Funding carry

What goes wrong in cross-venue funding carry

Shorting funding on one venue and buying it on another looks like free money on a screener. Here are the five ways it goes wrong in practice, and what our own trading has taught us about guarding against each.

1. The hedge that never fills

A cross-venue carry has two legs, and it is only neutral once both are filled. Between the first fill and the second, the position is directional. Usually that gap lasts seconds. When a hedge order is rejected, stuck or simply not filled, it can last much longer, and that is where most of the damage in carry trading comes from: not from the funding, but from exposure nobody meant to hold.

The fix is not simply to hedge faster. Crossing the spread on every hedge pays the full spread every time and eats the carry. Waiting patiently leaves the position exposed. The workable answer is a ladder: rest at your own price first, then join the best price, then cross, and after a hard time limit sweep regardless of cost.

An unfilled hedge escalates in steps, from patient to urgent. The final step fills the hedge whatever it costs once the position has been exposed for too long.

Our own trading has shaped how we handle this. The ladder alone is not enough: an order can be blocked or lost without anyone noticing. AlgoBee therefore runs a risk engine next to the strategy. It watches unhedged exposure on every pair, forces the hedge through when a position has been exposed for too long, stops new positions while unhedged exposure is too large or too old, and raises an alarm that keeps repeating until the pair is balanced. The point is not to avoid every small cost, but to cut off the few really adverse outcomes that decide a year's result.

2. The spread does not stay

The funding rate is reset every interval. A spread that looks like 40% a year today can be 5% by the weekend, or reverse. That matters because entering and leaving costs money once, while funding pays a little every day. A 20% annual spread earns about 5.5 basis points a day; with 20 basis points of costs, the position needs almost four days just to pay for itself. If the spread halves, it needs more than seven.

Two checks help. How many days of funding cover the costs, which the screener shows for every pair. And how often the spread actually paid over the past week, which is what the weekly report ranks on. A wide spread that paid half the time is worth less than a moderate one that paid every day.

3. Two venues, two margin accounts

The two legs offset each other in price, but not in collateral. Each venue only sees its own leg. In a sharp move, the losing leg can hit its liquidation level while the winning leg's gain sits on another venue, where it cannot help. After a liquidation you are left with a naked position on the other venue, at the worst possible moment.

Keep enough collateral on each venue for the moves the coin actually makes, not the average day. Rebalance collateral as prices move. And prefer structures where exposure per venue stays bounded, for example by pairing positions whose funding runs in opposite directions on the same venue.

4. Costs you did not count

A round trip is four fills: in and out on both legs. Each pays a fee, crosses a spread and may slip. Fees differ a lot between venues, from zero on some on-chain venues to several basis points per fill on others. The price gap between the venues at entry, the basis, adds or subtracts again at exit. Count all of it before trading; the calculator does it with live rates and your own fees.

5. Data you should not trust blindly

Venues quote funding per one hour, four hours or eight hours, as a fraction or as a percentage, and live data does not always follow the venue's own documentation. A unit mistake can make one venue's rates look ten or ten thousand times smaller than they are, and every pair involving it goes wrong at once. Check a few rates against the venue's own website before relying on any feed, including ours, and check again when a venue changes its API.

Venues also go down, slow down or return partial data. A system that stops for every venue when one of them fails will leave you blind at the wrong time; each venue should fail on its own.

A short checklist

Decide how long a pair may stay unhedged, and make sure something enforces it without you watching.
Judge a pair on days to cover costs and how often it paid, not on today's spread.
Size collateral per venue for a bad day, not an average one.
Count four fills, both spreads and the basis at entry and exit.
Verify funding units against the venue before trusting a number.

See how AlgoBee runs funding pairs Try the funding calculator

For information only, not investment advice. Trading perpetual futures with leverage can lose more than the collateral you post.