Hyperliquid HIP-4 Forecast Market Insights: Design Differences, Volume Comparisons, and Economic Model Analysis with Polymarket and Kalshi.
Author | Mario Chow @IOSG
Data as of October 5, 2026. Prices, staking, order books, and trades come from the Hyperliquid API; volume for individual markets comes from node trade archives published by Liquidiction; Kalshi and Polymarket volumes come from The Block. Charts are the same as the English version, see above.
Summary
The thing most similar to HIP-4 is not Polymarket, but an options ticket hanging next to your perpetual position. Yes shares and your leveraged position sit in the same account, use the same margin, and are matched on the same engine, so a bet and the position it is meant to hedge are always on the same screen. Three other things also set it apart: anyone who stakes 500,000 HYPE can open a market; settlement conditions are hard-coded into a template before the first trade; entry is free and exit costs about one basis point, the exact reverse of how Kalshi and Polymarket charge.
For the first four months, only Hyperliquid itself was running. At launch on May 2, 2026, all markets were operated directly by the protocol, 691 in total, and it was not until August 29 that it opened to external venues. Since then three have paid the stake, and so far there is no fourth.
Whether you can see it depends on where you are. Prediction markets have a separate page at app.hyperliquid.xyz/outcomes, and the frontend decides by region whether to put it in the trading list alongside perpetuals and spot. Accessing via a South Korean network, this column is visible in the market list; via the network we used for testing, it cannot be found in either the list or the menu, but the page itself still opens and trades. Hyperliquid restricts frontend access in several countries, and an entry point you cannot see on your side may be visible to someone two time zones away.
From May to now, HIP-4 has done $317 million in cumulative volume; Kalshi can do that much in under four hours. In September, for every $1,000 traded on Kalshi and Polymarket, HIP-4 traded about 70 cents.
The design is its best part; the problem is scale. More than a quarter of HIP-4's all-time volume came from a football World Cup that has already ended. The daily Bitcoin market that carried the first four months is down 93% from its May peak. New venues pulled September back to $51 million, flat with July, and the one contributing the most is paying traders roughly 0.6 cents for every dollar traded. October started slower, at about $1.25 million per day. Fees across the whole market are about $1,300 per day, less than half the interest a single venue earns by leaving its stake idle.
1. Product Design: Four Differences from Kalshi and Polymarket
At the base fee tier, the rates reverse-engineered from trades with actual fees deducted are as follows:
This is exactly double the spot rate: spot base tier is 0.070% taker and 0.040% maker. Doubling on one side and zeroing on the other means a round trip costs the same as spot, except it is all charged on exit. Common discounts simply multiply: a 4% referral discount turns 0.140% into 0.1344%, and stacking a 10% staking discount on top gives 0.12096%, and every trade we checked matched to the decimal. There is no maker rebate, which is the one place market makers are worse off here than in other parts of the exchange.
No venue has currently set its fee coefficient above the default, so venue revenue is carved out of the protocol fee, not added on top. The only charge you might hit on entry comes from the frontend app: under spot rules, the builder fee cap is 1%, and one app charges 0.81% on entry, roughly six times the protocol exit fee.
Settlement deserves a closer look, because that is where venue risk lies. Polymarket relies on UMA token holders voting to adjudicate, and earlier this year a $375 million Microstrategy market became a heated dispute over exactly that. Kalshi, as a regulated exchange, adjudicates itself, with regulators above it. HIP-4 splits the problem apart: crypto price markets read the exchange mark price at the agreed minute and settle automatically; all other markets are settled by the market opener signing off on the outcome, with no dispute period and no one who can sign on his behalf.
Of the 228 markets open today, 223 fall into the latter category. The constraint relies on money and time, not procedure. As long as a venue still has unsettled markets, the stake stays locked; it cannot be withdrawn for six months; and only wording pre-approved by validators can be used. That wording is deliberately long and pedantic, for example the "company IPO" template lists a full page of conditions just for what counts as a listing. Rules written before trading are better than arguments afterward. Handing the pen to a named party with no appeals channel is indeed worse than having a procedure, but only if that procedure itself does not go wrong.
June was the peak, because of the World Cup. Football markets alone traded $89.5 million, 28% of HIP-4's cumulative volume. Once the World Cup ended, it fell by more than 80% over two months. September recovered to $51 million, with all of the rebound coming from external venues; the first five days of October traded $5 million.
Beneath these events, the protocol has run the same market every day since May: whether Bitcoin will be above a certain price at six tomorrow morning. This is probably the most Hyperliquid-flavored product, and it fell for four straight months before bottoming.
The week of May 11 traded $16.6 million; the week of September 28 only $1.2 million, and the two weeks before that were about the same. The same question asked once a day, with fewer and fewer listeners, only stopping when almost no one was left. Whatever HIP-4 becomes later, it will not be this.
2. Five Weeks of Open Deployment:
Subsidy-Fueled Volume and Duplicate Listings
Outcome was the first to go live on August 29. Skew registered markets the next day and has had almost no volume since. Trade.xyz's Events registered on September 5 and opened to the public on the 10th. Across all of HIP-4, over the most recent day Outcome accounts for 68%, Trade.xyz 21%, protocol-operated markets 10%, and the rest is Skew. Counting only external venues, Outcome has accounted for 92% since August 29, while Trade.xyz has gone from zero to nearly a quarter in the most recent day. It picks the most opinionated markets: binary contracts betting on its own stock and commodity perpetual prices, the NFL, and the only questions on the exchange about whether Anthropic and OpenAI can list before a deadline.
These volumes should be discounted. Outcome has a $1 million rewards program, paying makers and traders by market and by day, but only rewarding orders carrying its own app code. As of October 5, it had paid $273,409 to 2,487 wallets, about $7,400 per day, equivalent to roughly 0.6% of Outcome's total volume, four times the cost of a round trip. A quarter of the budget is spent, and daily payouts have not slowed. External venues traded $1 million to $2 million per day in September, busiest on weekends; by October it fell back to about $1 million per day, while rewards kept being paid.
Open listing has also produced the first duplicate markets. Both Outcome and Trade.xyz listed markets on the Fed's October meeting; over the past seven days Outcome traded $354,000 while Trade.xyz traded only $90. Both also listed a set of questions on Anthropic's market cap on its first day of trading, and neither saw much volume. The fact that anyone can list a question means two venues can list the same question, and their order books do not merge. Volume flows to the one paying subsidies.
Trading content is tilting toward sports. Over the past 30 days, football, American football, and esports accounted for 45% of volume, crypto price questions 42%, stocks and commodities 7%, and the Fed 6%. Open interest is even more skewed: two-thirds of capital at risk is on sports, mostly long-season bets like English Premier League, Champions League, and NFL champions, with crypto prices at only 17%.Optional templates are being added faster than venues. Validators have approved 28 templates, up from 18 in mid-September, adding spread and totals, the Ballon d'Or, a head-to-head between two AI models on the leaderboard, and first-day market cap for company IPOs. The next network upgrade will double the per-venue cap, bringing concurrently live markets to 200 and daily new listings to 1,000. Ninety-five deployers have registered on testnet; mainnet still has three.
III. Order Flow: Seventy Percent of Volume Goes Through Outcome
For a frontend app to place orders on your behalf, it needs you to authorize a separate signing key, and that key is named after the app. So a wallet that has registered on outcome.xyz will forever carry a label like Outcome-9d3c74de, publicly visible to anyone. The exchange's own mobile login works the same way, showing up as Mobile QR; on desktop, every step is signed with your own key, leaving no label. We only looked at Outcome because it accounts for roughly ninety percent of external venue volume, and only its numbers carry weight.
About seventy percent of volume comes from wallets registered on outcome.xyz, about one-fifth from wallets with no label at all, and one-tenth from scripts and third-party tools that named their own keys. The larger the size, the higher Outcome's share: among the top 20 wallets it accounts for three-quarters of the dollar volume, and about half among the rest. This is exactly what the rewards program would produce: it only rewards orders routed through Outcome's code, and the market makers claiming rewards happen to be the largest wallets.
This method has limitations. A label only shows that a wallet once registered through some app; it does not show that a given trade was routed by it. The unlabeled fifth cannot be treated as users of the exchange's official page either: most of them are quoting two-sided on many order books at once, using scripts signed with their own keys. The sample is 121 wallets drawn from recent trades on the 60 most active order books, weighted by dollar volume over 48 hours, and is naturally biased toward market makers. Fees do not answer the question either. Less than 1% of dollar volume in the sample paid a builder fee, nine dollars total over two days, and Outcome's own code charges nothing, so a free routing app and no app at all look identical.
IV. Benchmarking Against Kalshi and Polymarket: A Gap of About 1,400x, Still Widening
The gap is about 1,400x, and in dollar terms it is still widening, even as HIP-4 itself grows. Kalshi did $59.3 billion in September volume, up $20.6 billion from August. Polymarket plus its U.S. business did $13 billion, with the bulk now in the U.S. app. HIP-4 did $51 million.
The incumbents' advantage is not market design; they win on the two things that actually drive volume. First, events people care about, thousands of them, listed within hours of the news breaking; second, distribution: apps, brand, sports audiences, and for Kalshi a compliance wrapper that U.S. brokerages can plug into directly. HIP-4 has 28 templates, three venues, and a user base that was already trading perps.
The incumbents are also starting to move toward Hyperliquid. In early September Polymarket launched perpetual futures for non-U.S. users, 67 markets covering crypto, equities, indices, and commodities, with leverage up to 20x. Putting bets and leveraged positions under one roof was supposed to be HIP-4's unique thing. Now it is a feature race, and Polymarket is entering it holding events.
V. The Economics: Market-Wide Fees Are Less Than Half of a Single Venue's Staking Yield
At the pace of the past 30 days, all HIP-4 traders pay roughly $480,000 a year in fees. Venues take at most half, split three ways. The stake a single venue must lock up to get in would earn $1 million a year just sitting there; and the venue with the most volume is paying out rewards at a rate of $2.7 million a year to buy that volume.
By these numbers, no venue is running a business. They are buying an option, and the option has value: Hyperliquid says deployers will eventually be able to raise fees themselves, volume can grow enough to support fees, and private-market comparables in this sector are valued at $22 billion and $21 billion, with Kalshi reportedly in talks at $40 billion. But it is still an option, and the cost of holding it is a $47 million stake that expires worthless if the volume never comes.
VI. Conclusion: Design Leads, Market Hasn't Arrived
The building blocks are good; the market hasn't arrived.
On the dimensions an engineer would pick, Hyperliquid has built the better product: one account, one margin pool, bets and the positions they hedge in the same interface; settlement terms written into the template before the first trade; Yes and No on the same question sharing one order book, so liquidity isn't split in two. Nothing in the past five months suggests design is the bottleneck.
The strongest argument for it is that you never have to leave. Margin, hedges, and bets are all in one account, and for a desk already trading perps on Hyperliquid, adding a prediction position is one click, no switching platforms, no new balances, no shuffling funds around. Whatever happens in the competition with Kalshi, that has value, and it is why it will persist at some scale even if it never becomes a business. But it is less unique than it was in August: Polymarket now sells leverage alongside bets too, and it started from the side that has users.
It isn't deep enough. Across the 25 most active order books, the median bid-ask spread is 0.29 cents, and about $4,700 sits within one cent of the best ask. A $1,000 market buy fills a few basis points above the best ask; a $10,000 buy pays 2% to 3% more, and in four or five of those 25 order books it simply can't be filled against visible quotes. Retail-size trades work fine here; large capital does not. Market makers keep requoting, and patient limit orders will do better than these numbers, but an order book this thin is what a market five months old looks like.
There also isn't much to trade. Prediction market volume is made of topics, and topics come either from an editorial team listing hundreds of events a week or from the sports calendar. HIP-4 has 28 approved wordings, three venues that each staked $47 million for access, and two of them are already listing the same questions. Kalshi can open a market for a news story in an afternoon.
The user base is also wrong in a specific way. Eight trades per address per day is a trading desk, not a betting public. If the goal is to give crypto traders a capped-upside tool, that's fine, and it's what this design does best. If the goal is the crowd Kalshi and Polymarket are fighting over, it isn't, and the fact that two-thirds of capital at risk sits on season-long soccer lines shows the venues are still pushing in that direction.
The most likely path is the boring one. HIP-4 stays a feature of Hyperliquid, earns pocket change, and its significance is letting perp traders avoid opening an account elsewhere. That's worth doing, but it isn't worth valuing as a prediction market business.