A perpetual shows a funding rate of 0.01% every eight hours. You short it, buy spot to hedge, and expect to collect about 0.03% a day. Then the backtest reports a loss.
That result can be right. The rate is a payment between long and short holders at a particular time, not a daily coupon guaranteed to anyone who noticed it. To see why, take one deliberately small example apart: a $10,000 delta-neutral BTC carry position held for 24 hours.
The signal: 0.01% is a snapshot
Assume the strategy observes a positive 0.01% predicted funding rate at 00:00 UTC. On this venue, funding settles at 00:00, 08:00, and 16:00 UTC. The strategy decides to short the perpetual and buy an equivalent amount of spot.
The prediction is an estimate of the next settlement, not a promise about all three. Funding can change before a payment is calculated. If the backtest carries 0.01% forward unchanged, it has turned one observation into three payments by assumption.
| Input | Assumption in this example | What the backtest needs |
|---|---|---|
| Observed rate | 0.01% at 00:00 | The exact rate known at decision time |
| Settlement times | 00:00, 08:00, 16:00 UTC | The venue's actual schedule and settlement records |
| Position timing | Order arrives just after 00:00 | Whether it qualifies for that settlement |
| Later rates | Unknown at entry | Each rate as it becomes available, not today's final history |
If entry happens after the 00:00 cutoff, the position may receive only the 08:00 and 16:00 payments. At the assumed rate, that is $2 rather than $3. If the rate falls to 0.002% before both later settlements, those payments total just $0.40.
The hedge: matching dollars is not matching exposure
A $10,000 short perpetual paired with $10,000 of spot is delta-neutral only at the prices and quantities used to size it. For linear BTC contracts, a short of 0.1 BTC against 0.1 BTC spot is a useful starting point. If the strategy instead sizes both legs by dollar value and one leg fills at a different price, a residual exposure remains.
That residual matters even over a day. A 0.2% BTC move is $20 on a $10,000 unhedged exposure, many times the example's $3 maximum funding. Rebalancing can shrink the exposure, but every rebalance adds trades, fees, and slippage. I’ve seen tidy carry curves built on a hedge that is exact at entry and fictional after the first move.
The payments: credit the right account, on the right side
Suppose the short receives funding at all three settlements. At 0.01% each, the gross credit is $3. A backtest should reproduce the venue’s sign convention: positive rates usually mean longs pay shorts, but conventions and product details belong in the contract specification.
Funding is generally calculated from the position value at settlement. If BTC moves, the notional changes. So does the dollar payment. Applying a fixed $1 credit three times ignores that variation; using the entry notional may be a tolerable approximation for a short hold, but it should be named and measured.
The round trip: two legs, four trades
At 5 basis points per taker fill on each venue, opening and closing both legs costs about $20 on $10,000 per leg: $5 to enter each leg, then $5 to exit each. That is before spread and market impact. If spot is maker and futures is taker, the bill changes, but a backtest needs evidence that the maker orders would have filled.
The arithmetic is intentionally plain. Even if all three payments arrive at the advertised rate, this particular one-day trade loses $17 before accounting for execution quality. A fee discount, a longer holding period, or more favorable funding could change the result. Each also changes the strategy being tested.
The exit: the backtest has to pay to leave
Funding strategies often look strongest when the backtest enters at the start of a favorable regime and exits at its end. In reality, the signal can turn negative, the basis can widen, and the exit order still has to trade. A negative funding payment near the exit may erase several earlier credits.
Model the decision rule for closing, then charge both legs to close at prices the strategy could plausibly get. Track funding by settlement, fees by fill, and the hedge’s residual price exposure separately. That decomposition tells you whether the trade failed because carry disappeared, execution was expensive, or the hedge stopped being neutral.
A positive funding snapshot is a useful input. It is not a P&L forecast. The backtest earns credibility when every payment has a timestamp, every leg has an execution cost, and the strategy survives the boring arithmetic of staying in the trade.
← All posts
