IOSG Ventures: Does the prediction market need Web3?

Author: Sid, IOSG Ventures

Special thanks to Aravind Menon for the insights. This article is original content from IOSG and is intended for industry learning and exchange purposes only, and does not constitute any investment advice. Please cite the source if you wish to reference it, and contact the IOSG team for authorization and reprint guidelines.

Lifecycle of a Bet

Investopedia states: "A prediction market is a market where people can trade contracts based on the outcomes of future unknown events." Essentially, it is a betting/gambling market. To better understand the betting market, let's break down the lifecycle of a bet:

Belief

In the belief stage, a prediction is just an opinion. When a person turns their opinion into a bet with money, they can earn a return if the outcome supports that belief. Beliefs are formed through the interaction of various complex factors such as cognition, social influences, emotions, and the environment. Opinions can arise from immediate beliefs or thoughtful considerations, and because there is no monetary loss for the person expressing the opinion, opinions are given more freely.

Betting

There are two scenarios for the creation of a bet:

• Wanting to profit from one's own belief

• Having a very attractive outcome that is unrelated to the belief

The first type of bet may stem from a calculated opinion, while the second comes from a "small bet, big win" attitude.

For any contract to succeed, there needs to be both a side betting for and against:

• You bet $50 on Chelsea to win a match, and someone (or many people) needs to be willing to bet a total of $50 against Chelsea winning (assuming odds are 50/50)

• In margin trading on GMX, traders open a long position, while GLP is the counterparty

• Casino games like roulette and blackjack require a "house" as the counterparty

Sometimes, to attract counterparty participation, we need to take some incentive measures, as the outcomes of events are not always equally likely to occur. These incentives can take various forms, such as odds, bond curves in AMMs (Automated Market Makers), or even funding rates in perpetual/margin trading platforms.

The structural design of market predictions becomes more complex when focusing on specific types of outcomes. Sports betting, for example, requires unique odds settings because almost no two events will have nearly the same outcome. Additionally, each major event (e.g., the outcome of a league championship) may involve many smaller events (the results of each league match), further increasing complexity.

In prediction events, it is also necessary to execute contracts correctly. What if your opponent refuses to pay? This is why derivatives are essentially legally enforceable contracts. On the blockchain, contracts can be executed trustlessly based on outcomes.

Thus, to place a bet, it is necessary to:

• Have an event (or non-event) occur and publish the event/game contract

• Ensure enough participants have opinions on these events (maker demand: market participants provide market orders)

• Ensure these participants have counterparties (taker demand: market participants execute existing market orders)

• Ensure settlement

• Ensure no market manipulation

Outcome

"Gambling games promote the 'illusion of control': that is, gamblers believe they can exert skill over outcomes that are actually defined by chance." -- Dr. Luke Clark The outcome is the conclusion of the event bet. Once the outcome is determined, the bet is complete.

Do Prediction Markets Need Web3?

Let's examine the necessity of Web3 based on the criteria for creating gambling markets mentioned above:

Event/Game Creation

Aside from permissionless event publishing, there is no clear blockchain use case here. Permissionless posting is a flaw rather than a feature, as it creates high redundancy for the same event, worsening the experience for bettors. The creation of bets can be based on an event or can create games like on-chain roulette or blackjack. (Permissionless posting refers to anyone being able to publish information or trade without centralized review or permission)

Events can also be price discovery. We have seen prediction markets for unreleased tokens on Aevo, which provide a good indicator of the market's view on token prices.

Parcl is also creating a prediction market for better price discovery in real estate. It provides homeowners with an approximate value of their house and offers a budget range for buyers intending to purchase real estate in a particular city.

The use case for price discovery is also a function of liquidity in event contracts, which is why the next section is important.

Maker Demand

The blockchain cannot control maker demand, which is entirely driven by offline behaviors such as marketing or gaming built into the product.

Companies focused on price discovery must strive to generate as much maker trading volume as possible to achieve the most accurate price for specific assets.

Counterparty

Now we enter an interesting topic. Counterparties can be incentivized to gamble through attractive odds, especially when the outcome of the event is nearly certain. In the image below, it can be seen that due to a huge mismatch in the Polymarket order book, it is possible to win $200 with a $0.50 bet.

One way is to have each market operate as an independent market running on Balancer AMMs, like Augur Turbo. Here, the LPs (liquidity providers) act as counterparties for different markets. While this structure avoids over-reliance on odds calculations (or acquisition), it worsens the experience of publishing prediction events.

For price discovery order books like Aevo, if there is no liquidity, the platform sometimes has to act as the counterparty itself. This is not ideal, especially when the market bottom is unknown.

Another method is to create a counterparty LP pool like "The House." This is what Azuro and WINR have done. There is a liquidity pool that acts as the counterparty for bettors. Parcl has a USDC liquidity pool that serves as the counterparty for traders betting long or short on real estate prices in different cities.
Both of these protocols have proven their effectiveness: Revenue generated by LPs on Azuro on Polygon (Source: Dune)

The value of WINR's LP token (WLP) has increased from $1 to about $1.27 (if LP started around July 1, 2023, indicating a 27% return) (Source: Dune)

These models demonstrate some good product-market fit, where the front end only needs to focus on bettors placing bets on the platform without managing the order book or making trade-offs brought by AMMs. You can think of these models as Uniswap v4, where different front ends use the underlying liquidity (similar to hooks). The WINR protocol has a casino betting front end and another margin trading protocol offering up to 1000x leverage, ensuring high pool utilization but potentially posing significant risks to the pool.

Ensuring Settlement

Once an event is completed, the bets need to be settled. In an AMM structure, everything is on-chain and settled on the contract. For the Polymarket order book model, the order book is maintained off-chain. If necessary, Polymarket can block withdrawals. For Azuro's front end, similar to Bookmaker.xyz, no deposits are required. Each bet is treated as an independent transaction. The only off-chain component is the calculation of odds and the data source.

Ensuring No Manipulation

If there is a centralized data provider, and this data source is manipulated by the provider, it could adversely affect the outcomes for market makers and takers. This is one of the main reasons most Web3 prediction markets use oracle systems like Chainlink. Using oracles involves trade-offs between latency and data integrity. When choosing oracles, platforms can choose between first-party and third-party oracles, which involves trade-offs in latency. In fast-moving events, whether there is a delay is a very important influencing factor.

In casino games, the completeness of randomness is crucial, and its fairness cannot be affected by its source.

Chainlink and other oracles like Supra and Pyth minimize the possibility of manipulation through aggregation, but in a vast market, the authenticity and reliability of data sources remain a concern. These oracle systems strive to provide reliability by aggregating multiple data sources, reducing the risk of single points of failure, thereby protecting the market from undue manipulation. Nevertheless, ensuring the authenticity of data sources and preventing manipulation remains an ongoing challenge in prediction markets.

Existing Applications of Success and Failure

When we observe the crypto market alongside prediction markets, a comparatively successful example is the use of cryptocurrencies as assets for betting on platforms like Stake.com and Rollbit. (The blue numbers are the predicted numbers)

Although applications like Polymarket have achieved some success, it is not a platform that can maintain consistent trading volume due to the significant gap between the event environment and the platform. Source: Dune

The product-market fit (PMF) of cryptocurrencies with prediction markets has begun to emerge in "House" pool systems like Azuro and WINR. An obvious application scenario is that new front ends focused on specific types of prediction markets only need to focus on the demand side. They can leverage systems like Azuro and WINR, which in turn provide top-tier yields for stablecoin holders (projected annual yields of 40-60% at the current rate).

In most countries, regulations on gambling applications and online casinos are very strict. Protocols like Azuro and WINR may also face lower regulatory pressure than companies like Rollbit.

The level of participation provided by the front end translates directly into the level of participation in the crypto market. Currently, there are no fully permissionless and trustless crypto prediction markets.

What we look forward to seeing is the success that applications like Parcl may achieve, bringing transparency to a relatively illiquid asset class. From a fundamental perspective, it seems to have the right structure to achieve its price discovery goals.

The main application scenarios for Web3 include the counterparty pool structure supporting the construction of various prediction markets and the successful application of prediction markets for better price discovery.

As the market capitalization of cryptocurrencies grows and more people have disposable capital on-chain, the prediction market industry may become profitable or at least very useful.