Funding rate arbitrage: how it works and where it goes wrong

A perpetual future has no expiry date. Nothing forces its price to converge with the spot market, so exchanges invented a mechanism that does the job instead: the funding rate. Every few hours traders on one side of the market pay traders on the other.

When the rate is positive, longs pay shorts. When it is negative, shorts pay longs. The payment does not go to the exchange — it moves between traders, and the exchange only settles it.

That is the whole basis of funding arbitrage. If one venue pays shorts generously while another charges longs almost nothing, you can hold both sides at once and collect the difference.

The position

Take one coin and open two positions of equal size:

• short on the venue where the rate is higher — there you receive;
• long on the venue where the rate is lower — there you pay little, or receive as well.

Price movement no longer matters. If the coin doubles, the long gains exactly what the short loses. If it halves, the same thing happens in reverse. The position is delta-neutral: your profit and loss comes from the funding payments, not from the direction of the market.

Reading the spread

Every venue publishes a rate per payout period. To compare venues you have to convert those rates to a common scale — annual percent:

APR = rate / interval_hours * 8760 * 100

The spread of a pair is simply the difference between the two legs:

spread = APR(short leg) − APR(long leg)

An example from real data. Bitget pays +0.0050% every 8 hours, Lighter pays +0.5144% every hour:

The spread is about 4500% a year. That number looks unreal because it is an extrapolation: it assumes the current rate holds for twelve months, and rates like this usually live for hours.

The interval trap

Here is the mistake that quietly ruins the whole calculation.

Not every venue pays every eight hours. Binance, Bitget, OKX and Aster all run a mix: some instruments settle every 8 hours, many every 4, a few every hour. On Binance more than sixty percent of pairs are not on the eight-hour schedule.

Two rates that look identical are not identical at all:

If a tool assumes eight hours everywhere, it understates half the market by a factor of two, and no amount of correct arithmetic afterwards will fix it. Always check where the interval comes from — the venue publishes it, sometimes in a separate endpoint rather than next to the rate.

What eats the profit

The headline number is gross. Several things take their share before anything reaches your account.

Fees on both sides. You open two positions and close two positions — four taker fills unless you work with limit orders. At 0.05% per side that is about 0.2% of the notional, which a modest spread needs days to earn back.

The weaker leg limits everything. A pair is only as large as the smaller of its two books. A spread of 3000% on a venue where 4000 dollars fits is a three-thousand-percent return on four thousand dollars, not on your capital.

The rate can flip. This is the main risk and it is not hedged by anything. The rate that pays you today can turn negative tomorrow, and then the position starts costing money on both legs at once. Funding arbitrage is not a set-and-forget trade — it needs watching.

Price gap between legs. Perpetuals on different venues do not trade at exactly the same price. Entering when the gap is against you eats part of the future income immediately.

When a spread is fake

A very large number is usually a symptom, not an opportunity. Three checks filter out almost everything worthless.

Compare the prices of the legs. If they differ by more than about ten percent, you are almost certainly looking at two different instruments under one ticker — a coin with a 1000x multiplier on one venue against the plain coin on another. The rates are real, the pair is not.

Look at liquidity, not at the rate. Extreme rates live on instruments where nobody trades. Check open interest and daily volume on the weaker leg before anything else.

Check how long the spread has held. An instantaneous rate says nothing about what the position will actually pay. A spread that has held its level across a day of samples is a very different proposition from one that spiked once.

A realistic view of the returns

Advertised annual percentages are extrapolations of a moment. What matters is the realised number: what the pair actually paid over the last day, week and month.

That is why our arbitrage table shows both — the current spread and what was genuinely paid over each window. Sorting by the realised figure rather than by the instantaneous one changes the top of the list completely, and it changes it in the direction of things that are actually tradeable.

Funding arbitrage is a real, mechanical source of return. It is also a business of small edges, thin books and sudden reversals — and the people who lose money on it are usually the ones who read the headline percentage and skipped everything below it.