The idea
A concentrated-liquidity position is long the base asset in a way you did not choose: as price falls the pool keeps buying more of it. Open a short of matching size on a perpetual, and the short's profit offsets that drift. The position becomes delta-neutral and you keep the fees.
That is the theory, and it is sound. The question is whether it survives costs.
Is a delta-neutral LP strategy profitable on Solana? (3-month backtest)
Three months of SOL price history, roughly 16,000 data points, replayed against the same LP strategy with different hedging rules. The perp's trading fees — about 0.12% of notional per round trip — are included. Returns are shown relative to running unhedged.
| Strategy | Return vs. no hedge | Max drawdown |
|---|---|---|
| No hedge | baseline | 5.1% |
| Constant 60% hedge | about 60% lower | 2.1% |
| Hedge gated on trend signals | worse than both | up to 6.4% |
| Perfect oracle, 1 step ahead | net loss | — |
| Perfect oracle, ~2 hours ahead | net loss | — |
| Perfect oracle, ~8 hours ahead | less than half | — |
The constant hedge did exactly what a hedge should: it cut drawdown by more than half. It also cut the return by 60%. That is not a flaw — that is the price of insurance, stated honestly.
Why "hedge only when it's falling" does not work
The obvious refinement is to switch the hedge on during downtrends and off during rallies. We tried several variants gated on trend strength and direction. Every one finished behind both simple strategies.
The reason is turnover. The naive gate produced 598 switches over the period, and its fees came to about six times what the hedge saved. A signal that flips often enough to be responsive is flipping often enough to bleed.
What this means in practice
The choice is binary. Either you carry the hedge continuously and accept a lower return in exchange for a calmer equity curve, or you run unhedged and accept the drawdown. There is no clever middle that a signal can find — not in this data.
So our bot ships the hedge as an explicit, off-by-default setting. It is there for people who genuinely need the drawdown suppressed — someone running borrowed capital, or who will close a position in panic at 5% and not at 2%. For everyone else the data says leave it off and pick your range width instead.
The honest caveat
The three months in the sample were a rising market. There was no crash in them. A hedge earns its keep precisely in the scenario this data does not contain, so the true long-run answer may be kinder to hedging than this backtest is. We are reporting what the backtest showed, not a law of nature — and we would rather say that than quietly present a bull-market backtest as a general result.
What the short costs on Jupiter Perps
Two costs, both on the notional of the short. A trading fee of about 0.06% to open and another 0.06% to close. And an hourly borrow fee that floats with how much of the pool's stablecoin custody is lent out: we measured about 6.1% a year on 11 June 2026, roughly $5 a month on a $1,000 short. The bot reads the live rate rather than assuming one. On top of that, the collateral behind the short is capital that is not earning LP fees.
If you do hedge, the mechanics that matter
- Never drive open and close off one threshold. A single boundary with jittery exposure flip-flops and pays fees on every flip. Use a dead band instead — the bot closes below $10 of notional and reopens only above $12.50.
- Size it against real exposure, recomputed as price moves — a static ratio drifts out of neutrality the moment the range does its job.
- Reserve the margin. Collateral for the short is not deployable capital. If you forget that, the LP side quietly eats the margin and the hedge cannot be maintained.
- Count the borrow fee. Jupiter Perps charges a short an hourly borrow fee rather than funding; a rate that looks trivial per hour is not trivial across a quarter of continuous exposure.