You may have noticed some strange new markets floating around: perpetual markets, or “perps”
A normal prediction market asks a question, trades between 0 and 100%, and eventually resolves. A perp is different. It tracks a real-world number directly — the price of Bitcoin, an approval rating, how green the UK power grid is running, or the share of openrouter tokens that is open-weight. Those are just the beginning.
If you think it’s going up, go long. If you think it’s going down, go short. There’s no resolution date, and the price doesn’t react to your positions. So, many of you are probably thinking:
What’s the point?
Leverage
Many of you have participated in our older format numeric markets, and the worst part of the experience was correcting a market from 70→85, and then winning 1 mana after 6 months. For smaller, open ended movements, it’s just not worth it. Perps fix this by allowing you to add leverage.
Take notice: 5x leverage on M100 bet, means that your position will move like it’s actually M500. If it goes up 10%, you’ll actually gain 50%. Likewise, if it goes down 20%, you’ll be liquidated.
At 25x, if the oracle price drops only 4% you will be liquidated. See:
64,718.20 → 62,129.47 = liquidated at 25x leverage.
However, high risk, high reward. A 4% increase is a 100% gain!
Liquidation
If the number moves far enough against you, your position is closed and you can lose the margin you put up. Your liquidation price is shown before you confirm the trade — check it.
If you go long with no leverage, you cannot be liquidated. Using the bitcoin example, 1x leverage is like owning bitcoin, you own it until you sell it, with one exception: funding rates.
Funding rates
At regular intervals, the more crowded side pays the other side a small amount. This rewards taking the unpopular side. The current rate and next payment time are shown above the chart.
if the losing side can't fully back every winner's profit, winning positions are trimmed slightly ("auto-deleveraging") — profits you've already made stay yours.
How much does it have to move to outpace the funding rates?
Click the “hold cost” pill above the graph to see. Each of those pills attempt to simplify the delivery of this information, and make all important info possible to visualise on the chart. Mess around with it, hover things, and learn.
There’s a lot of information stuffed into that page, and unfortunately nobody has figured out a way to make perps make sense. We're trying something new, so your feedback is essential.
Extras
Perp profits show up in your portfolio but don't count toward leagues for now.
Every perp shows exactly where its data comes from, right below the chart, and if a data feed goes stale, trading pauses until fresh data arrives — nobody gets to trade against an outdated number.
This is a brand-new mechanism, so start small while you get a feel for it. Leverage is a chainsaw, not a butter knife. We’re going to increase to 100x leverage once we’re satisfied things work, and there’s mana in the pool. You may get randomly liquidated if there are weird spikes in the oracles, only post margin if you’re happy losing it.
If you're interested in the AMM math, here's a writeup by Stephen! https://manifold.markets/maniperp.pdf
There’s a lot of changes to come.
We hope to improve the frequency and latency of updates, so be careful if you’re building a bot to slightly outpace our updates.
We may add fees or a spread if required to prevent position closing to dodge funding, or to add extra liquidity to the pool
Limit orders, eventually
What should our next perp be? If there's a public, reliable data feed for it, we can probably trade it. We will monitor this page for suggestions!
Over the last month we've also been working on:
Fixed limit orders being cancelled by mistake. Some limit orders were wrongly cancelled with an "insufficient balance" error even when the owner had plenty of mana.
Selling out of a position always works now. Occasionally a sale would fail because of a microscopic leftover fraction of a share; those slivers are now cleaned up automatically.
Fixed a bug that blocked some numeric markets from being created. If yours failed to create before, try again. Or, tell us to make a perp if that’s more appropriate!
Fixed a bug where some users never received their identity-verification bonus. Everyone affected is being credited the missing mana.
Elections page polish: the bet buttons on the maps now work properly on phones, we added the Florida Senate special election, clarified the House district bet dialogs, and added a share button (which includes your referral code).
Mod improvements: mods can now fix answers even after a market has resolved, and the moderation tools scroll properly on mobile.
Tidied up the home page now that the World Cup is over.
A batch of behind-the-scenes fixes to make betting faster and more reliable, especially on markets with lots of limit orders.
Looking for work?
A reminder from last month: we run a jobs board at manifold.markets/jobs. If you're looking for work, or just open to hearing about something interesting, register your interest. It takes 30 seconds and there's zero downside.
As always, I'd love to hear your feedback! Comment here, message me, or come chat in our Discord. I’ve also seen some people sharing Manifold on bluesky — thank you!
Tod (Genzy) & The Manifold Team
@Jack1 Silly idea, but how about the water level / the tide at some location? It would be funny to see how something that predictably fluctuates works as a perp.
@Gen I didn't realize that; neat!
Water level feels more... intuitive to me. It might be a cool training ground to let people calibrate themselves on.
the funding rates *look* crazy low and slow. we're play money! we don't need to be bound by real money interest rates. make things snappy and rewarding, assume traders around here like to trade for the sake of trading. (I get that funding is about holders, not day traders, but the current rewards aren't enough to challenge mana's annual interest rate status quo)
re: bot fees. I may not run a bot, but I'd still like to avoid this hacky solution for all parties, even unscrupulous human traders. is there a way to pay in/out on a gradient at sell time? we don't in theory always know how the Oracle will move but we do know the relative positions of long/short onsite at sale time. I'm not totally understanding why funding periods are locked to incremental oracle updates anyway. funding could meaningfully be once a day if we accounted for hold time gradients afaict
@Stralor also how does funding work in practice? does it affect position / cost basis, or is it withdrawn to/from balance? both seem not ideal in subtle ways
paying out from balance is a whole can of unfair worms that should never happen
paying into balance is reasonable and fair to not increase exposure
however! paying out from basis also means you can get shanked by sudden momentum shifts and lose not just the value of your investment (fair and expected) but also its long-term HODL value, meaning you can't fire and forget, which is the whole damn point
paying into basis is cool in concept but increases exposure while not really being rewarding without day trading (same problem)
a third option is perhaps manipulating leverage. this seems to be the solution settled on for profit gain, but it's not clear if it's also used in the reverse for funding. doing this has similar pitfalls to basis adjustment
in a world with those three options (maybe you've found a fourth?), my hope for the solution is: pay to balance, pay from basis/leverage, BUT track losses from funding out and refill during momentum shifts.
@Stralor How it works in practice: Funding never touches your balance in either direction. At each funding event, every position on the crowded side is scaled down by the funding rate, size and cost basis together, and every position on the other side is scaled up by the matching amount. Because both legs scale proportionally, your entry price, leverage, and liquidation price don't move at all.
- Funding can never liquidate you: Paying funding just shrinks your position, never moves your liquidation price. A momentum shift can cost you some position size, but funding itself can't. This makes it safer to hold positions longterm
- Receiving funding compounds into your position: you realize it when you close
on rates looking low, they are deliberately limited at the moment. We just increased the cap ~100x on the /Manifold/uk-grid-carbon-intensity-gcokwh market so that it can reach 1%/hr. It should probably be even higher. If it's too high though, then all incentives point to farming the imbalance (you can open/close positions before/after funding periods to collect or dodge funding), which has other issues which are amplified in a zero fee environment.
I'm not totally understanding why funding periods are locked to incremental oracle updates anyway
it's anti-dodge, because we have no fees. Otherwise you can exit your position before funding periods and immediately re-enter after, if the price didn't move, you have nothing to lose. Fees are how real money prevent this. Locking funding periods to oracle updated guarantees risk during that moment
The accrue-by-hold-time and settle funding on close is also a solution, but the weird part is that if you hold the position for months, after the counterparties have been closed/liquidated/or ADL'd, mana from those counterparties needs to be stored somewhere. If we pay it on a schedule then everything is fully backed the instant it moves.
Keep the ideas flowing!
@Gen explain deleveraging for me. are we losing upside if we don't sell fast enough? feels like those old school proportional markets.
it reduces notional?
it reduces margin?
it says it reduces exposure... but I'm not sure what that means other than impacting your leverage unless it closes some of your risk (margin) in the process, but that would further bleed the market dry
take 2 examples. let's say 10 mana at 10x leverage. so a 100 notional. ignoring funding, in a world where I sold that at 50 value and was never deleveraged, I'd get 50 out. but if in that exact same scenario I was deleveraged down to 1x, my position is still 10? and my payout on sale is, what, 14 or something?
relatedly, what happens if you hold a position so long it passes it's 'notional' value? is that where deleveraging comes in? or can it grow beyond that? nvm on this. I made a 1x bet and I understand notional now. it's not payout.
@Stralor it seems deleveraging happens at funding intervals? that might be a helpful insight for me. idg the math yet, but if funding is the reason... are they truly lockstep or only influential on each other?
@Stralor Sorry, I think I addressed this briefly on the comment you made on one of the markets.
Autodeleveraging (ADL) reduces notional (reducing your leverage) but keeps margin in tact. It can only take your profits. Since unrealised profit is proportional to size, trimming size trims the unrealised profit with it (socialising profits to cover excessive losses which are uncovered by losers). The extreme of this is that the other side cannot cover anything, and you're force-settled for only your margin back.
We have this because we don't have any insurance fund, no printed mana, balance clawbacks, or any bs like that. We can add to the pool, which can operate like those things, but when the pool is empty the profits are socialised to cover the losses. The new fees are continually adding back into the pool, so this should happen less.
our ADL scales every profitable position on the winning side by the same factor so total claims = total backing. Pro rata, not like real-money versions where the highest leveraged person gets force-closed first. Partial ADL just shrinks everyone's claims against the pool proportionally.
So to answer the core question:
are we losing upside if we don't sell fast enough?
Yes, kind of, unrealised profit is what is at risk. Less unrealised profit means less at risk. Selling before the trigger means you avoid the ADL
@Gen This creates a wired incentive to sell and re-enter the market every 30 minutes (in the UK grid market). Wouldn't it be better to tax unrealized profits only once?
@Simon74fe The entry fee should negate this, but idk specifically for that market. That market needs a ton of tweaking still and I haven't revisited it since the funding rate changes
@Gen could ADL in theory take all profits from a 1x bet? would mean in a bad world that you're paying fees to get just your margin... even if you're right and 'won'
@Stralor Yes, in theory. The profit haircut would be the same on 1x or 100x
It's the funding-adjusted margin, not the original deposit, so if you have been paying funding you could get back less than you put in
the 10bps fee is tracked separately and not refunded, so you could lose that too.
This is why we currently have a cap on entering based on notional capacity, rather than letting everyone enter and then just settling it however - super parimutuel style. So far it seems to be not terrible to be unable to open a position, but it would be terrible to open that position and then literally gain zero mana...
Maybe we will change this though? Idk, still tweaking things
@JasonMendoza2008 But prediction markets aren't about individuals predicting. The whole point is that the group consensus, futarchy boosters tell us, can actually predict the future. These markets just function like normal investing, it removes the group predictive element.
@ChurlishGambit I think you could layer a smart visualization of the data on these to show the aggregated predictions. they're more futures contracts / options than simple investments. you're mostly right of course but even normal markets here react most strongly to post-predictive events
@ChurlishGambit to me at least prediction markets are all about adding scattering to the world of trading (which is about predicting the future price, not investing) to make more profits by opening up fun arbitrage opportunities. A bit like when I create duplicate markets. Or when NASDAQ actually operates 10 NASDAQ exchanges. Reading more into it seems weird to me. Goal is to maximise profit, not understanding the philosophy of XYZ.
@JasonMendoza2008 This isn't something I'm making up. This place were founded by e/acc "Rationalist" people: https://manifold.markets/about
"Our mission
Provide the most accurate, real-time predictions on any event.
Combat misleading news by incentivising traders to be fast and correct.
Help people make more informed decisions by improving their model of the future."
These markets seem to be moving away from the mission.
This is pretty interesting! Is there a good way to interpret "manifold's prediction" from these markets? Like, if everybody is long on a market, that's us collectively saying we think the real quantity will go up. But can that be quantified and summarised? Is it just a case of total mana long minus total mana short?
@Fion the “hold cost” kind of does this. It will move directionally in line with what going long/short actually means over time. If it costs x% to hold long for a day, that means the aggregate prediction is roughly that it will move x% that direction that day. This doesn’t exactly work at the beginning when the liquidity is partially artificial and the open positions are slim
To learn more about perps:
@realTomBayes Lmfao I have to subscribe to read something written by someone with negative profit on the website? come on
@JasonMendoza2008 @jgyou here is a version so you can read it for free: https://supercycle.blog/p/9d9d20b6-8b1e-46f4-b552-b42d81bad4b8?postPreview=paid&updated=2026-07-13T20%3A53%3A31.680Z&audience=only_paid&free_preview=false&freemail=
i have to make money bro!
not to rag on you, but do you? 🙃
I may have started working at your age but I sure hope you don't have to
@marvingardens I don’t think you can specifically block the market type, no, but we can look into adding this if you like. We will be tweaking the importance scores to make sure that they only surface when appropriate/relevant, and soon(ish) when we revisit personalisation for content discovery, we are intending to make the browse page much more personalised by default (instead of global “best”).
@Quroe not counting towards leagues in August, at least. We don’t want leagues to be decided entirely by perps randomness or errors.
They still count toward leaderboard stats!
@Gen Ah, I was getting my signpost from the ranked Boolean value on the perps markets' details page. I'm not sure if anybody else digs into those like I do, but maybe consider flipping that switch on the perp markets until next month.
@JussiVilleHeiskanen we’re not going to massively subsidise them (with the exception of bitcoin today, I guess), so it shouldn’t meaningfully impact mana supply
@Gen I think I will take an approach of betting one mana this week, two mana next week, four mana the week after that. That should be the ticket, no way that could spiral.
At regular intervals, the more crowded side pays the other side a small amount. This keeps the market price tethered to the real number and rewards taking the unpopular side. The current rate and next payment time are shown above the chart.
@Simon74fe Are you talking about the funding thing? I thought if you could predict accurately the future, you would just eat the liquidity pool.
@Quroe In many other versions of perps there is the underlying asset value ("oracle"/"real") and the perp market value ("market"/"market price") which can diverge. Manifold's implementation just uses the oracle value as the market value. I think "tethers" is a little misleading here since on Manifold by definition they can never diverge.
@Simon74fe that market is definitely the weirdest, least natural thing to trade. It will mostly be about trading the funding rates, but that’s also partly why I included it - people can play around with it, and if it sucks, we can try to find ways to improve it or discontinue it.
We’re still figuring things out, but it was too good to keep hidden while we iterate!
@Simon74fe Agreed. Seems theres too many people on one side which prevents others from taking positions on that side at all and removes those positions from those on that side as soon as it moves and they profit some since the losing side can't pay enough. The funding cost isn't near high enough for how predictable the value is.
@JasonMendoza2008 The explanation in the post is pretty useless for actually trading on these but the pdf they posted has the details of the implementation they are using (though I wouldn't stake my life on the code and the pdf matching lmao)
https://manifold.markets/maniperp.pdf
@HastingsGreer Link is broken. Edit: the fix is to refresh on the 404 page by dragging down.
@Quroe Click it then click reload.
Which is kinda remarkable. I'm not sure how to even implement a CDN that fails that way if I needed to
Do perps exist and work on Dev.Manifold.Markets? I'd like to learn how they work there without risking any real mana. I learn by experience.
@Quroe Yes, I've just un-hidden these for you to test on:
https://dev.manifold.markets/ManifoldMarkets61be/bitcoin-price-usd-5UhA
https://dev.manifold.markets/ManifoldMarkets61be/uk-grid-carbon-intensity-gcokwh
Hi @Gen what does « This keeps the market price tethered to the real number » mean. I thought you said the market price WAS the real number but here you use both as if they were distinct so I’m not sure I understand. Thanks.
@JasonMendoza2008 i think there is a misspelling in the sentence you are refering, as the white paper says the price on Manifolds is the medium of Kraken, coinbase and an other exchange i forgot. But the funding here still reflect the crowd on manyfold.
@JasonMendoza2008 that sentence was an error. Thank you for pointing it out! I have removed it.
The only purpose is to reward the unpopular side, as we don’t allow the market price to diverge.