Prediction markets have a natural affinity with crypto. This synergy arises from crypto’s decentralized infrastructure, aligning with the incentives of prediction markets and their value as decentralized truth-finders. In fact, prediction markets indirectly leverage collective intelligence to forecast events, offering a transparent alternative to potentially biased traditional media.

Even @VitalikButerin has recognized their significance as “social epistemic tools,” enabling a clearer public understanding of events.

Source: @VitalikButerin Twitter

Crypto, with its censorship-resistant, transparent, and trustless nature, provides a solid technological foundation for prediction markets. This fusion creates an ecosystem for accurate, minimally-biased, and decentralized real-world predictions. However, a robust, decentralized settlement system remains crucial for their success.

The Rise and Stumbles of “web3” Prediction Markets

Early “web3” prediction markets, such as @AugurProject, faced many hurdles, limited adoption, and ultimately wound down or remained extremely niche products.

The initial approach was towards a purely decentralized model, which theoretically offers one of the highest levels of censorship resistant and tamper-proof results. However, this often comes at the cost of complexity, leading to slower settlement times, and higher transaction fees, at a time when transactions on @ethereum Mainnet were already much more expensive.

Both factors indeed resulted in a significant barrier to entry and denied the protocols a mainstream adoption, especially when centralized platforms offered user-friendly experiences and lower costs.

Moreover, the allure of altcoins’ volatility overshadowed the appeal of prediction markets and regulatory uncertainty further impaired adoption (in 2018, the @CFTC expressed its view that prediction markets are generally prohibited unless they operate on a small scale and serve academic purposes and as consequence of this, for example, Polymarket ended up paying a $1.4 million civil penalty in 2021).

Current Environment and New Prediction Market Models

Over the past eight years, the American general public has grown more polarized than ever before, drawing heightened attention to the political landscape. This division has naturally fueled interest in political betting, particularly as gambling itself has expanded in the U.S. following a key 2018 Supreme Court ruling that paved the way for legal sports betting nationwide. More recently, a Supreme Court decision in favor of @Kalshi has further legitimized political prediction markets.

Source: Dune — Polymarket Monthly Active Traders

The upcoming 2024 U.S. Presidential Election, in particular, holds special significance for the crypto community. After four years of strong opposition from a Democratic White House, many in the crypto space are heavily invested in seeing a pro-crypto figure elected to office. This election is viewed as pivotal for the future of crypto regulation, adding another layer of urgency to prediction markets.

Therefore, let’s look at how the crypto prediction market landscape evolved.

Semi-centralized Prediction Market Models

Semi-centralized models, like BET @DriftProtocol and @azuroprotocol aim for a middle ground between total decentralization and efficiency. The idea is that by sacrificing some degree of decentralization, they can offer faster and more efficient settlements compared to purely decentralized models and often provide a more user-friendly experience.

Both Drift and Azuro exemplify semi-centralized approaches to prediction markets. However, they differ in their specific implementations:

Azuro adopts a model where designated data providers are responsible for supplying the information used to settle bets. While Azuro aspires to incorporate a wider network of data providers in the future, it currently relies on a single provider. This reliance on a central entity for crucial information introduces a single point of failure.

Drift, on the other hand, utilizes a system where an elected governance representative determines the outcome of bets. This representative operates within Drift’s decentralized governance framework, introducing a level of community oversight. However, this system of indirect representation is still potentially susceptible to having a limited number of points of failure, potential bias of the individuals elected, and reduced transparency and accountability compared to a more distributed consensus mechanism.

In essence, while both models aim to streamline efficiency by introducing centralized elements, they differ in their chosen points of centralization. Azuro centralizes the data provision aspect, while Drift opts for an indirect representation system.

Both approaches represent trade-offs between decentralization and efficiency, and their long-term sustainability depends on their capacity to mitigate the inherent risks associated with these centralized components.

Optimistic Oracles Model

@Polymarket, currently utilizing an optimistic oracle system (UMA), attempts to strike a different balance between decentralization and efficiency. By operating under the assumption that the first reporter is generally honest, this allows for faster settlements. Furthermore, the option to raise disputes theoretically ensures the accuracy of the reported outcome, while introducing economic security elements through UMA’s dispute mechanisms.

However, as observed in various Polymarket cases, this dispute process can be lengthy and complex, often hinging on the subjective interpretation of what constitutes “official information” and “credible reporting.”

For example, in the case of the Barron Trump and $DJT Token market, UMA, relying on limited public evidence, had settled the market as “no” in response to the question of Barron Trump’s involvement with the $DJT memecoin. Polymarket intervened when further evidence emerged contradicting UMA’s initial resolution and ultimately reversed this decision, refunding users who bet “Yes.”

The 2024 Venezuelan Presidential Election presented a complex situation where UMA’s resolution, based on international recognition of Juan Guaidó as president, diverged from the reality in which Nicolás Maduro maintained control. This case highlights the difficulties decentralized oracles can face in nuanced political situations where global perceptions don’t necessarily mirror actual circumstances.

In the Justin Bieber Baby Gender prediction market, a conflict arose between UMA’s proposed outcome, based on external sources, and Polymarket’s explicitly stated rules (Polymarket mandated official confirmation from the Biebers or their representatives for determining the market’s resolution).

Disagreements also surfaced in the @LayerZero_core Airdrop market, centered on whether a small fee to claim tokens disqualified the event as a true “airdrop.” UMA’s “Yes” ruling was met with resistance from the community, which disputed the result, even though it ultimately got settled according to the original outcome proposed.

Again, the @Ethereum ETF market experienced discord when UMA’s resolution hinged on the SEC’s approval of related filings but preceded the ETF becoming fully operational.

Finally, something similar happened with the very recent @Eigenlayer airdrop: originally the rules of the specific market referenced September 30th UTC, but when the official tentative date for the transferability of $EIGEN was announced, it fell exactly one minute after it. This led to chaos and a reverse of valuations between “Yes” and “No”, with the latter trading at almost 100%.

In the end, an early launch from Eigenlayer (it happened some minutes before than what was originally scheduled) put everything into question again, and the vote ultimately resolved for “Yes” on UMA.

Source: Polymarket

Ultimately, the model’s reliance on the initial reporter’s honesty and accuracy, which can’t be guaranteed, remains a point of contention, and it has sometimes led to contested outcomes, which were not welcomed by users.

More generally, even setting aside the intricacies of individual market settlements, optimistic oracles introduce a specific set of potential issues. In particular, the assumption underlying optimistic oracles is that people will act according to the best overall outcome, in line with game theory.

However, game theory performs best when the game is repeated between the same actors without strict limitations, which is not necessarily always the case in settlement scenarios, where individual actors could exploit opportunities to their own advantage.

Source: Mr Banks Economics Hub

Moreover, the simple base case for the Prisoner’s Dilemma fails to account for more complex scenarios, where a sub-segment of people can consistently act against the best interest of the people not belonging to it and maximize their group’s returns.

Contesting the proposed outcome might not always be viable, due to costs, risks, and the friction associated with disputes. Even when disputes occur, an additional human layer, like UMA’s DVM (Data Verification Mechanism), is necessary to handle escalations.

The DVM relies on its economic guarantees: it requires a substantial stake to corrupt the oracle, at the risk of malicious voters getting slashed, yet this theoretical security can falter under specific scenarios, such as coordinated attacks (exacerbated by supply concentration) and misaligned incentives.

Source: Polymarket User Comment on “Will Trump Launch a Coin before elections?”

In addition to this, should the vote not reach the required quorums, it will roll over, requiring even more time of the settlement and adding on uncertainty.

In short, while optimistic oracles present intriguing game theory dynamics that align with crypto’s goal of serving as an incentives layer to coordinate human actions, the growth of the prediction markets subsector is revealing cracks in this model’s uncertainties and imperfections that affect protocols’ UX and feedbacks from end users.

A New Decentralized Approach: Edge Oracle

OutcomeMarket by @wintermute is an example of a new decentralized approach which solves for the shortcomings of previous models. Powered by the Edge Oracle, the model maintains the separation of the Oracle and Prediction Market layers, reducing conflicts of interest while introducing a novel approach that strives for a more robust balance between decentralization and efficiency. How does this work?

First, Edge utilizes a set of pre-designated trusted sources and a decentralized network for consensus, filtering out misinformation.

Furthermore, Edge implements LLMs (Large Language Models) for objective interpretation and settlement determination based on the aforementioned verified news sources to reduce bias. Leveraging LLMs in this context allows for more transparency and objectivity: even if LLMs can develop biases during their training process, these processes can be audited and additional safeguards can be implemented. The usage of LLMs on hard news sources should also reduce biases to a minimum, as LLMs’ activity is limited to source interpretation. Last but not least, Edge is a decentralized network, relying on a plurality of nodes to form consensus, guaranteeing settlement reliability.

In essence, Edge positions itself as a “reader of truth” rather than an “arbiter of truth,” leveraging automation and AI for efficiency while maintaining decentralization.

The 2024 U.S. Presidential Elections will serve as the first crucial testing ground for this approach, which will ultimately help shape the future of web3 prediction markets.

An Inefficiency Problem

One of the side yet relevant discourses that have always characterized prediction markets, especially in crypto, is the one about the perception of risk-free opportunities.

Source:@VitalikButerin Twitter

More precisely, we can distinguish between “sure bets” and “value bets”.

Sure bets are defined by a >100% chance of winning money from a combination of bets; usually, they stem from arbitraging different platforms or, at least, different markets. For example, earlier this year, it was possible to bet that Kamala was going to win at 40% odds and that Trump was going to lose at 46% odds.

Value bets, on the other hand, are defined by an extremely high but <100% chance of winning and do not require arbitraging. For instance, the odds of an alien race being discovered before the end of the month were quoted at 99.8%.

All of this is very common in prediction markets and even more common in the nascent sub-sector of “web3” prediction markets. That said, until now, these possibilities have persisted for a simple reason: betting on humans not meeting aliens in the next 10 days sounds like a very likely bet to win, but it is still not 100% certain and what is the opportunity cost of taking this bet?

A 0.2% return in 10 days translates to a 7.3% APR, only slightly above the risk-free yield and below what is considered the base yield in DeFi. On top of this, there is platform/counterparty risk.

It will be interesting to observe whether these types of inefficiencies are effectively on the way to being resolved, at least partially, thanks to the introduction of base yield on some prediction markets, similar to what some brokers already offer for users’ deposits. Even though this adds potential additional counterparty risks, it may help to widen the gap between the risk-free yield and the value betting yield in favor of the latter.

Disclaimer：

1. This article is reprinted from [Chaos Labs]. All copyrights belong to the original author [@0xGeeGee]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.