Prediction Markets: Bottlenecks and the Next Major Unlocks

Summary of Content

How Prediction Markets Work
Bottlenecks Preventing Greater Adoption of Prediction Markets
- Supply Side
- Demand Side
Solutions to these problems
- Supply Side
- Demand Side
Other ways to increase adoption

Prediction Markets: Bottlenecks and the Next Major Unlocks

Augur, the first on-chain prediction market, was among the earliest applications ever launched on Ethereum. The aspiration was to allow anyone to bet any size on anything. Plagued with issues, Augur’s vision fell short of reality many years ago. Lack of users, poor settlement UX, and high gas fees led the product to shut down. Since then we have come a long way: blockspace is cheaper, and order book designs are much more efficient. Recent innovations solidify that crypto’s permissionless and open-source nature allows for an unconstrained and global liquidity layer in which anyone can become a market participant through liquidity provisioning, market creation, or betting.

Polymarket has been the emergent market leader, with approximately $900 million in volume to date, and SX Bet has amassed $475 million thus far. Despite this, there is substantial room for growth, especially when compared to the massive scale of sports betting, a subcategory of traditional prediction markets. Sportsbooks in the United States alone have handled over $119 billion in 2023. The volume figure is much more prominent when considering land-based and online sports betting volume in all other countries combined and other types of prediction markets such as politics and entertainment.

This article aims to break down how prediction markets work, the current bottlenecks that would need to be addressed for further adoption at scale, and some ways we think they can be solved.

How do prediction markets work?

There are several ways to design prediction markets, and most can be divided into two categories: order book models and concentrated AMM (Automated Market Maker) models. Our thesis is that order book models are the more optimal design choice because they allow for better price discovery, enable maximum composability, and ultimately lead to volume at scale.

For order book models, each market has only two possible pre-defined outcomes: Yes and No. Users trade these outcomes in the form of shares. At the time of market settlement, the correct share is worth $1.00, and the incorrect one will be worth $0.00. Before market settlement, the price of these shares could trade anywhere between $0.00 and $1.00.

In order for share trading to occur, liquidity providers (LPs) must exist; in other words, they must provide bids and asks (quotes). These LPs are also referred to as market makers. A market maker provides liquidity in exchange for a small profit in the spread.

An example of a specific market: if there is an even chance something will happen, such as a coin toss landing on heads, both “Yes” and “No” shares should theoretically trade at $0.50. However, much like in any financial market, there tends to be a spread and thus slippage. If I wanted to buy “Yes” shares, my execution price might end up being closer to $0.55. This is because my counterparty, a liquidity provider, is purposely overpricing the true odds to make a potential profit. The counterparty might also be selling “No” shares for $0.55. This spread of $0.05 per side is compensation that the liquidity provider gets for offering quotes. Spread is driven by implied volatility (expectation of price movement). Prediction markets essentially have guaranteed realized volatility (actual price movement) simply due to the design where shares must eventually reach a pre-destined value of $1 or $0 by some pre-defined date.

To illustrate a market maker scenario for the coin toss prompt above:

1. A maker sells 1 “Yes” share for $0.55 (this is equivalent to buying a “No” share at $0.45)

2. A maker sells 1 “No” share for $0.55 (this is equivalent to buying a “Yes” share at $0.45)

3. The maker now has 1 No Share and 1 Yes Share that it paid $0.90 for in total

4. Regardless of whether the coin lands Heads, the maker will redeem $1.00, earning a $0.10 spread

The other primary way of settling prediction markets is through concentrated AMMs, used by both Azuro and Overtime. For the sake of this article, we will not be touching on these models too much, but the analogy in DeFi is GMX v2. Capital is pooled, and acts as the single counterparty to platform traders, and the pool relies on external oracles to price the odds that are offered to users.

What are the current bottlenecks for prediction markets?

Prediction market platforms have been around and heard about for a long enough time that escape velocity would have happened already if there was true product market fit. Current bottlenecks boil down to the simple conclusion of a lack of interest on both the supply (liquidity provider) and demand (bettor) side.

On the supply side, problems include:

Lack of liquidity due to volatility: Polymarket’s most popular markets tend to be conceptually novel ones with scarce relevant historical data, making outcomes challenging to predict and price accurately. For instance, predicting whether or not a CEO, such as Sam Altman, will return to his position after rumors of potential AGI mishandling is difficult because no past events closely mirror this scenario. Market makers will put up larger spreads and less liquidity on uncertain markets in order to compensate for implied volatility (i.e. wild price action in the Sam Altman CEO market, where consensus flipped 3 times in less than 4 days). This makes it less appealing to whales who want to bet in size.

Lack of liquidity due to few subject matter experts: Although hundreds of market makers earn rewards daily on Polymarket, many longer-tail markets lack liquidity simply because there are not enough parties equipped with specialized knowledge who also want to market make. Example markets include “Will x celebrity get arrested or charged for y?” or “When will celebrity x tweet?”. This will change over time as more prediction market types are introduced, data becomes more abundant, and makers become more specialized.

Information Asymmetry: Since makers provide bids and asks that any taker can trade against at any time, the latter party has the advantage of making positive EV bets when they acquire advantageous information. In DeFi markets, these types of takers can be referred to as toxic flow. Arbitrageurs on Uniswap are good examples of toxic takers because they use information edges to continuously extract profit from liquidity providers.

One Polymarket market, “Will Tesla announce a Bitcoin purchase before March 1, 2021?” saw a user purchase $60,000 worth of Yes shares at ~33% odds. This market was the only market the user had ever participated in, and it can be assumed that this user had advantageous information. Legalities aside, there is no way the maker who offered this quote could have known that the taker/bettor had this advantageous information at the time that it did, and even if the maker had originally set their odds at 95%, the taker probably still would have placed the bet, because the true odds were 99.9%. This results in a guaranteed loss scenario for the maker. In prediction markets, it is hard to predict when toxic flow will occur and at what size, therefore making it harder to provide tight spreads and deep liquidity. Makers need to price the risk of toxic flow occurring at any time.

On the demand side, the main issue is:

Lack of a leverage vehicle: Without a leverage vehicle, the value proposition of prediction markets to retail users is relatively inferior to that of other crypto speculation vehicles. Retail wants to make “generational wealth,” which is more likely achieved on memecoins than betting on a prediction market with a capped upside. Since its inception, betting on $BODEN and $TRUMP early has yielded far more upside than Biden or Trump Yes shares for winning the presidency.

Lack of exciting short-term markets: Retail bettors have no interest in placing bets that settle months from now, and this conclusion can be proven in the sports betting world, where a lot of retail volume occurs on live betting (super short-term) and daily events (short-term) nowadays. Not enough short-term markets appeal to a mainstream audience, at least not yet.

What solutions are there to these problems? How can we increase volume?

On the supply side, the first two problems, related to lack of liquidity due to volatility and lack of liquidity due to few subject matter experts, will naturally lessen over time. As the volume grows across various prediction markets, the number of specialized makers and those with higher risk tolerance and capital will also grow.

However, instead of waiting for these problems to lessen over time, the lack of liquidity can be addressed head-on through liquidity coordination mechanisms, originally invented in the DeFi derivatives space.  The idea is to allow passive stablecoin depositors to earn yield through vaults, which deploy market-making strategies across different markets. This vault would act as the primary counter-party against traders. GMX was the first to do this through a pool-style liquidity provisioning strategy that relied on oracles for pricing, and Hyperliquid was the second notable protocol to deploy a native vault strategy, but with the kicker that the liquidity was provided on a CLOB. Both of the vaults have been profitable over time because they are able to act as the counterparty to mostly non-toxic flow (retail users that tend to lose money over time).

Native vaults make it easy for protocols to bootstrap liquidity themselves without relying on others. They also make long-tail markets more appealing; one reason Hyperliquid became so successful is that newly listed perpetual assets contain tons of liquidity from Day 0.

The challenge with building a vault product for prediction markets is to prevent toxic flow. GMX prevents this by attaching high fees to their trades. Hyperliquid implements market maker strategies with large spreads, has a 2 block delay for taker orders to give time to makers to adjust their quotes, and prioritizes maker order cancels within a block. Both protocols create an environment where toxic flow does not enter because they can find better price execution elsewhere.

In prediction markets, toxic flow can be prevented by providing deep liquidity at wider spreads, selectively providing liquidity to markets less susceptible to information advantages, or employing a sharp strategist with access to information advantages.

In practice, a native vault could deploy $250,000 of additional liquidity as a bid at 53 cents and as an ask at 56 cents. Wider spreads help increase potential vault profits because a user would be betting while accepting worse odds. This is opposed to placing quotes at 54 and 55 cents, where the counterparty trader might be an arbitrageur or sharp looking for good prices. This market is also relatively less susceptible to information asymmetry issues (less insider info and insights usually come out to the public relatively quickly), so the expectation of toxic flow is lower. The vault could also use information oracles that provide insight into future line movement, such as pulling odds data from other betting exchanges or gathering information from top political analysts on social media.

The result is deeper liquidity for bettors, who are now able to bet larger sizes with less slippage.

There are several approaches to solving, or at least reducing, the information asymmetry issue. The first few are around order book design:

Gradual Limit Order Book: One way to combat toxic flow is by increasing the price based on the combined speed and size of an order. If a buyer is certain that an event will occur, the logical strategy would be to buy as many shares as possible at prices below $1.00. Additionally, purchasing quickly would be sensible in the event that the market eventually accesses the advantageous information.

Contro is implementing this GLOB idea and is launching as an interwoven rollup on Initia.

If the Tesla $BTC market occurred on a GLOB model, the taker would have had to pay much more than 33% per “Yes” share, as there would be “slippage” given the combined speed (one clip) and size (huge) of the order. He still would have made a profit regardless of slippage because he knew his “Yes” shares would eventually become worth $1, but it at least contains the maker’s loss.

One could argue that the taker still could have incurred little slippage and paid close to 33% per “Yes” share if they had just performed a DCA strategy over a long period of time, but in this scenario, it at least gives some time to the market maker to withdraw its quotes from the book. A maker could withdraw for several reasons:

1. it suspects that there’s toxic flow because there’s such a large taker order coming in

2. it is convinced that there is toxic flow because it checked the taker’s profile and saw that it had never made a bet before

3. it wanted to re-balance its inventory and no longer wanted to be too one-sided given how many “Yes” shares it was selling and thus “No” shares it was accumulating – perhaps the maker originally had $50,000 worth of orders on the ask side at 33% and $50,000 worth of orders on the bid side at 27% – and its original goal is not to be directionally biased, but rather, neutral so it could earn profits through symmetric liquidity provisioning

A winner’s rake: There are many markets whereby a portion of the profits of those with advantageous information get re-distributed. The first example is in peer-to-peer web2 sportsbooks, specifically Betfair, where a fixed percentage of a user’s net winnings will be re-distributed back to the company. Betfair’s rake actually depends on the market itself; on Polymarket, it could be rational to charge a higher rake on net winnings for markets that are more novel or long-tail.

This re-distribution concept exists similarly 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 who participated in the transaction, which could either be the liquidity providers or the user making a trade. Orderflow auctions have seen plenty of PMF to date, and CowSwap * is pioneering this category through MEVBlocker.

Static or dynamic taker fees: There are currently no taker fees on Polymarket. If this were implemented, proceeds could be used for liquidity provisioning rewards in markets with high volatility or greater susceptibility to toxic flow. Alternatively,  taker fees could be set higher on long-tail markets.

On the demand side, the best way to solve the lack of upside is to create a mechanism that allows for it. In sports betting, parlays have become increasingly popular for retail bettors because they offer a chance at “winning big.” A parlay is a type of bet combining multiple individual wagers into one bet. For the parlay to win, all of the individual wagers must win.

There are three main ways to increase upside for users in crypto-native prediction markets:

• Parlays

• Perpetuals

• Tokenized leverage

Parlays: This is technically infeasible to implement on Polymarket’s books because upfront capital is needed for bets, and the counterparty for each market is different. In practice, a new protocol can source odds from Polymarket at any given point to price any parlay bet’s odds and act as the single counterparty for parlays.

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

These bets give limited upside if placed on their own, but when combined into a parlay, the implied return skyrockets to ~1:650,000, meaning that the bettor could win $6.5 million if every bet is correct. It is not too hard to envision how parlays could gain PMF amongst crypto users:

• The cost to participate is cheap, and you can put up very little to win a lot

• Sharing parlay slips would become viral on Crypto Twitter, especially if someone wins big, which creates a feedback loop with the product itself

Supporting parlays presents challenges, namely counterparty risk (what happens when multiple bettors win big parlays at the same time) and odds accuracy (you don’t want to provide bets where you are underestimating the true odds). Casinos have tackled the challenge of offering parlays in the sports world and it has become by far the most profitable component of sports betting. The profit margins are ~5-8x higher than offering single-market bets, even if some bettors get lucky and win big. The other added benefit of parlays is that there’s relatively less toxic flow compared to single-game markets. The analogy here is: why would a professional poker player who lives and dies by expected value put money into lottery tickets?

SX Bet, a web3 sports betting appchain, launched the world’s first peer-to-peer parlay betting system and has done $1M in parlay volume in the past month. When a bettor “Requests Parlay”, SX creates a private virtual order book for the parlay. Programmatic market-makers listening through the API will then have 1 second to offer liquidity on the bet. It would be interesting to observe increased liquidity and traction around non-sport parlays.

Perpetual Prediction Markets: This concept was explored briefly in 2020 when former leading exchange FTX offered perpetuals for American Election outcomes. You could long the price of $TRUMP and, if he won the American Election, each share could be redeemed for $1. FTX had to change margin requirements as the odds of him actually winning changed. Creating a perpetual mechanism for markets as volatile as prediction markets creates a lot of challenges for collateral requirements because prices can be worth $0.90 one second and $0.1 the next. Thus, there might not be enough collateral to cover the losses of someone who longed the wrong way. Some of the order book designs explored above could help compensate for the fact that prices could change so quickly. The other interesting part of the FTX $TRUMP market is that we can reasonably assume that Alameda was the main market maker in these markets and that without natively deployed liquidity, the books would have been too thin for substantial volumes to occur. This highlights how valuable a native liquidity vault mechanism might be for prediction market protocols.

Both LEVR Bet and SX Bet are currently working on perpetual sports betting markets. One plus about leverage on sports is that the price of “Yes or “No” shares will fluctuate less violently, at least most of the time. For example, a player making a basketball shot might boost a team’s odds of winning the game from 50% to 52%, as on average a team might make 50 shots per game. The 2% boost on any given shot is a manageable increase from a liquidation and collateral requirement perspective. Offering perpetuals at the end of a game is a different story, as someone might hit a “game-winner” and the odds could flip from 1% to 99% in half a millisecond. One potential solution is to only allow for leveraged betting up to a certain point because any event after would change the odds too much. The feasibility of perpetual sports betting also depends on the sport itself; a single hockey goal changes the expected game outcome much more than a single made basketball shot.

Tokenized Leverage: A lending market that allows users to borrow against their prediction market positions, especially ones of long-term duration, may increase the volume among pro-traders. This can also lead to more liquidity, as market makers could borrow against positions in one market to make in another. Tokenized leverage probably wouldn’t be an interesting product to retail bettors unless there was an abstracted looping product, such as the ones that gained traction for Eigenlayer. The entire market is probably too immature for abstraction layers like this to exist just yet, but these types of looping products will eventually come.

Aside from the pure supply and demand aspects, there are other minute ways to increase adoption:

From a UX perspective: Switching the settlement currency from USDC to a yield-bearing stablecoin would increase participation, especially in long-tail markets. This has been discussed on Twitter a few times; holding positions for markets that expire at the end of the year has a significant opportunity cost (e.g. earning 0.24% APR by betting on Kanye West to win the presidency versus earning 8% APR on AAVE).

Moreover, increased gamification that aims to improve retention can really help drive more users in the long run. Simple things such as “daily betting streak” or “daily contests” have worked well in the sports betting industry.

A few sector-level tailwinds will also increase adoption in the near future: The combination of growing virtual and onchain environments will unlock a whole new level of speculative demand because the number of short-term events will eventually be unlimited (think AI/computer-simulated sports), and the level of data will be abundant (this makes it easier for market makers to price outcome odds). Other interesting crypto-native categories include AI gaming, onchain gaming, and general onchain data.

Accessible data will lead to an increased level of betting activity from non-humans, more specifically autonomous agents. Omen on Gnosis Chain * is spearheading the idea of AI agent bettors. Since prediction markets are a game where the outcome is defined, autonomous agents can become increasingly skilled at calculating expected value, likely to a precision far better than a human. This mirrors the idea that AAs would probably have a harder time predicting which memecoins will take off because there is more of an “emotional” element to what makes them successful, and humans are much better at feeling emotions than AAs at the moment.

In summation, prediction markets are a fascinating user product and design space. As time goes on, the vision of allowing anyone to bet any size on anything will become a reality. If you’re building something in the space, whether it be an entirely new protocol, a liquidity coordination platform, or a new leverage mechanism, please reach out! I am an avid user and would love to give feedback.

Thank you 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 the feedback and for reviewing this piece!

Disclaimers:

*Denotes a 1kx portfolio investment.

This article is for general information purposes only and should not be construed as or relied upon in any manner as investment, financial, legal, regulatory, tax, accounting, or similar advice. Under no circumstances should any material at the site be used or be construed as an offer soliciting the purchase or sale of any security, future, or other financial product or instrument. Views expressed in posts are those of the individual 1kx personnel quoted therein and are not the views of 1kx and are subject to change. The posts are not directed to any investors or potential investors, and do not constitute an offer to sell or a solicitation of an offer to buy any securities, and may not be used or relied upon in evaluating the merits of any investment. All information contained herein should be independently verified and confirmed. 1kx does not accept any liability for any loss or damage whatsoever caused in reliance upon such information. Certain information has been obtained from third-party sources. While taken from sources believed to be reliable, 1kx has not independently verified such information and makes no representations about the enduring accuracy or completeness of any information provided or its appropriateness for a given situation. 1kx may hold positions in certain projects or assets discussed in this article.