THE FIRST KPI MARKET
WHAT HAPPENED
Yesterday, we created the world’s first KPI-based decision market. 63 traders bet $16,275 on whether MetaDAO co-founder Kollan House would improve his sleep score if he stopped watching YouTube videos before bed.
Going into the market, he stated that he would only stop watching YouTube videos before bed if the market said that doing so would raise his sleep score to 90% or above (he was previously averaging an 85%).
To bet in the market, people bought “Up” and “Down” tokens. These tokens pay out relative to his sleep score. In this case, if Kollan had stopped watching YouTube videos and achieved a 94% sleep score, Up would pay out $0.94 and Down would pay out $0.06.
However, the market never eclipsed 90% and averaged closer to 80%. Because of this, two things happened:
WHAT WE LEARNED
1: These markets are more intuitive to trade
I thought people might get confused by the Up and Down tokens, but by and large people seemed to quickly understand how they worked.
I had this thesis - “if you can understand @polymarket, you can figure out MetaDAO ” - and that seems to be true for these markets.
2: AMM asymmetry is a problem
Because of how constant-product AMMs work, the further the price gets from 0.5 the harder it gets to move even further. It got to the point where it was 8x harder to push the up than down. You can see this here:
I think this biased the price downwards - the price would probably have been more than 80% if this weren’t the case.
There are a few things we can do to resolve this. One is to set the threshold around the 50% mark.
We’re already doing this for the Jito grants programs: NCNs will need to be deemed 50% or more likely by the market to deliver above their benchmark in economic activity in order to get funded.
To have a threshold around $0.5 for this market, we could have had Up pay out $0.05 for every percentage of sleep score above 80% so that $0.50 Up would imply a 90% sleep score.
The longer-term solution is providing less AMM liquidity and moving more towards order books, like Polymarket and Kalshi use.
3: There’s a long tail of product improvements that need to be made
When you’re building a product, you tend to dismiss the small stuff with “we can solve that later.” This is good because it allows you to ship fast, but these small details really matter in prediction markets because people have money at stake.
An example of this is the time-weighted average price oracle. Because it’s what decides whether the market is finalized or reverted – whether your (un)profitable trades are locked in or you get your USDC back – traders tend to have strong opinions on how it should work.
Even though each of these improvements is individually small-impact, in aggregate they could be a huge unlock for participation.
CLOSING THOUGHTS
Thank you to everyone who traded. If futarchy works, you’ll have been a part of history. I’ve memorialized your wallets below. You should now be able to redeem your money back by pressing redeem next to Up and Down:
Thank you also to everyone who provided feedback!