A proposal for approaching prediction market questions differently w/ UMA

This is being written over the weekend, off the clock, and without any UMA, Polymarket, or Risk Labs approval or association. I’m just a guy writing about something I find interesting.

** I’ve never participated on Polymarket and will not do, so as to avoid any concerns.

UMA is an optimistic oracle that escalates disputes to a tokenholder vote. Voters don’t lie because they’re self-interested. How far can we take this? Can it be used to combat propaganda? Could we ever come to trust it more than any other single information source, on the basis of it becoming so reliable and unshakable?

There have been some market upsets when Polymarket sends requests to UMA’s oracle. Users think one thing, but the oracle resolves to a different one. Why does this happen?

My leading theory is that the resolution criteria were different in ways that require some familiarity with Polymarket and UMA.

One recent example was a dispute over the price of ETH in the format of a binary option. Polymarket had a tradition/precedent where they would reference the mentioned Coingecko market unwaveringly, whereas UMA has historically been very precise when it comes to price data. This finally came up when Coingecko fell just over $1600, but all exchange data (Coinbase, Binance, and DEXs) were under $1600 at the moment in question.

UMA resolved to the more precise exchange data rather than CoinGecko. This caused a market upset. Everyone had access to the same information, but the information was being weighted differently.

How do we fix that? The question here is no longer about “how do we get an oracle that arrives at the truth,” it’s about “everyone having the same interpretation from start to finish.”

Our natural inclination is to create more rules and more specificity. I think there maybe a better way. To explain it, I’ll lay out three approaches and compare:

Approaches (1) and (2) have dominated the thinking thus far with Polymarket’s use of UMA’s optimistic oracle.

Approach (3) tries to acknowledge the messiness of both reality and natural language.

Here are what I believe to be the benefits of approach (3):

• Everyone on equal footing: Precedent is a different word for “unwritten rules” or for having a rule that is “typically interpreted a certain way.” Casual and new prediction market users are at a disadvantage because the outcome is based on knowledge that falls outside of the question & rules. Precedent is a UX detraction (not to mention impossible to enforce.)
• Better hedging: An over-reliance on a single data source could result in insurance that fails to pay off. If someone had hedged binary options on the above ETH market, relying on CoinGecko would have been a mistake.
• Wisdom of the crowd: Instead of an individual asking, ahead of time, “how should I instruct people to resolve this market?” Crowds can do so instead.
• Debate starts from day 1 — Since there are identical criteria there is no difference between the truth needle on the prediction market and the truth needle at UMA.

The Good Path

Polymarket provides live streaming information (in the form of price) about the likelihood of an event happening. UMA provides validated information after an event has ended. Both rely on the wisdom of the crowd with financial skin in the game.

If we take route (3), the future I imagine is one where Polymarket markets automatically propose their answer to the OO. Meaning that this price information is used as the de facto proposal.

Furthermore, UMA voters can participate on Polymarket. If they see a mis-pricing based on such unequal information. Ideally, Polymarket users will also be voting tokenholders.

I think that UMA and Polymarket can be singing the same song, all the time, if rules and specifications stop getting in the way.

What’s Next

There are a lot of obvious criticisms you can throw at this plan. What I’d really like to do, though, is ask the UMA DAO to consider funding some experiments around this approach.

I’d like to hear feedback from people on this — Please leave a comment below or tag me in Discord.