Hyperliquid HIP-4 outcome markets: how they work and how to get the data

HIP-4 added outcome markets to Hyperliquid: Yes and No contracts on a question, traded on the same order book as perps and settled at the end. This page covers how an outcome settles, what trades on HIP-4 today, how outcomes are named, and how to get their prices, trades and traders over REST, SQL and MCP.

The short version#

• What: fully collateralized contracts with two sides, usually Yes and No. No leverage, no liquidations.
• Price: 0 to 1, the implied probability. Buying Yes at p is selling No at 1 - p.
• Settlement: the winning side pays 1 USDC, the other side 0.
• Growth: weekly fill notional went from $25.3M to $93.9M between late September and late October 2026.
• Data: outcome fills from 2 May 2026 on api.ironflow.sh, readable without a key, plus SQL and MCP.

1. How an outcome works#

Hyperliquid's HIP-4 pagedescribes outcomes as "fully collateralized contracts that settle within a fixed range". Each market has two sides, each with its own token, usually named Yes and No. The two books are merged: "an order to buy Yes at price p is equivalent to an order to sell No at price 1-p". Matching uses price-side-time priority.

• Settlement. Yes converts into settleFraction quote tokens and No into 1 - settleFraction. A binary outcome settles at 1 for yes and 0 for no.
• No leverage. Outcomes are "an alternative form of derivative trading that does not involve leverage or liquidations".
• Questions. A question groups outcomes where exactly one settles to Yes, such as the winner of a league.
• Builders. Builder codes earn builder fees on sell orders. On 4 November 2026 the page said outcome trading fees were zero for initial testing.

2. What trades on HIP-4#

The first market was a daily binary on the BTC mark price, settling at 06:00 UTC. By November 2026 outcomes covered daily BTC price levels, football and NFL matches and season winners, central bank decisions and company events. In the 7 days to 4 November the most traded outcome sides included the Fed's October hold, Champions League and NFL winners and an Anthropic IPO valuation market.

Weekly fill notional, wallets and fills on outcome markets, from our index (weeks start Monday). Fill notional counts every fill, so each side of a trade adds its own notional; counted once per trade id, the week of 26 October comes to $71.4M.

Since 2 May 2026 our nodes have recorded 20.7 million outcome fills across 39,636 outcome sides. In the 7 days to 4 November, 1,188 of the 4,321 wallets that traded outcomes also traded perps.

3. How outcomes are named#

Hyperliquid's API names each side with a #N coin, such as #240640. That number says nothing about the question, so Ironflow also gives every side a readable symbol built from the question, the outcome and the side, which is what requests take. The #N coin stays in each market's hl_coin field.

• BTC-76637-20261104-Yes: BTC above 76,637 at the 4 November settlement.
• BTC-74000-20261109-0800-No: a BTC level with a settlement time in the name.
• AEK Athens vs Real Madrid-20261104-Real Madrid-Yes: one outcome of a match.
• Federal Reserve's Open Market Committee October 2026-Hold-Yes: one outcome of a rate decision.

/v1/markets returns each side's metadata: the question and outcome ids and descriptions, the side name, the deployer and the quote token. Search by any part of the name with q.

4. Markets, prices and trades over REST#

# volume, wallets and every listed outcome, last 24 hours
curl "https://api.ironflow.sh/v1/predictions/summary?window=1d"

# listed BTC outcome sides, with their #N coins
curl "https://api.ironflow.sh/v1/markets?market_class=prediction&q=BTC&limit=5"

# hourly candles for one outcome side
curl -G "https://api.ironflow.sh/v1/candles" \
  --data-urlencode "market=BTC-74000-20261109-0800-Yes" -d interval=1h -d limit=24

Funding, open interest and mark prices are perp concepts: /v1/funding, /v1/open-interest and /v1/mark-prices answer 400 INVALID_MARKET_CLASS for an outcome. Use trades and candles instead.

5. Settled outcomes#

Most outcomes live for a day or a few weeks, so most outcome history is of settled markets. Settled sides leave the active list and stay queryable. List them with include_inactive=true; each carries active_from_ms and active_to_ms. Ask for trades or candles inside that range:

# find a settled outcome and when it traded
curl "https://api.ironflow.sh/v1/markets?market_class=prediction&include_inactive=true&q=BTC-76637-20261104"

# its last hour of trading (to = its active_to_ms; keyless works for 24 hours
# after it settles, a free key for 30 days)
curl -G "https://api.ironflow.sh/v1/trades" \
  --data-urlencode "market=BTC-76637-20261104-Yes" -d from=1793768455287 -d to=1793772055287

Without a key the API reaches 24 hours back; a free key reaches 30 days, which covers the full life of most outcomes.

6. Traders and volume in SQL#

/v1/query runs one read-only SQL query over the same index. Outcome rows carry market_class = 'prediction'. For questions about wallets and volume over days, use hl.wallet_daily, one row per wallet, day and market: it answers in about 100 ms where a scan of every fill takes seconds.

-- the most traded outcome sides this week
SELECT market, round(sum(volume_usd)) AS volume_usd, uniqExact(address) AS wallets, sum(fills) AS fills
FROM hl.wallet_daily
WHERE market_class = 'prediction' AND day >= today() - 7
GROUP BY market ORDER BY volume_usd DESC LIMIT 20

-- wallets that trade both outcomes and perps
SELECT address, sumIf(fills, market_class = 'prediction') AS outcome_fills,
       sumIf(fills, market_class = 'perp') AS perp_fills,
       round(sumIf(volume_usd, market_class = 'prediction')) AS outcome_volume
FROM hl.wallet_daily
WHERE day >= today() - 30
GROUP BY address HAVING outcome_fills > 0 AND perp_fills > 0
ORDER BY outcome_volume DESC LIMIT 50

Send either as GET /v1/query?sql=... or as JSON with POST /v1/query. GET /v1/query/schema lists every table and column.

7. Python#

import requests

r = requests.get("https://api.ironflow.sh/v1/predictions/summary", params={"window": "1d"})
d = r.json()
print(f"24h: ${d['notional_total']:,.0f} across {d['markets_active']} outcome markets, {d['wallets_total']} wallets")
top = sorted(d["active_markets"], key=lambda m: m["notional_24h"], reverse=True)[:5]
for m in top:
    print(f"{m['display_symbol']}: last {m['last_price']:.2f}, ${m['notional_24h']:,.0f}")

8. MCP#

AI clients reach the same data at https://mcp.ironflow.sh/mcp without a key. list_markets with class prediction finds outcomes, get_recent_trades and get_candles read one side, and run_query runs the SQL above. In Claude Code:

claude mcp add --transport http ironflow https://mcp.ironflow.sh/mcp

Questions#

Related guides#

• Hyperliquid stock perps (HIP-3)
• Hyperliquid historical data: API, S3 archive and Python
• Hyperliquid trades and fills data

References#

• Hyperliquid docs: HIP-4 outcome markets. Settlement, merged books, fees and the first market, quoted above.
• Hyperliquid docs: HIPs. HIP-1 to HIP-4.