Funding-Rate Arbitrage, Honestly
"Collect funding with no price risk" is one of the most-repeated ideas in crypto, and it is genuinely a real strategy. It is also routinely oversold. Here is the mechanism, and then the costs that decide whether it actually pays — using our own realized-funding data.
The two shapes of the trade
Spot-vs-perp (cash-and-carry). When a perp's funding is persistently positive, you go short the perp and buy the same amount of spot. You're delta-neutral: price moves cancel, and you collect funding from the shorts-get-paid side while holding spot. This is the cleaner version and the one most "funding yield" products run.
Perp-vs-perp (cross-exchange). The same coin funds differently on different venues (median 17 annualized points across majors). You go long the low-funding venue and short the high-funding venue, and collect the difference. No spot needed, but now you're exposed on two venues.
The costs nobody puts in the headline
- Four executions. Two legs to open, two to close. As a manual example, assume
the same fixed notional on each leg and a 0.05% fee for each execution. The total is
4 × 0.05% = 0.20%of one leg's notional, or 200 USDT when each leg is 100,000 USDT. This is a fee assumption, not a current exchange quote; spread, slippage and other costs are additional. - Capital on both sides. You post margin twice, and must keep both legs from being liquidated if price runs — so you can't use much leverage, which caps the yield on capital.
- Persistence needs its own measurement. The July 2026 article reported that BTC funding on Binance and Bybit had the same sign about 74% of the time over its preceding 90-day window (ETH: 77%). Those historical percentages have not been rechecked in this correction. Same-sign frequency does not establish the sign, size, persistence or reversal frequency of the difference between the two rates. That requires a separate analysis of aligned settlement cash flows. A difference can narrow or reverse; rebalancing adds execution costs.
- Execution and slippage on the thin mid-caps where spreads are widest — exactly the coins whose funding is most volatile and least persistent.
17 × 30 ÷ 365 ≈ 1.40 percentage points.
For two legs each held at a fixed 100,000 USDT notional, that is about 1,397.26 USDT of gross funding
difference before the illustrative 200 USDT execution fees and other costs. Both figures use
one leg's fixed notional as the denominator, not the total margin or capital committed
to the pair. This is not a return on margin, a forecast or a compounded return. Actual funding
requires the realized settlement payments and position notionals; the same-sign percentages above
cannot tell us whether this assumed difference would persist.
When it actually works
- Persistent, large, liquid. A big spread on a liquid coin that has stayed one-signed for a while beats a bigger spread on a thin coin that whips around. Persistence > magnitude.
- Low fee tier or maker execution. The trade is fee-sensitive; running both legs as maker (0.02% versus 0.05–0.06% taker) rather than crossing the spread can be the difference between edge and no edge.
- Spot-vs-perp on a clean positive-funding regime is generally more robust than chasing cross-exchange mid-cap spreads, because you remove one venue's counterparty and liquidation risk.
- Scale and automation. The per-unit edge is small, so it rewards size and tight operational execution — which is why it's mostly a desk/bot strategy, not a manual one.
How to screen for it
The starting point is the current cross-venue spread per coin, ranked. Our live monitor sorts every cross-listed perp by that spread, and the historical funding-direction study describes observed payment signs, not the persistence of a tradable two-leg difference. Before drawing a conclusion about costs, align actual settlement cash flows over a common window and include all four executions, spread, slippage and capital requirements. A snapshot or a same-sign percentage alone cannot establish a positive net result.