A scanner that normalises live funding rates across seven perpetual DEXes, then filters for the pairs where the trade actually clears its own execution cost — using three signals that came out of trading it with real capital.
Read-only demo · live market data, 90-second refresh · cannot place a trade, no capital at risk
Heuristics are advisory and may be wrong. Not financial advice.
A perpetual future pays funding — a periodic transfer between longs and shorts that keeps the contract near spot. Venues arrive at different rates for the same asset. Long one, short the other, and price movement cancels: the position is delta-neutral and the funding difference is the return.
Finding those pairs is easy — any two APIs will do it. Deciding which of them is actually tradeable is the whole problem, and it is not visible in the funding rates at all.
Before anything can be compared, it has to be made comparable. Each venue publishes funding on its own interval, in its own units, and encodes leverage caps differently. Get one wrong and the number is off by 8×.
| Venue | Funding interval | Rate published as | Leverage cap from |
|---|---|---|---|
| Variational | 4h / 8h per asset | annualised, RFQ-quoted | not published |
| Ethereal | 1h | hourly | maxLeverage |
| Nado | 1h | x18 fixed-point | 1/(1−w) |
| Lighter | 1h settlement | 8h-equivalent | 10000/f |
| Lighter on Robinhood Chain | 1h | 8h-equivalent | 10000/f |
| Aster | 1h / 2h / 4h / 8h per symbol | per-interval | 100/pct |
| edgeX | 4h | per-interval | riskTierList[0] |
Aster publishes four different intervals across its symbols, so the interval must be read per symbol or every APR on that venue is wrong. Lighter's published rate was verified against Hyperliquid's hourly rate — the ratio came out at exactly 8.000, confirming it is 8h-equivalent despite hourly settlement.
Raw spread is a poor signal on its own. These three columns encode what the measurements actually showed.
Not the funding difference: the price difference between the two venues. A delta-neutral round's basis P&L is exactly the change in this number while the position is open. On thin books it moves further in an hour than the funding pays in a day, which makes it the binding constraint — not the spread.
Does the funding want you to short the more expensive leg? When gap and funding pull the same way, a narrowing gap pays on top of the carry. When they oppose, they fight. Two live rounds on consecutive days, same method, differing only in this flag: +$12.73 and −$4.61.
Reads each venue's own measured history rather than the current tick. Reversed is the strongest warning: the 7-day average points the opposite way. One asset read −703.9% average against +1914.3% current — and turned back within hours.
Six cross tables plus a combined view, each with gap, alignment, leverage cap, verdict, round-trip cost and break-even days.
Per-asset history, order-book depth at your size, funding series on a symlog axis, and a structural verdict.
Tick countdowns per leg and a funding accumulator that credits the rate in force at each tick boundary — not elapsed time × current rate. Across one live round the hourly ticks ran 2.36 → 1.80 → 1.18 → 2.18 → 0.67 → 0.49; the closing rate would have implied $14.16 against an actual $8.68.
Liquidation distance from the venue's real maintenance-margin fraction, not the nominal leverage. A 3× cap and a 10× cap need nearly the same margin for the same safety.
Seeded with 99 hourly snapshots collected since 16 August 2026 across ~700 pairs, so the history-dependent columns are populated on first load. It keeps measuring while it runs. Each snapshot holds funding rate and price per venue per symbol — nothing else.
GET /api/archive available days ·
GET /api/archive/<day> hourly snapshots ·
GET /api/scan full scanner payload ·
GET /api/dive/<symbol> per-asset detail