Perpetual futures are missing one thing
A traditional futures contract has an expiry date. On that day its price has to converge to spot, or somebody collects free money — and that constraint is what stops futures from drifting away from spot for long.
Perpetual futures removed the expiry. Convenient, but the constraint left with it. Nothing mechanical prevents the contract from trading above spot indefinitely, with longs winning and shorts bleeding forever.
The funding rate is what exchanges put in its place.
It is a transfer between traders, not a fee
At each settlement the contract squares up:
- When the rate is positive, longs pay shorts
- When the rate is negative, shorts pay longs
The part people miss: the money never reaches the exchange. It moves from one side of the contract to the other. How much moves depends on how far the contract has drifted from spot — the richer the contract, the more longs pay, the more expensive it gets to stay long, and the harder the price is pulled back toward spot.
Not every contract settles on the same clock, and this one bites repeatedly. The standard is every eight hours — 00:00, 08:00 and 16:00 UTC — and most pairs still work that way. But Binance moved a batch of contracts to four hours, and a few to one hour, from 18 September 2025, adding an 8 / N frequency-normalisation term so that the same premium does not double a day’s total cost just because it is collected more often. Apply the eight-hour assumption to a four-hour contract and your per-period figure comes out exactly twice too large, with nothing anywhere raising an error. The formula and the three places people get it wrong are in how the funding rate is calculated.
Most venues compute the rate from two pieces: an interest-rate component, and a premium index measuring the gap between the contract and the underlying spot price. The result is clamped to a cap, so a single violent hour cannot produce an unbounded payment. The exact formula and cap differ by exchange, which is why the same nominal position earns different amounts in different places.
In a bull market almost everyone wants leverage on the long side, so the contract usually trades above spot and positive funding is the normal state rather than the exception. In 2021 that condition held for months at a stretch: ETHUSDT returned 37.562% raw, 375.621% amplified. The same table shows 0.787% for 2022 — the six-year review is about that gap and nothing else.
Where the arbitrageur sits
Holding a short perpetual on its own does collect funding while the rate is positive. It also loses far more than it collects the moment price rallies, so on its own it is a directional bet with extra steps.
The position only makes sense once you buy the same notional in spot at the same time:
| Price move | Spot long | Perp short | Net |
|---|---|---|---|
| Up 10% | +10% | −10% | 0 |
| Down 10% | −10% | +10% | 0 |
The two legs cancel each other out. That is what delta-neutral means in practice: the direction of the market stops mattering, and the only cash flow left in the position is the funding collected at each settlement. Worth flagging that the term means something different in an options context — same mathematical definition, two unrelated risk and cost profiles, separated out in delta neutral explained.
That is the entire idea. It does not forecast the market and does not need to — it collects a payment that the contract rules oblige somebody to make. Actually running it on an exchange is a different problem, and account structure plus API permissions are the first wall: that part is funding rate arbitrage on Binance.
What is left once price risk is gone
Hedging removes price risk. It leaves these:
Funding turns negative. In a bear market the contract trades below spot and the shorts start paying. The correct response is to step out of the position, not to hold it and hope. A strategy that only knows how to be in the trade is not a strategy. Whether you can simply invert the structure and collect on the way down is covered in negative funding rates — mostly you cannot.
Both legs have to fill together. A single filled leg is a naked position — precisely the directional exposure the structure exists to avoid. This has to be atomic at the system level: if the second leg fails, the first is unwound immediately, not queued for a human to notice.
Margin has to survive the drawdown. A sharp rally puts the short leg underwater. Thin margin gets liquidated, and while the spot leg is up by the same amount, the liquidation happens inside the futures account — spot cannot reach across to save it.
The first two are engineering problems. The third is position sizing. What precision alignment, single-leg rollback, margin coupling and rebalancing actually cost to build is laid out in why delta neutral is hard in practice; how the surviving risks are distributed, and to whom, is in the risk breakdown.
One category sits outside those three: fees, spread and borrow interest are invisible on any funding-rate leaderboard. Six hidden costs puts a measured number against each one. Once these are clear, it is obvious why building this is considerably harder than describing it, and why most of the work sits in execution rather than in the idea.
Reading the historical record
Funding is not a constant, and any account of this strategy that quotes a single annual number is hiding something. Rates track sentiment: frequent and rich through bull markets, thin or negative when the market goes flat.
The year-by-year record for this platform, including the worst year rather than a selection of the good ones, is published in full on the performance page, verifiable settlement by settlement — the average there covers only the five complete, period-verifiable years.