How we grade HIP-3 markets
Method v1, in force since 2026-09-16. Every grade records the method version it was computed under, so an older grade is never rewritten by a later change.
The reference price is the venue's own oracle — the deployer's reading of the underlying market, and the price funding pulls the perp toward. It is not an independent feed. While the underlying market is closed the oracle is priced from Hyperliquid's own book, so tracking is measured only during open sessions, and closed-hours quality is measured by the jump when the market reopens.
Weights
| Tracking | 40% |
| Reopen fidelity | 20% |
| Cost of carry | 20% |
| Liquidity | 20% |
Thresholds
| Measure | Good | Fair |
|---|---|---|
| Tracking, median gap | ≤ 10.0 bps | ≤ 30.0 bps |
| Reopen, median jump | ≤ 25.0 bps | ≤ 75.0 bps |
| Funding, annual rate | ≤ 10.0% | ≤ 25.0% |
| Liquidity, 24h volume | ≥ $10.0M | ≥ $1.0M |
- A market cannot rate good on tracking if its 95th percentile gap exceeds 50.0 bps, however small the median — one bad hour a week is still a bad hour.
- “Tracks closely” on a market page means within 10.0 bps of the oracle.
- Liquidity is demoted one step when open interest falls below $250.0K.
From ratings to a score
Each dimension rates good, fair or weak, worth 100 points good, 60 points fair, 20 points weak. The score is the weighted average of those points over the dimensions that apply; where one does not apply — a 24/7 underlying has no reopens — its weight is shared across the rest.
Grades
A ≥ 85.0 · B ≥ 70.0 · C ≥ 55.0 · D below that. A market is graded only once it has 14 days of samples and 2 reopens; until then it reads "Not graded".
Markets whose underlying never closes are graded without the reopen dimension, its weight shared across the rest. Pre-IPO markets have no traded underlying and are never graded.