How to pick a funding arbitrage pair

To pick a funding arbitrage pair, start from the 7-day spread, not the current rate. Then confirm four things: the spread is steady rather than a one-day spike and stays positive after fees at your real hold period; the thinner leg has enough open interest; both legs are the same asset; and you can use both venues.

Each check below maps to a column, filter or badge in our funding rate screener. The example pairs are hypothetical, not a forecast. Rates change all the time, and nothing here is investment advice.

1. Rank by the 7-day spread, not the current rate

The "now" column is one noisy snapshot of the spread (the short venue's funding APR minus the long venue's). On centralized exchanges it is the venue's estimate for the next payment, not money paid; some DEXs publish the predicted rate, others the last settled one. One hot hour can top a "now" sort and vanish before both legs are open.

The 7-day spread averages the hourly spread over the hours when both venues have data. The screener ranks by it, builds net APR on it, and shows it only when at least 56 of the 168 hours have data. Compare it with the 30-day column: close means the gap has lasted, far above means it is recent.

2. Is the spread steady or a spike?

A high weekly average can come from one wild day. The screener's badges catch that (details are in the methodology):

• Prior: the average spread over the last 7 days, leaving out the last 24 hours (six days). It needs at least 96 hours of shared data; with less, the pair is marked new.
• Spike: the 24-hour spread is more than 3× the prior, or above 15% APR when the prior is under 5%.
• Steady: a prior of at least 5% APR, a positive 24-hour spread that is not a spike, and a positive current spread.
• Mixed: the spread was not negative in fewer than 60% of the week's hours. Weak means a prior under 5% APR; cooling, a solid prior whose last 24 hours or current spread are no longer positive.

Both pairs below pay 17 bps of round-trip taker fees and are held 14 days with $10,000 per leg.

The 4.43% is the 17 bps round trip annualized over a 14-day hold: 0.17% × 365 / 14.

On paper X earns almost twice as much, but one day carries its 7-day figure: six days near 14% and one at 90% average about 25%. Such jumps usually fade. Back at its prior, X nets about 9.6%, less than Y, on a thinner leg with only $30,000 of open interest. The "Steady only" filter hides pairs like X.

3. What is the net APR at your real hold period?

A pair pays taker fees on four fills (open and close, both legs). They are a one-off cost, not a daily one, so they weigh more the shorter you hold:

fee_rt = 2 × (taker_long + taker_short)

net_apr = spread_7d − (fee_rt / 10,000) × 365 / hold_days

For Pair Y (a 16% spread, 17 bps of fees):

Set the screener's hold to your real plan; the default is 14 days. Fee schedules matter as much as the spread: taker fees on the venues we track run from 0 to 7.5 bps per fill, so a round trip can cost, say, 5 bps on one pair and 24 bps on another. The screener assumes base-tier taker fees. Maker orders are cheaper, and the calculator lets you switch either leg to maker; VIP tiers are not modelled.

4. How does breakeven compare with your hold?

Breakeven is how many days of the 7-day spread pay back the fees:

breakeven_days = (fee_rt / 10,000) / (spread_7d / 365)

Pair Y breaks even in about 3.9 days. Our rule of thumb: prefer a breakeven under a third of the planned hold. Y passes on a 14-day hold (about 28% of it) and fails on a 7-day one, where it nets only 7.1%. X's 2.5 days exist only on paper; at its prior of 14% it needs about 4.4. Slippage and the entry gap between the legs are not included, so the real payback takes longer.

5. Is the thinner leg liquid enough?

"Min OI" is the smaller open interest of the two legs, in dollars. Below $50,000 the score drops and a "low OI" badge appears, but that is a floor, not a target: a $10,000 leg on Pair X's thinner market would be a third of all its open positions. A rough guide: if your leg would be more than a few percent of the thinner leg's open interest, expect to move the price on entry and exit, and your own position can push that venue's funding against you.

Check book depth at your size and 24-hour volume. Each coin's page shows every venue with open interest and volume side by side.

6. Same ticker, same asset?

One ticker can mean different things on two venues. In one real case a coin traded at $11.65 on one venue and $0.106 on another under the same name, and Gate's CAT is Caterpillar stock, not the CAT token. The screener compares the legs' mark prices: a gap above 2% (200 bps) adds a "≠ asset?" warning, and above 10% the pair is not shown at all.

Contract size is the quieter trap. Contracts on 1,000 coins (1000PEPE, kPEPE) are converted before we compare prices, but on the venue you trade contracts, so match the legs in coins, not in contract counts. Read both contract specs before the first order.

7. Do the intervals and caps of both legs fit together?

Most DEXs settle funding hourly. CEX markets usually pay every 4 or 8 hours, and some switch to hourly when funding hits its cap. Intervals are set per market and shown on each coin's page. Our APRs spread each payment over its hours, but the cash arrives in lumps, and only positions open at the funding time pay or receive it: close an 8-hourly leg a minute early and it misses the whole payment. Pairs with different intervals carry a flag.

Caps shape spreads too. A venue capped at 4% per hour (Hyperliquid, RISEx) can print rates far beyond the ±0.30% per interval that Binance allows on BTC and ETH. Very wide spreads often come from the loosely capped leg, and they can turn just as fast.

8. Can you actually use both venues?

• KYC. Binance, Bybit, Bitget, Gate and MEXC require identity verification. "DEX" does not always mean no KYC: GRVT is widely reported to require it, which we have not verified.
• Countries. Many venues bar some countries; use only those you are allowed to use.
• Invites. Polymarket Perps needs an invite code and is off by default; RISEx mainnet access is invite-based too.
• Hidden costs. Variational charges no trading fees, but you pay its market maker's quote spread, which our net APR leaves out (the "spread not incl." badge).
• Venue risk. Venues can halt withdrawals, get hacked or delist a market: keep on each only what you can afford to lose (see funding arbitrage risks).

The exchanges table lists each venue's type, KYC rule and known regional restrictions. Tick the venues where you have accounts, and the screener's "My accounts" button shows only pairs you can open.

9. How big, at what leverage, and when do you exit?

Each leg needs margin on its own venue: capital = 2 × N / leverage, where N is the notional of one leg. Pair Y at $10,000 per leg and 3x ties up about $6,667, and its 11.6% net APR becomes about 17.4% on capital (net APR × leverage / 2). Leverage lifts that figure and brings liquidation closer: at 3x on each leg, a move of roughly 25–30% against one leg can liquidate it while the other leg's profit sits on another venue.

Set the exit before you enter. Sensible triggers: the spread has flipped and stays flipped (a negative 24-hour spread, not one bad hour); margin on either venue falls below your buffer; or a better pair clears its own breakeven, since switching means paying a full set of fees again. Test your numbers in the calculator.

The checklist

1. Rank by the 7-day spread; treat "now" as a hint.
2. Prefer steady pairs: check the prior, spike ratio and positive share.
3. Read net APR at your real hold period.
4. Keep breakeven under a third of your hold (a rule of thumb).
5. Compare the thinner leg's open interest and book depth with your size.
6. Confirm both legs are the same asset and match them in coins.
7. Note the funding intervals and caps of both legs.
8. Make sure you can use both venues and know every cost.
9. Set size, leverage, a margin buffer and an exit rule before the first order.