The oldest trade in derivatives markets exists in crypto too, and it is frequently presented as a way to earn a return without taking a view. The first half is accurate. The second describes the direction risk it removes and not the several others it introduces, which are the ones that have actually cost people money.
The same asset trades at one price for immediate delivery and another for delivery at a future date, and the difference between them is the basis. It is not a mispricing: it reflects the cost of carrying the asset until that date, the demand for leveraged exposure, and whatever participants collectively believe about the interval. In crypto the second of those dominates, which is why the basis is usually positive and sometimes very much so.
The trade that exploits it is mechanical. Buy the asset for immediate delivery, sell the dated future against it, and hold both until the future expires. At expiry the two prices must converge, because the future becomes a claim on immediate delivery, and the difference you locked in at the start is what you keep. The direction of the price between now and then does not matter to the outcome, which is the property that makes the trade attractive.
That property is real and it is narrower than the phrase market neutral suggests. The position is neutral to the direction of the asset. It is exposed to the financing of both legs, to the solvency of whoever holds each side, to the possibility of being closed before expiry, and to the cost of maintaining margin through a move. Those exposures are what the rest of this is about.
Convergence at expiry is guaranteed by the contract. Everything that can go wrong happens on the way there, and the way there is where the position actually lives.
A perpetual contract has no expiry date, so there is no convergence event to wait for. What replaces it is the funding payment: when the contract trades above the spot price, holders of long positions pay holders of short positions periodically, and that payment is what keeps the two prices tethered.
The trade adapts accordingly. Buy the asset for immediate delivery, sell the perpetual against it, and collect the funding payments for as long as the contract trades above spot. The direction exposure cancels exactly as before, and instead of a single locked-in difference realised at expiry, the return arrives as a stream of small payments whose size and sign are not fixed in advance.
That difference matters more than it first appears. The dated version tells you at entry what you will earn if nothing breaks. The perpetual version does not: the funding rate can fall, and it can turn negative, at which point the position that was collecting payments starts making them. Nothing forces it back, and a position held through a period of negative funding pays for the privilege of remaining neutral.
It is worth being precise about who is paying, because it explains when the trade works and when it stops. A positive basis exists because participants want leveraged long exposure and are willing to pay for it, and the trade is the other side of that willingness. You are supplying the leverage they are demanding, and the payment is the price of that supply.
This is not a market inefficiency and there is no reason for it to disappear. Leverage has a price in every market that offers it, and somebody has to be on the other side. What the framing does explain is the pattern: the return is large when demand for leverage is high, which is when markets are rising and confident, and it compresses toward nothing when that demand fades.
It also explains the correlation that makes the strategy less diversifying than it appears. Everybody running this trade is short the same thing: demand for leverage. When that demand collapses, every one of these positions stops earning at the same moment, and the ones that are financed with borrowed capital come under pressure simultaneously.
None of them involve the price moving, which is the exposure the trade removed. All four have caused real losses.
You are short the future, so a basis that widens after you enter produces an unrealised loss on that leg. Convergence at expiry means the loss reverses eventually; margin calls do not wait for eventually. A position that is right about the destination and undercapitalised for the route is closed at the worst point, and this is the most common failure.
On the perpetual version, the payment stream that constituted the return reverses. There is no expiry to wait for, so a position held through such a period simply accumulates cost, and closing it means crossing two spreads to exit a trade that produced nothing.
Holding spot on one platform and the short on another means the position is neutral in aggregate and not neutral on either platform. A large move produces a gain on one and a margin call on the other, and satisfying that call requires moving capital between them at exactly the moment transfers are slowest and both venues are busiest.
The position depends on both legs being honoured. If the venue holding either one fails, the hedge disappears and what remains is an unhedged position in whichever direction was left, discovered at a moment chosen by somebody else.
The appeal of the trade is that its payoff can be stated in advance and does not depend on being right about anything. That description is accurate about the payoff and silent about the path, and the path is where the capital is consumed. A position that returns a modest percentage over three months, and that requires you to survive a period of adverse marks in between, has a risk profile shaped very differently from what the headline return suggests.
There is a specific shape to it that is worth naming. The trade produces small, regular, predictable gains and occasional large losses concentrated in the moments when everything else is also going wrong. That is the classic profile of a strategy that looks excellent on a track record and is dangerous in size, because the record contains many of the good outcomes and few or none of the bad one.
Which does not make it a bad trade. It makes it a trade whose risk arrives all at once, and sizing it as though the risk were spread evenly across its history is the mistake that turns a sound position into a serious loss.
Small regular gains and rare large losses is a real strategy profile and it is also what a strategy looks like just before the rare part.
Because the position is direction-neutral, it appears to justify a larger size than a directional one, and platforms frequently support that with margin arrangements that recognise the hedge. This is reasonable in principle and it is precisely where the trade becomes dangerous, because the leverage that makes the return interesting is also what makes an adverse move in the basis a solvency event rather than an inconvenience.
The arithmetic is unforgiving. A trade capturing a modest percentage needs substantial leverage to produce a return worth the operational effort, and at that leverage a basis widening of a few percent against the position consumes the margin. The event that ends the position is therefore much smaller than the events people size against, and it is an event in the basis rather than in the price.
The defensible version is to size against the historical range of the basis rather than against the expected return, and to hold enough unused margin to survive the widest widening in that range without being closed. That produces a smaller position and a much lower probability of the one outcome that matters.
The gross basis is not the return. Both legs pay trading fees on entry and again on exit, which is four fees rather than two. Both legs cross a spread, and on the futures leg the spread is frequently wider than on spot. If the two legs are on different platforms, capital has to be moved, which costs network fees and time. And margin held against the short leg is capital that is doing nothing else.
Adding those up honestly removes a substantial share of a typical basis, and for smaller accounts it can remove all of it. This is the calculation that separates the trade from the description of the trade, and it takes ten minutes with a real fee schedule rather than an assumed one.
There is also an opportunity cost that never appears in the arithmetic. Capital committed to this position is unavailable for anything else for the duration, and the duration is set by the expiry or by how long funding stays positive rather than by you. A modest return on capital that could not be redeployed is a different proposition from the same return on capital that could.
Five numbers, and the trade either survives them or it does not. The gross basis, annualised, so it can be compared with anything else. The full round-trip cost of all four executions plus any transfers. The historical range of the basis on this instrument, which tells you what adverse move you have to survive. The margin required and how much spare you would hold above it. And on the perpetual version, how often funding has been negative historically and for how long.
The reason to write them down rather than estimate them is that the trade's whole appeal is its predictability, and a predictable return computed from assumed costs is not predictable at all. Traders who do this calculation frequently find that the basis they were going to capture was smaller than the costs of capturing it, which is a useful discovery to make on paper.
No. It removes exposure to the direction of the price and leaves exposure to the basis widening before expiry, to funding turning negative, to margin calls on one leg while the gain sits on the other, and to a venue failing. Every real loss in this strategy has come from one of those four rather than from direction.
From participants who want leveraged long exposure and are willing to pay for it. You are supplying the leverage they are demanding. That is why the return is large when markets are confident and compresses toward nothing when demand for leverage fades, and it is not an inefficiency that should be expected to disappear.
A dated future converges at expiry, so the difference you lock in at entry is what you earn if nothing breaks. A perpetual has no expiry, so the return arrives as funding payments whose size and sign are not fixed. The perpetual version can start costing money and nothing forces it back.
Yes, and then the stream that was your return reverses and you are paying. There is no expiry to wait for, so the only exits are holding through it or closing, and closing means crossing two spreads to end a trade that produced nothing. Check how often and for how long this has happened historically before entering.
Because the position is neutral in aggregate and not on either platform individually. A large move produces a gain on one and a margin call on the other, and meeting that call requires moving capital between them at exactly the moment transfers are slowest and both venues are busiest.
Enough less than the platform allows that you can survive the widest historical widening of the basis without being closed. The event that ends this position is a move in the basis rather than in the price, and it is much smaller than the moves most people size against. Sizing against expected return rather than adverse range is the common error.
More than in most trades, because both legs pay on entry and on exit, which is four executions rather than two, and both cross a spread. Add transfers if the legs sit on different platforms. For smaller accounts the total frequently exceeds the basis being captured, which is worth discovering with a calculator rather than afterwards.
Because everybody running the trade is short the same thing, which is demand for leverage. When that demand collapses, every one of these positions stops earning simultaneously, and the ones financed with borrowed capital come under pressure at the same moment. The strategy diversifies less than it appears to.