the form of academic institutions, media, government, or markets, does not seem to be the solution. All these institutions work precisely because someone thinks they don't work, or at least someone thinks they might be wrong some of the time.
Futarchy lessons
Witnessing first-hand the importance of the cost of capital and how it interplays with risk is an important evidence against which to judge a system like Futarchy. Futarchy and "decision markets" in general are an important and potentially very useful social application of prediction markets. A slightly more accurate prediction of who will be the next president has little social value. But there is a lot of social value in conditional predictions: if we do a, how likely is it that it will lead to something good X, and if we do B, what is the chance? Conditional predictions are important because they not only satisfy our curiosity but also help us make decisions.
While election prediction markets are far less useful than conditional predictions, they help shed light on an important question: how resistant are they to manipulation, or even just bias and erroneous views? We can answer this question by looking at how difficult arbitrage is: Assume that the current probabilities given by the forecast market are (in your opinion) wrong (perhaps due to ill-informed traders or obvious manipulation attempts). How much impact can you have, how much profit can you make, by setting things up right?
Let's start with a concrete example. Suppose we are trying to use a prediction market to choose between decision A and decision B, where each decision has a probability of achieving some desired outcome. Say you think decision A has a 50% chance of achieving the goal, and decision B has a 45% chance. However, the market (wrong in your view) thinks there is a 55% chance of decision B and a 40% chance of decision A.
Assuming you are a small player, your individual bets will not affect the outcome, only if many people bet together. So how much should you bet?WolframAlphaThe standard theory here relies on the Kelly formula. Essentially, you should act to maximize the expected logarithm of the asset. In this case, we can solve the resulting equation. Let’s say you put some of your money into buying token A at $0.4. From your perspective, your desired new log wealth is:
The first term is the 50% chance (from your point of view) that the bet pays off and that part of your investment grows 2.5 times (because you bought the dollar for 40 cents). The second term is that there is a 50% chance that the bet will not pay off and you lose some of the money you bet. We can use calculus to find a way to maximize this, for the lazy you can use
, the answer here is: r = 1/6. If someone else buys and the price of A in the market rises to 47% (and B falls to 48%), we can redo the calculation for the last trader who flips the market to correctly favor A:
Here, the expected log wealth maximization r value is only 0.0566. The conclusion is clear: when decisions are close and there is a lot of noise, it turns out that it only makes sense to invest a small portion of your money in the market. This assumes that rationally, most people invest less in uncertain gambles than Kelly's formula says. Capital costs are even higher. However, if the attacker really wanted to force outcome B for personal reasons, they could spend all their funds on that token. In general, the game can easily favor the offensive side of more than 20:1.
In reality, of course, attackers are rarely willing to bet all their funds on one decision. And futarchy is not the only vulnerable mechanism, the stock market is equally vulnerable, and non-market decision-making mechanisms can also be manipulated in various ways by determined wealthy attackers. Regardless, we should be wary of thinking that futarchy will push us to new heights of decision accuracy.
Interestingly, the math seems to suggest that futarchy works best when the intended manipulator wants to push the outcome to an extreme. An example of this might be liability insurance, since someone wishing to obtain insurance improperly would effectively attempt to reduce the market estimated probability of an adverse event to zero.
Can prediction markets get better?
The final question to ask is: are prediction markets destined to make the same mistake? Like it gave Trump a 15% chance of overturning the election in early December, and even after the Supreme Court (including three of his appointees) told Trump to fuck off, it gave a 12% chance of overturning the election that serious? Surprisingly, I don't think prediction markets will repeat their mistakes, and I see some reason for optimism.
1. The market is naturally selected
First, these events gave me a new perspective on how market efficiency and rationality arise. Proponents of market efficiency theory often claim that markets are efficient because most participants are rational (or at least more rational than any group of superstitious people), which is an axiom. But, instead, we can look at what's going on from an evolutionary perspective.
The crypto market is a young ecosystem. Despite Elon’s recent tweets, the ecosystem remains disconnected from the mainstream and does not yet have sufficient expertise in the minutiae of election politics. Those who are experts in electoral politics have a hard time entering the cryptocurrency space, and there are many reverse forms of cryptocurrency that are not always correct, especially in the political field. But what happened this year is that in the cryptocurrency space, prediction market users who correctly predicted a Biden victory saw their capital increase by 18%, while those who incorrectly predicted a Trump victory lost their capital by 100% (or at least is part of their bet).
Therefore, there will be selection pressure. After ten rounds of such forecasts, good forecasters will have more capital to wager and poor forecasters will have less capital to wager. This does not rely on anyone "getting wiser" or "learning their lessons" or any other assumptions about the human ability to reason and learn. As a result of choice dynamics alone, over time players who are good at making the right guesses will dominate the ecosystem.
It is worth noting that the prediction market performs better than the stock market in this respect: the "upstarts" in the stock market are often lucky with a thousand-fold return, which adds a lot of noise to the signal, but in the prediction market, the price is Limits between 0 and 1 limit the impact of any single event.
2. Better players and better technology
Third, the demonstrations we have seen of prediction markets functioning correctly will ease the concerns of participants. Users will see that the Augur oracle gives correct output even in very controversial situations. People from outside the crypto industry will see that the process works and be more inclined to participate. Perhaps even Nate Silver himself will use some DAI and use other prediction markets like Augur, Omen, Polymarket, etc. to supplement his income in 2022 and beyond.
in conclusion
Fourth, prediction market technology itself can be improved. Here's a market design suggestion I've come up with myself that can be capital efficient while betting on many unlikely events, helping to prevent unreasonably high probabilities for unlikely outcomes. Other ideas are sure to pop up, and I look forward to seeing more experiments in this direction.
in conclusion
With the first incredibly direct experimentation with prediction markets and how they collide with the complexities of individual and social psychology, it shows a lot about how market efficiency works in practice and what its limitations are , and what can be done to improve it.
