1kx: What are the bottlenecks and breakthrough points for prediction markets like Polymarket?

Original Title: Prediction Markets: Bottlenecks and the Next Major Unlocks

Original Author: Mikey 0x, 1kx

Compiled by: Elvin, ChainCatcher

Content Summary

1. How prediction markets work

2. Bottlenecks hindering the wider adoption of prediction markets

• Supply side
• Demand side

1. Solutions

• Supply side
• Demand side

1. Other methods to increase adoption

Prediction Markets: Development Bottlenecks and the Next Key Opportunities

Augur, as the first on-chain prediction market, was one of the earliest applications launched on Ethereum. Its vision was to allow anyone to bet any size on anything. However, due to numerous issues, Augur's vision was not realized years ago. A lack of users, poor settlement user experience, and high gas fees led to the product shutting down. However, since then, we have made significant progress: block space is cheaper, and order book designs are more efficient. Recent innovations have solidified the permissionless and open-source nature of cryptocurrencies, allowing anyone to participate in the global liquidity layer by providing liquidity, creating markets, or placing bets.

Polymarket has become the market leader, with a trading volume of about $900 million to date, while SX Bet has accumulated $475 million so far. Nevertheless, compared to the enormous scale of traditional sports betting, there is still significant room for growth. In the U.S. alone, sports bookmakers handled over $119 billion in trading volume in 2023. When considering all other countries' offline and online sports betting volumes and other types of prediction markets, such as politics and entertainment, this figure becomes even more striking.

This article aims to break down how prediction markets work, the current bottleneck issues that need to be addressed, and some methods we believe can solve these problems.

How do prediction markets work?

There are several ways to design prediction markets, most of which can be categorized into two types: order book models and centralized AMM models. Our view is that order book models are the superior design choice because they allow for better price discovery, achieve maximum composability, and ultimately lead to scalable trading volumes.

Order Book Model

In the order book model, each market has only two possible predefined outcomes: Yes (Y) and No (N). Users trade these outcomes in the form of shares. At market settlement, the correct shares are worth $1, while the incorrect shares are worth $0. Before market settlement, the prices of these shares may trade between $0 and $1.

For share trading to occur, there must be liquidity providers (LPs); in other words, they must provide buy and sell orders (quotes). These LPs are also known as market makers. Market makers provide liquidity in exchange for small profits from the spread.

For a specific market example: If the probability of something happening is even, such as the outcome of a coin toss being heads, then theoretically, the "Yes" and "No" shares should trade at $0.50. However, like any financial market, there is usually a spread, leading to slippage. If I want to buy "Yes" shares, my execution price may end up close to $0.55. This is because my counterparty, a market maker, intentionally overestimates the true odds to earn potential profits. The counterparty may also sell "No" shares at $0.55. A $0.05 spread on each side compensates the market maker for providing quotes. The spread is driven by implied volatility (the expected price movement). Prediction markets inherently have guaranteed volatility (actual price movement), simply because shares must ultimately reach $1 or $0 on a predetermined date.

Example of a market maker scenario:

• The market maker sells 1 "Yes" share at $0.55 (equivalent to buying 1 "No" share at $0.45)
• The market maker sells 1 "No" share at $0.55 (equivalent to buying 1 "Yes" share at $0.45)
• The market maker now holds 1 "No" share and 1 "Yes" share, having paid a total of $0.90
• Regardless of whether the coin is heads or tails, the market maker will redeem $1, earning a $0.10 spread
  Another major method of settling prediction markets is through centralized AMMs, which Azuro and Overtime both use. This article will not delve further into these models, but the analogy in DeFi is GMX v2. Capital is pooled, acting as the sole counterparty for platform traders, with the liquidity pool relying on external oracles for user pricing.

What are the current bottlenecks in prediction markets?

Prediction market platforms have existed and been discussed for long enough that if they truly met product-market fit, escape velocity would have already occurred. The current bottlenecks can be simply summarized as a lack of interest from both the supply (liquidity providers) and demand (bettors) sides.

Issues on the supply side include:

1. Insufficient liquidity due to volatility: Polymarket's most popular markets are often conceptually novel markets that lack relevant historical data, making outcomes difficult to predict and price accurately. For example, predicting whether a CEO, like Sam Altman, "will return to his position after rumors of potential AGI mishandling" is challenging because there are no past events that closely resemble this situation. Market makers will set larger spreads and less liquidity in uncertain markets to compensate for implied volatility (i.e., the price of the Sam Altman CEO market fluctuated wildly, with consensus flipping three times in less than four days). This makes it less appealing for large players wanting to place significant bets.

2. Insufficient liquidity due to a lack of subject matter experts: Although hundreds of market makers earn rewards daily on Polymarket, many long-tail markets lack liquidity due to a shortage of knowledgeable participants. For example, markets like "Will a certain celebrity be arrested or charged for something?" or "When will a celebrity tweet?" As more types of prediction markets are introduced, data becomes richer, and market makers become more specialized; this situation will change over time.

3. Information asymmetry: Because the buy and sell quotes provided by market makers can be traded by any taker at any time, the latter has the advantage of making positive expected value bets when they acquire favorable information. In DeFi markets, these types of takers can be referred to as harmful traffic. Arbitrageurs on Uniswap are a good example of harmful takers, as they continuously extract profits from liquidity providers by leveraging their information advantage.

In a Polymarket market, "Will Tesla announce the purchase of Bitcoin before March 1, 2021?" a user bought "Yes" shares worth $60,000 at about 33% odds. This market was the only one the user had participated in, suggesting that this user had favorable information. Regardless of legality issues, the market maker providing quotes at that time could not know that the taker/bettor possessed such favorable information, even if the market maker initially set the odds at 95%, the taker might still bet because the true odds were 99.9%. This puts the market maker in a certain loss situation. In prediction markets, it is difficult to predict when harmful traffic will occur and how large it will be, making it harder to provide tight spreads and deep liquidity. Market makers need to price in the risk of harmful traffic that could occur at any time.

The main issues on the demand side are:

1. Lack of leverage tools: Without leverage tools, prediction markets have relatively low appeal to retail investors compared to other crypto speculation tools. Retail investors want to create "generational wealth," which is more likely to be achieved by betting on memecoins than on capped prediction markets. For example, betting early on $BODEN and $TRUMP brought more upside than betting on "Yes" shares for Biden or Trump winning the presidential election in prediction markets.

2. Lack of stimulating short-term markets: Retail bettors are not interested in bets that settle months later, a conclusion supported by the sports betting world, where much retail trading now occurs in live betting (ultra-short-term) and daily events (short-term). Not enough short-term markets attract mainstream audiences, at least not at present.

What are the methods to solve these issues? How do we increase trading volume?

On the supply side, the first two issues related to insufficient liquidity due to volatility and insufficient liquidity due to a lack of expertise will naturally decrease over time. As trading volumes in various prediction markets grow, the number of professional market makers and those with higher risk tolerance and capital will also increase.

However, rather than waiting for these issues to ease over time, it is better to directly address the liquidity shortage issue through liquidity coordination mechanisms initially invented in the DeFi derivatives space. The idea is to allow passive stablecoin depositors to earn yields through a treasury that deploys market-making strategies across different markets. This treasury will act as the primary counterparty for traders. GMX was the first protocol to achieve this through a pool-based liquidity provision strategy relying on oracles for pricing, while Hyperliquid is the second well-known protocol to deploy a native treasury strategy, characterized by liquidity being provided on a CLOB. Both of these treasuries have been profitable over time because they can act as counterparties to most harmful traffic (which often tends to lose money over time).

Hyperliquid's treasury PNL has been growing over time
Native treasuries allow protocols to easily bootstrap liquidity without relying on others. They also make long-tail markets more attractive; one reason Hyperliquid has been so successful is that newly listed perpetual assets have included substantial liquidity from day one.
The challenge of building treasury products for prediction markets lies in preventing harmful traffic. GMX addresses this by attaching high fees to its trades. Hyperliquid employs a market-making strategy with large spreads and a two-block delay on taker orders to give market makers time to adjust their quotes and prioritize canceling market maker orders within one block. Both protocols create an environment where harmful traffic does not enter because they can find better price execution elsewhere. In prediction markets, harmful traffic can be prevented by providing deep liquidity with wider spreads, selectively providing liquidity to markets less susceptible to information advantages, or employing savvy strategists with information advantages.

In practice, a native treasury can deploy an additional $250,000 in liquidity, bidding at $0.53 and asking at $0.56. A wider spread helps increase potential treasury profits because users accept worse odds when betting. This is different from setting quotes at $0.54 and $0.55, where the counterparty might be an arbitrageur or savvy trader looking for a good price. This market is relatively less affected by the information asymmetry issue (less insider information and insights that typically get disclosed to the public quickly), thus having lower expectations of harmful traffic. The treasury can also utilize information oracles that provide insights into future line movements, such as obtaining odds data from other betting exchanges or collecting information from top political analysts on social media.

The result is deeper liquidity for bettors, who can now place larger bets with smaller slippage.

There are several methods to address or at least reduce the information asymmetry issue. The first few are related to order book design:

1. Gradual Limit Order Book (GLOB): One way to combat harmful traffic is to improve pricing by combining the speed and size of orders. If buyers are confident that an event will occur, the logical strategy would be to buy as many shares as possible at a price below $1. Additionally, if the market ultimately acquires favorable information, quick purchases are also wise.

Contro is implementing this GLOB concept and launching it as a cross-rollup on Initia.

If the Tesla $BTC market occurs on the GLOB model, the taker will have to pay more than 33% for "Yes" shares because the "slippage" generated by considering the combination speed (a fragment) and size (huge) will be factored in. They will still profit because they know "Yes" shares will ultimately appreciate to $1, but this at least accounts for the market maker's losses.
One might argue that if the taker simply executes a long-term DCA strategy, they can still tolerate minimal slippage and pay close to 33% for each "Yes" share, but in this case, at least it gives the market maker some time to withdraw their quotes from the book. Market makers might withdraw for several reasons:

• Suspecting harmful traffic due to such a large taker order coming in
• Being certain of harmful traffic because they check the taker's profile and find they have never placed a bet before
• Wanting to rebalance their inventory and no longer wishing to be overly one-sided due to the number of "Yes" shares they are selling and the "No" shares they have accumulated—perhaps the market maker initially had a $50,000 order on the ask side at 33% odds and a $50,000 order on the bid side at 27% odds—their initial goal was not directional bias but neutrality to earn profits through symmetric liquidity provision

2. Winner's Take: Many markets redistribute a portion of the profits of those with favorable information. The first example is in peer-to-peer web2 sports betting, particularly Betfair, where a fixed percentage of users' net winnings is redistributed back to the company. Betfair's take actually depends on the market itself; on Polymarket, it might be reasonable to charge a higher net winnings take for newer or long-tail markets.
This redistribution concept exists in DeFi in the form of order flow auctions. A backrunning bot captures value from information asymmetry (arbitrage) and is forced to give back to those participating in the trade, which could be liquidity providers or users placing trades. Order flow auctions have seen much PMF to date, with CowSwap* pioneering this category through MEVBlocker.
3. Static or Dynamic Taker Fees: Polymarket currently has no taker fees. If this were implemented, the revenue could be used for liquidity provision rewards in high-volatility markets or markets more susceptible to harmful traffic. Alternatively, higher taker fees could be set for long-tail markets.
On the demand side, the best way to address the lack of upside potential is to create a mechanism that allows it. In sports betting, parlays have become increasingly popular among retail bettors because they offer the chance to "win big." A parlay is a bet that combines multiple individual bets into one. To win a parlay, all individual bets must win.

Users won over $500,000 with an initial bet of $26

In native cryptocurrency prediction markets, there are three main methods to increase users' upside potential:

1. Parlays
2. Perpetual Markets
3. Tokenized Leverage

Parlays: Technically, implementing this on Polymarket's book is not feasible because bets require upfront capital, and the counterparties for each market are different. In practice, a new protocol could obtain odds from Polymarket at any given time, price the parlay bets, and act as the single counterparty for the parlays.

For example, a user wants to bet $10 on the following:

These bets have limited upside when placed individually, but when combined into a parlay, the implied return skyrockets to about 1:650,000, meaning if every bet is correct, the bettor could win $6.5 million. It’s not hard to imagine how parlays could achieve PMF (product-market fit) among crypto users:

• The cost of participation is low; you can put in a little money to win a lot
• Sharing parlay tickets could go viral on Crypto Twitter, especially if someone wins a big prize, creating a feedback loop with the product itself
  Supporting parlays brings challenges, such as counterparty risk (what happens when multiple bettors win large parlays at the same time) and odds accuracy (you don’t want to offer bets where you underestimated the true odds). Casinos have solved the challenges of offering parlays in the sports world, and it has become the most profitable part of sports betting. The margins are about 5-8 times higher than offering single market bets, even if some bettors are lucky enough to win big. Another added benefit of parlays is that there is relatively less harmful traffic compared to single markets. The analogy here is: why would a professional player, who lives off expected value, put money into a lottery?

SX Bet, a web3 sports betting application chain, launched the world's first peer-to-peer parlay betting system and achieved $1 million in parlay trading volume in the past month. When a bettor "requests a parlay," SX creates a private virtual order book for the parlay. Programmatic market makers listening via API then have one second to provide liquidity.

Perpetual Prediction Markets: **This concept was briefly explored in 2020 when the leading exchange FTX offered perpetual contracts for U.S. election outcomes. You could go long on $TRUMP's price, redeeming $1 per share if he won the U.S. election. As his actual winning probability changed, FTX had to adjust margin requirements. Creating perpetual mechanisms for such volatile markets presents many challenges for margin requirements, as prices can swing from $0.90 to $0.10 in a second. Thus, there may not be enough collateral to cover losses for those going long in the wrong direction. Some of the order book designs discussed above can help mitigate the fact of rapid price changes. Another interesting aspect of the FTX $TRUMP market is that we can reasonably assume Alameda was the primary market maker for these markets; without locally deployed liquidity, the order book would be too thin for large trades. This highlights the value of local liquidity treasury mechanisms for prediction market protocols.

LEVR Bet and SX Bet are currently developing perpetual sports betting markets. One advantage of sports betting is that the price fluctuations of "Yes" or "No" shares tend to be smaller, at least most of the time. For example, a player making a shot might increase the team's chances of winning the game from 50% to 52%, as on average, a team might take 50 shots per game. A 2% increase in the odds of any given shot is manageable from a liquidation and margin call perspective. Offering perpetual contracts at the end of a game is another matter, as someone might hit a "game-winning shot," and the odds could flip from 1% to 99% in half a millisecond. One possible solution is to only allow leveraged betting to a certain extent, as any subsequent events could cause the odds to change too drastically. The feasibility of perpetual sports betting also depends on the sport itself; a hockey goal is more likely to change the expected outcome of a game than a basketball shot.

Tokenized Leverage: **A lending market that allows users to borrow against their prediction market positions, particularly long positions, could increase trading volume among professional traders. This could also lead to more liquidity, as market makers could borrow against a market's position to provide liquidity in another market. Tokenized leverage may not be an interesting product for retail bettors unless there is an abstract looping product, like those gaining attention from Eigenlayer. The entire market may still be too immature to support such an abstract layer, but these types of looping products will eventually emerge.

Besides purely supply and demand aspects, there are other minor ways to increase adoption:

From a user experience perspective: Switching the settlement currency from USDC to a yield-bearing stablecoin would increase participation, especially in long-tail markets. This has been discussed several times on Twitter; holding market positions that expire at the end of the year has a significant opportunity cost (e.g., earning a 0.24% annual interest rate by betting on Kanye West to win the presidential election instead of earning 8% on AAVE).

Additionally, increasing gamification aimed at improving retention could genuinely help attract more users in the long run. Simple things like "daily betting streaks" or "daily competitions" have worked well in the sports betting industry.

Some industry-level tailwinds will also increase adoption in the near future: the growth of virtual and on-chain environments will unlock a whole new level of speculative demand, as the number of short-term events will ultimately be infinite (think AI/computer-simulated sports), and the level of data will be rich (making it easier for market makers to price outcomes). Other interesting crypto-native categories include AI gaming, on-chain gaming, and general on-chain data.

Accessible data will lead to an increase in non-human gambling activities, more specifically, an increase in betting activities by autonomous agents. Omen on the Gnosis Chain is a pioneer in the concept of AI agent bettors. Since prediction markets are games defined by outcomes, autonomous agents will become increasingly proficient at calculating expected values, potentially more accurately than humans. This reflects a viewpoint that AI may struggle to predict which memecoin will take off, as there are more "emotional" elements among the factors that make them successful, and currently, humans are better at sensing emotions than AI.

In summary, prediction markets are a fascinating user product and design space. Over time, the vision of allowing anyone to bet any size on anything will become a reality. If you are building something in this space, whether it’s a brand new protocol, a liquidity coordination platform, or a new leverage mechanism, please reach out to us. I am an enthusiastic user and would be happy to provide feedback.

Thanks to Peter Pan, Shayne Coplan, Sanat Kapur, Andrew Young, taetaehoho, Diana Biggs, Abigail Carlson, Daniel Sekopta, Ryan Clark, Josh Solesbury, Watcher, Jamie Wallace, and Rares Florea for their feedback and reviewing this article!

Disclaimer:
* Indicates investment from the 1kx portfolio.