Delta-Neutral Liquidity Provision

Putting tokens into a Uniswap pool makes you a market maker who never moves a quote. Traders swap against your inventory and pay you a fee; in exchange, the pool automatically sells you whichever asset is falling and sells away whichever asset is rising. That leaves you long the volatile token in a curved, price-dependent way. Delta-neutral liquidity provision is the attempt to strip that exposure out with a short perpetual future, so the position becomes a pure business: fee income against the money that arbitrageurs extract from a market maker who cannot update. The hedge is the easy part. The extraction is what decides whether the trade works.

What you actually own

In a constant-product pool, the dollar value of your position moves with the square root of the price. Half your value sits in the volatile token and half in the stablecoin, so at any instant your delta is simply the number of tokens the pool is currently holding for you — for a $1,000,000 ETH/USDC position with ETH at $2,500, that is $500,000 of ETH, or 200 ETH. Short 200 ETH of perp and you are flat.

Flat for a moment. The square-root shape means the delta falls as price rises and rises as price falls — you are structurally short gamma, exactly like a short straddle. A static hedge decays into a position immediately.

Worked example: the break-even a pool must clear

The cost of being an un-updating market maker has a name and a size. For a constant-product pool it is loss-versus-rebalancing, and it accrues at roughly $\sigma^2/8$ per year on the value of the position. ETH at 60% annualised volatility gives $0.36/8 = 4.5\%$ a year — that much of the pool leaks to arbitrageurs, hedge or no hedge.

Now put $1,000,000 into ETH/USDC on the 5bp fee tier, in a range of $2,250–$2,750 around a $2,500 spot. A range that tight concentrates capital roughly 20x, so while price sits inside it your position behaves like $20m of full-range depth:

• Leakage: 4.5% of $20,000,000 = $900,000 a year, about $2,466 a day.
• Fees needed to cover it at 5bp: $900,000 ÷ 0.0005 = $1.8 billion of annual swap volume, or roughly $4.9m routed through your range every day.

That is the whole decision, and notice what dropped out: concentration multiplied the fees and the leakage by the same 20x. Tightening a range levers the position; it does not improve the trade.

Worked example: the hedge that became a position

Same $1,000,000 book: 200 ETH of delta at $2,500, hedged with a 200 ETH perp short.

ETH rallies to $2,750 — the top of your range. The pool has sold every ETH you owned into the rally, so the position is now 100% USDC and your delta is zero. The perp short is still 200 ETH. You are net short $550,000 of ETH in an uptrend, earning no fees because you are out of range, and losing $5,500 for every further 1%.

The fix is mechanical: recompute delta from the pool's current state on a price trigger, not from the entry. The good news is that this is cheap. At 60% vol, a 2% rebalancing grid is crossed on the order of $\sigma^2 T / h^2 = 0.36/0.0004 \approx 900$ times a year, and each crossing trades about 1% of a $500,000 delta — a few dollars of taker fee. Hedging costs pennies; the leakage costs six figures. Do not let a rebalancing-cost debate distract from the number that matters.

What erodes the edge

Your quotes are stale by construction. Most of the leakage is taken in the first block after a centralised exchange moves. You are the passive side of every CEX-DEX arbitrage, permanently.

Just-in-time liquidity. A searcher can mint liquidity in the same block as a large swap, collect the fee and burn the position, diluting exactly the volume you underwrote without carrying any of the risk.

Range exits. Out of range you earn nothing and hold 100% of the asset you least want. Wider ranges avoid this and earn proportionally less.

The hedge has its own carry. A short perp collects funding in bull regimes and pays it in bear regimes, so hedging quietly bolts a funding-rate position onto a fee-capture trade.

Everything else on-chain. Gas on mainnet, pool and bridge contract risk, oracle divergence between the perp's index and the pool's price, and venue risk on whichever exchange holds your hedge.

In interviews

Say what the position is before you say how you hedge it: square-root payoff, delta equal to the pool's current token holding, structurally short gamma. Then make the point that the hedge neutralises delta but not curvature, and that curvature has a price called LVR at roughly $\sigma^2/8$ a year. Convert that into the volume-to-TVL break-even out loud, and note that concentration scales fees and losses identically. If you have time, close on the range-exit failure — a hedge that was correct at entry becomes a naked short the moment price leaves the range, which is the operational risk that actually shows up in production.