Goal
Introduce a new market primitive to Manifold that enables trading and forecasting on continuous time-series data (and gets users excited for perps on MNX).
Background
Perpetual Futures
Perpetual futures (perps) are arguably the most significant financial mechanism to emerge from crypto. Unlike traditional futures contracts, which expire on a fixed date and require settlement or rollover, perps have no expiry. Traders can hold long or short positions indefinitely, with the contract's price kept in line with the underlying asset through a funding rate mechanism: when a perp trades above the spot price, longs periodically pay shorts, incentivizing the price downward; when it trades below, shorts pay longs. This creates a self-correcting system that tracks an external oracle price without ever requiring delivery or settlement.
The result is a simple, powerful instrument for gaining leveraged exposure to any asset — or, in Manifold's case, any measurable index.
AMMs vs. Orderbooks
Orderbooks work well for the most liquid markets, but automated market makers (AMMs) are far better suited to thinly traded or niche markets. This distinction is critical on Manifold, where markets are user-created and long-tail by nature. An AMM-based perp design ensures that even low-volume markets remain functional, with liquidity always available for traders on both sides.
There has been prior work on using AMMs for perps in the crypto world (e.g. on GMX and Ostium), but I believe this proposal is simpler and novel.
Why Perps on Manifold?
Perps on Manifold would unlock a new class of markets beyond binary or multi-choice prediction questions. Users could trade not only on financial assets like Bitcoin or stock prices, but on any continuously updating index: daily temperatures in San Francisco, presidential approval ratings, AI benchmark scores, box office totals, and more — anything with a reliable, regularly updated data feed.
Proposal
Add perps to Manifold using ManiPerp, a simple AMM. ManiPerp uses an external price oracle, dual liquidity pools for for long / short positions, and a funding rate that rebalances the pools based on open interest imbalance.
Read the ManiPerp paper: https://manifold.markets/maniperp.pdf
I wrote up a description of the AMM as a paper (please refresh after you open the page due to a nav bug 🫣): https://manifold.markets/maniperp.pdf
@SG I'm so confused, current price isn't involved in the update formula? That's just based on pool sizes.
Does trading move the current price?
Under what conditions does the perp price disagree with the underlying price?
How do you handle data sources that update infrequently or with lag?
@EvanDaniel trading just influences the funding rate; all trades happen at the oracle price, which is given by an external data source
I'm not actually sold on the perps mechanism specifically. But something in this vein would be neat! Related things I'd love to see:
Auto resolve to API data source. No new market types, just a new resolution mechanism. Usual CDF/PDF stuff, just auto-resolve.
Regular CFDs based on external data source. "Pays M$1 for each time $event happens between now and end date." "Pays $M1 per basis point of fed funds rate on the 1st of every month for 2026." That sort of thing.
@EvanDaniel Regardless of the details, I think it is extremely cool that y'all are thinking about new market types!
Who would update the values? For example, If I wanted to track Newsom odds to win the presidental election on electionbettingodds.com, Would I have to update that myself? Would the site use ai to? @SG
@Jack1 It would have to be automated. Users could point to an API data feed to setup a market.
Alternatively, perps markets could be run just by the house. But this is the age of agentic coding, so it's much easier now for anyone to spin up a server to publish a data feed. And Manifold users are pretty technically sophisticated.
@Jack1 i feel like you chose the least interesting possible example as election betting odds are perfectly serviced by existing prediction markets on manifold
perps would enable people to, for instance, move in and out of parametric insurance positions for whether sinclair will be in another active shooting event. or uhh more conventional risks like getting my car towed or break-in. so if i am irrationally scared i can calm my nerves by setting and forgetting a “bad things happen to me” portfolio position whereas normally i would have to set up an auto bidding bot that moves out of eoy positions by 1/365 every day,
pay my therapist the big bucks,
or seduce more beautiful transgender women and subby men maintain high rep for poly escape velocity while also squeezing as much free emotional labor as possible
advantage of user-sovereign markets in this regard would be to elide moral hazard somewhat … can get cheaper rates than AAA and more reliable insurance than geico