Crypto Carry: Funding on Perpetuals as a Risk Premium That Reverses
Where does the return on a perpetual-futures carry trade come from?
The observed carry comes from the futures price relative to spot and, for a perpetual contract, from funding exchanged between long and short holders. Both legs can reverse and carry trading bears basis and execution risk.
Evidence map
| Aspect | Finding |
|---|---|
| What it is | The treatment of the funding rate paid between long and short holders of a perpetual futures contract as a carry, comparable to the carry of currencies or commodities, with the same property of paying steadily and then reversing sharply. |
| Key result / formula | Schmeling, Schrimpf and Todorov define carry in crypto markets from the basis between futures and spot and from the funding rate of perpetuals, and study its behaviour across currencies and venues. |
| Why it matters for backtesting | Two things must be true before this can be tested in Stochastly. |
Key result / formula
The carry is on average positive for the short side of the perpetual, reflecting persistent demand for leveraged long exposure, and it is strongly time-varying, rising in periods of speculative enthusiasm. A strategy that harvests it earns a steady stream punctuated by sharp losses concentrated in the moments when positioning unwinds, which is the same shape documented for currency carry, and the authors relate the variation in crypto carry to market-wide measures of risk appetite and to conditions in the wider crypto market. The premium is therefore best read as compensation for bearing the risk of the unwind, not as a free payment.
Why it matters for backtesting
The user needs a funding rate series aligned to their bars, which exchanges publish but which is not part of an OHLCV file, and they need to model the position as the pair of legs that generates the carry, since a perpetual alone is a directional position and not a carry trade. Given both, the test on bars is sound and instructive: accumulate the funding received, subtract the return of the hedge, and read the full distribution, because the object of interest is the tail, which the mean does not show. The pitfall specific to carry is that the Sharpe ratio of a strategy with rare large losses is flattering over a short sample, so the verdict must come from a tail measure and from an explicit count of how many unwind episodes the sample actually contains (see [Tail Risk: CVaR / Expected Shortfall]). Without a funding series, an agent should state that the strategy cannot be evaluated here, and should not approximate it from price alone.
Source
Schmeling, Schrimpf & Todorov, "Crypto Carry", BIS Working Paper 1087, Bank for International Settlements, 2023. Primary source