Prediction markets (PMs) are one of the crypto use cases that most captured people’s imagination. It was mentioned in the original Ethereum whitepaper and Ethereum’s first ICO (Augur) raised $5.5M to build the world’s first decentralized prediction market. However, neither Augur nor any of its competitors have thus far been able to achieve meaningful traction. While we believe generalised PMs will eventually be a thing, initially it makes more sense to start off focused on a particular niche. So far, the only niches that have seen any traction are sportsbetting and insurance.

Augur V1 featured a P2P architecture that, coupled with high Ethereum network fees and the volatility of $ETH as collateral, made bootstrapping liquidity difficult. Augur v2, Polymarket, and Omen, all sought to leverage the simplicity of the Uniswap AMM breakthrough as a way to solve the liquidity cold-start problem. While sidechains/L2s alleviated some of the UX issues, these designs were still plagued by issues such as high IL due to the nature of prediction markets where positions are highly uncorrelated as markets often resolve spontaneously due to external catalysts. Oracle latency also means traders can front-run updates before market resolution, further harming LP profitability.

However, there have recently been some promising breakthroughs in the design space. Thales, built on top of Synthetix, has adapted the traditional AMM design for betting, incorporating features like vig, risk cap, odds range, skew impact, and time limit. As a result, its AMM is actually profitable. On the other side, designs such as Azuro resemble the Gains oracle-based DEX, using a single stablecoin pool and importing prices odds using oracles. Both designs can be seen as a form of decentralised casino. Similar to Gains, the challenges here are: a) constructing a robust risk model to prevent blow-up risk; b) preventing oracle frontrunning and other value leakage. We’re very interested in working with protocols experimenting with both AMM and oracle-based betting models. Where: This would make sense on any Cosmos chain, with speed and the existence of oracles being factors to consider.

Mentors: Jose Maria Macedo (Head of Delphi Labs) and Jordan Yeakley (Researcher at Delphi).