Two opposite positions. One net payment.

Where the return comes from

A perpetual future never expires, so nothing forces its price back toward spot. Funding does that job instead: at a fixed interval, whichever side of the market is crowded pays the other side. In a long-heavy market longs pay shorts; in a short-heavy one shorts pay longs.

Each venue sets its own rate from its own order flow. The same asset can be crowded long on one exchange and crowded short on another in the same minute — and that gap, not a price forecast, is what the strategy collects.

The hedged pair

One asset, two venues, opposite directions, equal size. If the asset rallies, the long leg gains what the short leg loses; if it dumps, the same in reverse. The pair holds no view on price — it exists only to sit on both sides of a funding difference and collect it.

How a pair is chosen

Candidates are ranked on funding both venues have already paid and published, then filtered on whether the trade can actually be put on and taken off at that size.

Settled history only

A pair is judged on funding that has already been paid, never on a predicted rate or a live snapshot. Snapshots flip around zero from one interval to the next — a reading, not evidence.

Consistent, not just positive

The spread has to hold across every part of the lookback window, not merely on average. One strong week hiding three flat ones is a losing trade once costs are paid, and an average will not show that.

Costs come first

Both legs in and both legs out is a real bill. A pair that cannot earn that round trip back in a handful of days of carry is never opened, however good the headline number looks.

Books that can take the size

Both order books must absorb the position several times over, on both sides. Leaving costs what entering costs, and a leg that cannot be closed cheaply is not really hedged.

Getting in and out

Both legs are worked as maker orders. The edge is a few basis points a day; crossing the spread twice would hand back days of it.

The two venues never quote exactly the same price, and that gap wanders and reverts rather than trending. Pairs are opened while the gap sits on the cheap side of its own recent range and closed while it sits on the expensive side, which covers much of the round trip on its own.

The thinner leg goes on first and is filled completely before the second is chased, so the window in which only one leg is live is measured in seconds. Legs are always equal size, and a single leg is never closed on its own to bank a move: a half-closed pair is a naked directional bet, which is the one thing this strategy exists to avoid.

When a position closes

There is no holding period and no price target. A pair is closed for one of two reasons.

The funding reverses

The pair's rolling net funding turns negative. The crowd has swapped sides and the position now pays out instead of in, so it is worked off in the next favourable window.

The funding stalls

The pair still pays, but has decayed toward nothing. Capital parked in a position that barely earns is a cost measured against the next candidate — so a stalled pair is compared with what can genuinely be opened today, and rotated only if something clearly better exists. Often it does not, and the pair stays.

What can go wrong

Delta-neutral means price direction is not the source of the return. It does not mean risk-free.

Funding turns

Rates can flip faster than a maker exit can be worked. The exit is priced patiently, which means it is not instant.

Execution

On a thin book the second leg can fill materially worse than the first. On an edge this small, one bad fill can cost more than a day of carry.

Venue risk

The legs live on two different exchanges. Downtime, a withdrawal halt, a delisting or a liquidation on one leg breaks the hedge while the other side stays exposed. Margin is monitored per leg, not per pair.

Basis drift

The price gap between the two venues can move against an open pair. It is bounded and mean-reverting rather than trending, but it is noise laid on top of a thin edge.