Late on 10 October 2025, Binance briefly quoted USDe, a token built to hold one dollar, below 66 cents. Traders who had posted it as collateral were hurt between 21:36 and 22:16 UTC, and Binance paid about $283 million in compensation two days later. On Curve the dip was about 0.3%, by one investor’s count. USDe rests on the plainest delta neutral strategy there is: hold an asset, short the same amount through a perpetual, collect what the longs pay. Mint and redeem kept working that night, Ethena says. The damage came from one venue’s price.
A hedged trade is indifferent to small price moves, and that is all it is. Four other exposures stay behind: funding or basis, the venue, the margin on the short leg, and the costs of execution and tax.
Four versions of the trade are priced here with dated figures from 31 October 2026 and examples labelled as examples: a staked ether position with a short perp, the basis trade on dated futures, the cash and carry through Ethena, and a hedged liquidity position. A word of warning first. This lesson is for readers who already use perps and understand margin. If you have never watched a position get liquidated, read it for the mechanics and do not copy the trades: a small mistake in sizing can lose the whole margin in hours. Nothing here is investment advice. The previous lesson, on farming airdrops and points, may have left you with a position whose price risk you do not want, and the same hedge can sit under it.
Delta neutral strategy in plain terms: the price term goes to zero
Delta is how much a position gains or loses when the asset moves by one dollar. Long 1 ETH has a delta of +1. Short 1 ETH through a perpetual has a delta of -1. Hold both and the sum is zero: ether rises $100, one leg makes $100, the other loses $100. What remains is whatever the two legs pay or cost you while you wait.
Two limits come with the word. Delta is a first-order measure, so it holds for small moves, and it can drift as price moves. That drift is called gamma: it is zero for a spot coin against a linear perp and very much not zero for a liquidity position, as trade 4 shows. And delta neutral is narrower than market neutral. A market-neutral book aims to have no exposure to the market as a whole across many positions; a delta-neutral position only cancels the price of one underlying.
A reminder rather than a repeat: our guide to on-chain perpetuals and funding explains how a perp stays near spot. Funding is the payment between longs and shorts that does the job. On Hyperliquid it is paid every hour, with an interest component fixed at 0.01% per eight hours. Multiply that out and a perp at fair value has longs paying shorts 10.95% a year on simple arithmetic (the docs quote 11.6% APR).
Who pays you when price does not matter
No neutral position creates money. Every return in this lesson has a payer, and naming the payer is the fastest test of a yield.
- Funding comes from traders who are long with leverage and agree to pay for it.
- The premium on a dated future comes from buyers who want exposure later and pay above spot to get it without tying up the cash.
- Staking rewards come from the Ethereum protocol and from fees paid by its users.
- blockchain that runs exactly as written when it is called, holding and moving funds according to its code, without an intermediary able to change the outcome.">smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">Pool fees come from swappers, and part of what they pay goes back out to arbitrageurs.
Funding is the least reliable of the four. The table shows Hyperliquid data from its public API, computed by us over a fixed window: 90 UTC days from 2 August to 30 October 2026, or 2,160 hourly payments per coin, annualised by simple multiplication. Long-run figures from Ethena’s funding risk documentation sit next to them.
| Funding, annualised | BTC | ETH |
|---|---|---|
| Hyperliquid, one hour, 31 Oct 2026, about 10:05 UTC | 10.95% | 4.35% |
| Hyperliquid, mean, 2 Aug to 30 Oct 2026 | 8.65% | 8.56% |
| Hyperliquid, hours with negative funding, same 90 days | 8.0% | 7.6% |
| Hyperliquid, days with negative funding, same 90 days | 2 of 90 | 3 of 90 |
| Hyperliquid, hourly range, same 90 days | -14.0% to +78.6% | -67.2% to +73.1% |
| Ethena docs, average over three years, data to 31 Dec 2024 | 7.80% | 9.15% |
| Ethena docs, days with negative funding, data to 31 Dec 2024 | 15.9% | 17.5% |
One hour is a snapshot, not a yield, and these numbers cover one venue. BTC and ETH both closed 11 October 2026 negative; ETH did so again on 30 October. Ethena puts the longest run of negative days in its data at 13. Its docs also carry a line to keep in mind for every trade below: “Negative funding rates are a feature, rather than a bug of the system.”
Trade 1: long a liquid staking token, short a perpetual
You buy stETH, which keeps earning staking rewards, and short the same ether amount on a perp venue. Our lesson on liquid staking tokens such as stETH covers the token itself. The attraction is two incomes on one position: staking on the long leg and funding on the short leg.
A worked example with the numbers of 31 October 2026
Example only. Ether at $2,243, the rounded Hyperliquid mark at about 09:50 UTC. You buy 10 stETH for $22,430 and short 10 ETH, posting $4,500 in USDC as isolated margin, about 5x. Total capital: $26,930. Staking is taken at 2.5%, inside the range of Lido daily readings between 24 and 29 October 2026 (2.24% to 2.78%), and funding at the 90-day ETH mean of 8.6%. Lido’s feed then printed 15.08% for 30 October, and DefiLlama repeats it; we treat that single reading as an unexplained outlier, not a yield. The yearly results assume these rates hold, which they will not.
| Funding over the year (example) | Funding income | Staking income | Total | On $26,930 capital |
|---|---|---|---|---|
| 8.6% (90-day mean) | $1,929 | $561 | $2,490 | 9.2% |
| 4.3% (half) | $964 | $561 | $1,525 | 5.7% |
| 0% | $0 | $561 | $561 | 2.1% |
| -5% | -$1,122 | $561 | -$561 | -2.1% |
Costs come off that. At Hyperliquid’s base taker fee of 0.045% the short costs about $10 to open and $10 to close. Assume 0.1% each way for the stETH leg, $22 and $22, plus gas, and a round trip takes about $65, or a quarter of a point of capital. That is small over a year and large if funding flips after two weeks and you leave.
What the short does not hedge
The perp hedges ether against the dollar. It does nothing for stETH against ether. Lido’s integration guide says tokens are minted at 1:1 and warns that the exchange price can sit below the inherent value. On 14 June 2022, The Block reported it at almost 7% below ETH, with the main Curve pool at 80% stETH and 20% ETH, as wallets labelled Three Arrows Capital sold. Applied to this example, a forced exit at a 7% discount costs about $1,570, or more than seven months of the income in the top row.
The second gap is mechanical. The two legs live in different places, so the stETH gain cannot top up the margin. Hyperliquid sets maintenance margin at half the initial margin at maximum leverage. ETH’s maximum was 25x on 31 October 2026, which gives 2%. With $4,500 on a $22,430 short, the position is closed near $2,640, a rise of about 18%, before counting any funding received. You would then hold 10 stETH with no hedge and most of the $4,500 gone, just after a sharp rally. Our guide to how liquidations work explains the process. Lower leverage pushes that price away and drags the yield on capital down: every extra dollar of margin earns nothing.
Ethena shows the professional answer. It says it “does NOT trade with any material leverage”, and its liquidation risk page works an example in which the stETH spread would have to reach 65% before a liquidation. That figure assumes an exchange valuing stETH 1:1 with ETH; with a 0.85 haircut the page gives about 25% to 41%. The same page says the discount peaked at 8% before Ethereum’s Shapella upgrade and has not passed 0.3% since.
Add slashing and smart-contract risk on the staking token. Lido’s guide says balances can fall if validators are slashed, and that daily rebases had never been negative at the time of writing. That is a record, not a promise.
Trade 2: the basis trade, where the yield is known at entry
A dated future settles to the index at expiry. If it trades above spot today, you can buy spot, short the future, and collect the gap as it closes, whatever price does in between. Unlike funding, the figure is known the moment you enter, as long as you hold to expiry.
The table uses OKX quarterly BTC-USD futures on 31 October 2026 at about 10:06 UTC, taking the midpoint of bid and ask against an OKX index of 77,274.6. The annualised column is our own simple calculation, premium over the time left to the 08:00 UTC expiry. We left out the December 2027 contract: its book showed 79,000 bid and 83,002.5 ask, and its last trade was a poor guide to anything. Check the spread before you trust a quoted basis.
| Expiry | Future (mid) | Premium | Days left | Annualised |
|---|---|---|---|---|
| 27 Nov 2026 | 77,609.7 | 0.43% | 26.9 | 5.9% |
| 25 Dec 2026 | 77,926.2 | 0.84% | 54.9 | 5.6% |
| 26 Mar 2027 | 79,000.7 | 2.23% | 145.9 | 5.6% |
| 25 Jun 2027 | 80,065.6 | 3.61% | 236.9 | 5.6% |
| 24 Sep 2027 | 81,158.7 | 5.03% | 327.9 | 5.6% |
Example with the March contract: buy 1 BTC at 77,274.6, short 1 BTC at 79,000.7. If bitcoin ends at $60,000, spot loses $17,274.6 and the short makes $19,000.7. If it ends at $100,000, spot makes $22,725.4 and the short loses $20,999.3. Both roads give $1,726.1, before fees. These OKX contracts are inverse, margined in bitcoin, so their dollar profit is not linear in price; a linear contract behaves as the example does, and an inverse one needs its size adjusted.
The hurdle is the Treasury bill, not zero
A locked 5.6% is only interesting against what cash pays. The 13-week US Treasury bill stood at 4.08% (coupon equivalent) on 30 October 2026. That leaves about 1.5 points for carrying exchange risk and margin. The Block made the same comparison on 9 April 2026: basis near 5% against about 4.5% risk-free, CME bitcoin open interest averaging $7.2 billion in early April, the lowest since February 2024, partly because the institutional version, buying spot ETFs and shorting CME futures, was being unwound. We have no October 2026 CME basis figure to give.
For a US reader, CME is a regulated road for this trade, and the ETF makes the long leg simple. For everyone, the catch is margin. If bitcoin rises 25%, the short shows a paper loss near $19,750 per coin, which must be funded in cash at once while the matching gain sits in a custody account or a fund. Basis desks that are forced to close early give up the convergence they were waiting for, and buying back the short in a squeeze can cost more than the premium.
On-chain, the perp stands in for the dated future
On-chain derivatives venues mostly list perps, so a DeFi user runs the same idea with funding in place of convergence. The trade-off is plain from the data above: 5.6% fixed to March on OKX against a Hyperliquid 90-day mean of 8.6% that nobody can promise for the next 90 days. One is a known price for a term. The other is a floating rate with a higher recent average.
Trade 3: the cash and carry through Ethena, and what a depositor earns
Our guide to Ethena USDe covers the mechanics, so one line here: USDe is backed by spot assets and an offsetting short derivatives position, and sUSDe is the staked form that receives the yield. Buying sUSDe hands trades like number 1 to a protocol, at least in part. The question is what reaches you.
| Ethena figure (DefiLlama) | Value |
|---|---|
| sUSDe APY, 31 Oct 2026 | 5.04% |
| sUSDe mean, last 30 days to 31 Oct 2026 | 4.98% |
| sUSDe mean, last 90 days to 31 Oct 2026 | 4.73% |
| sUSDe mean, last 365 days to 31 Oct 2026 | 4.30% |
| sUSDe highest reading, 7 Mar 2024 | 55.87% |
| Ethena TVL, 31 Oct 2026 | $4.95bn |
| Ethena TVL, 1 Nov 2025 | $9.50bn |
| Ethena TVL, record of 3 Oct 2025 | $14.98bn |
| USDe in circulation, 31 Oct 2026 | $4.51bn |
Example: $10,000 in sUSDe earns $504 over a year at the current rate, $430 at the 365-day mean. That is about one point above the Treasury bill, with protocol risk on top. These are DefiLlama’s reported rates, not a promise from anyone.
Funding averaged 8.6% on Hyperliquid while sUSDe paid about 5%. Part of the answer is that sUSDe is no longer a pure funding trade. Ethena’s docs describe moving backing into liquid stablecoins earning about the Treasury rate when funding is low, plus a reserve fund that steps in when revenue is negative so the loss is not passed to stakers. In a 2026 proposal, reported by The Defiant, Ethena said perps backed only about 11% of USDe. We could not read a live backing split or a reserve balance on 31 October 2026, so we do not break the gap down. TVL stands at a third of its peak: the data shows the exit, not the reasons for it.
Two terms matter to a depositor. Direct mint and redeem are reserved for approved market-making counterparties that clear KYC/KYB checks, so most holders leave through a market. And unstaking sUSDe runs through a cooldown. The long-standing period was 7 days, and DefiLlama still labels the pool that way, but the sUSDe contract returned 1 day when we read cooldownDuration on 31 October 2026. The docs let the admin set it as high as 90 days, so the exit can lengthen after you enter. During a cooldown you can sell sUSDe only if an outside market will take it.
10 October 2025: the price that broke furthest on one venue
USDe fell below $0.66 on Binance, according to The Block, and did not lose nearly as much elsewhere. BNSOL and WBETH lost their pegs on the same venue. The Block puts the market lows at 21:20 to 21:21 UTC; Binance says the severe depegging came after 21:36. Compensation went to futures, margin and loan users who held those assets as collateral in the 21:36 to 22:16 window, plus users with verified losses on internal transfers or Earn redemptions, and Binance said it was still reviewing cases.
Ethena Labs chief executive Guy Young disputes the word itself: “I do not think it is accurate to describe this is a USDe depeg when a single venue was out of line.” He said the deepest pools showed nothing unusual and, as CryptoSlate reported, that mint and redeem kept working and handled about $2 billion of redemptions in 24 hours. Dragonfly’s Haseeb Qureshi gave a less tidy picture in the same report: USDe “wicked down on every CEX”, though unevenly, with Bybit touching $0.95, while Curve dipped only about 0.3%. Those are their accounts, and we did not check them on-chain.
Binance answered with two index changes: redemption prices in the index weights and a soft floor for USDe in the reference index. The lesson for a delta-neutral trader is narrow. Users with leverage lost out because a stablecoin was valued at the price of a thin book on one venue. Use a yield token as margin and your risk becomes the venue’s price feed.
Trade 4: hedging a liquidity position, the one with real gamma
A reminder first. Providing liquidity earns fees, and impermanent loss is the shortfall against simply holding. What follows is the part those lessons leave out: what happens when you try to hedge the price away.
The delta of an LP moves against you
A constant-product pool holds half its value in each asset, so its delta is about half the position in the volatile token. Put $20,000 into an ETH-USDC pool with ETH at $2,000 (example price) and you hold 5 ETH and $10,000. A short of 5 ETH zeroes the delta at that price. The trouble is that the pool sells ether as it rises and buys as it falls, so your 5 ETH hedge stops being right as soon as price moves. The table shows the result with no rebalancing.
| ETH at end (example) | Pool value | Short 5 ETH | Net result | Impermanent loss, share of holding |
|---|---|---|---|---|
| $1,000 | $14,142 | +$5,000 | -$858 | 5.7% |
| $1,600 | $17,889 | +$2,000 | -$111 | 0.6% |
| $1,800 | $18,974 | +$1,000 | -$26 | 0.1% |
| $2,200 | $20,976 | -$1,000 | -$24 | 0.1% |
| $2,400 | $21,909 | -$2,000 | -$91 | 0.4% |
| $3,000 | $24,495 | -$5,000 | -$505 | 2.0% |
The loss comes in both directions and, in dollars, it equals impermanent loss exactly. That is the signature of negative gamma: the position behaves like a sold option. A fixed hedge does not remove impermanent loss. It strips out the price gain or loss around it and leaves that shortfall standing alone.
Concentrated liquidity makes this sharper. Inside a range the mix shifts faster, the position can end up entirely in one asset at the edge, and past that edge it earns no fees. Delta swings from long to flat over a short distance, so the hedge has to move more often.
Why rebalancing is not free
Rebalancing means trading after price has already moved: you add to the short after a rise and cut it after a fall. Each step locks in a small loss. The paper that named this cost, Automated Market Making and Loss-Versus-Rebalancing by Milionis, Moallemi, Roughgarden and Zhang, finds that over 99.991% of LP return variance in the Uniswap v2 ETH-USDC pool it studied comes from market exposure. Hedge that out and the bet left is fees minus what arbitrageurs take. For a constant-product pool the paper puts that loss, called LVR, at one eighth of the variance as a share of pool value per unit of time. With annual volatility that gives 4.5% a year at 60% and 8% at 80%.
Now price it. DefiLlama showed the Uniswap v2 USDC-WETH pool on Ethereum at 2.99% fees and a 30-day mean of 7.39% on 31 October 2026, on about $19 million of TVL. Take the $20,000 position. The short is about $10,000, so 8.6% funding brings $860. At 60% volatility a 5% rebalancing band is hit about 144 times a year (0.36 divided by 0.0025). Each trade moves about $250 of short, so taker fees at 0.045% come to roughly $16; we round up to $20 for slippage. Volatility here is an assumption, not a measurement.
| Example over one year | Fees | Funding | LVR | Rebalancing fees | Net on $20,000 |
|---|---|---|---|---|---|
| Fees 7.39%, funding 8.6%, volatility 60% | +$1,478 | +$860 | -$900 | -$20 | +$1,418 (7.1%) |
| Fees 2.99%, funding 8.6%, volatility 60% | +$598 | +$860 | -$900 | -$20 | +$538 (2.7%) |
| Fees 7.39%, funding 0%, volatility 60% | +$1,478 | $0 | -$900 | -$20 | +$558 (2.8%) |
| Fees 2.99%, funding 0%, volatility 80% | +$598 | $0 | -$1,600 | -$20 | -$1,022 (-5.1%) |
Execution fees are the smallest line, so the usual complaint about rebalancing costs misses the point: gamma is the bill. Much of the return in the good rows is funding, which makes a hedged LP partly trade 1 in another form. The margin for the short, perhaps $3,000, is extra capital the table does not count. Higher fee rates on concentrated pools are no free gift either, since the same concentration raises LVR. The Uniswap v3 0.05% pool showed 13.88% fees that day, and we have no dated yield for a hedged on-chain vault to set beside it.
What a delta neutral strategy pays: the four trades side by side
Indicative figures as of 31 October 2026, gross of tax. The first and last rows come from the examples above, so they depend on the assumptions stated there.
| Trade | Who pays you | Indicative gross yield | What breaks it |
|---|---|---|---|
| stETH plus short perp | Stakers’ rewards and leveraged longs | 2.1% to 9.2% on capital, funding 0% to 8.6% | Funding turning negative, stETH discount, short liquidated in a rally |
| Basis on dated futures | Buyers of forward exposure | 5.6% to March 2027, against 4.08% on bills | Margin call before expiry, exchange failure |
| sUSDe | Funding and income on the backing assets, less what Ethena retains | 5.04% reported, 4.30% over 365 days | Long negative funding, custody or exchange failure, longer cooldown or a thin market on exit |
| Hedged LP | Swappers, less arbitrageurs, plus funding | -5.1% to +7.1% in the examples | High volatility, low volume, negative funding |
The hidden risks shared by all four
Exchange counterparty risk comes first. Every short sits somewhere. On a centralised venue your margin and profit are a claim on that firm, and if it fails the hedge disappears while the long stays, which leaves you with full price exposure. Ethena’s answer is off-exchange settlement: backing stays with providers and the exposure is limited to unsettled PnL between cycles, daily in the Copper example. A retail account generally has no such structure, and an on-chain perp venue swaps this risk for contract and validator risk.
Funding can stay negative for days, 13 in a row at worst in the Ethena data above. Leaving has its own price: closing two legs in a thin market can cost more than the days of carry you were trying to save.
The short leg can be liquidated while the position is fine on paper, as trade 1 showed. Rallies are the danger, not crashes, which feels backwards to a reader used to long positions.
Execution has its own traps. The two legs are not filled at the same instant, so you carry naked delta for seconds or minutes. Taker fees, slippage on the long and gas all come off before the first funding payment arrives. Capital is split across venues and part of it earns nothing. Oracles add a last layer: liquidations follow a mark price, which on a stressed venue may not be the price you see elsewhere, as 10 October 2025 showed.
Rules and tax: funding has no settled treatment
US derivatives on crypto fall under the CFTC, and CME is one regulated venue. Tax is another matter. A Bloomberg Tax article of 11 September 2026 by attorney Rustin Diehl says no IRS guidance, act of Congress or tax court decision addresses bitcoin perpetuals directly, even when they trade on a US contract market. The author sees a plausible Section 1256 argument and names its main obstacle: that section excludes swaps, and a perpetual may be one, which would bring the notional principal contract rules into play. So the character of funding, received or paid, is unsettled, and more so on offshore venues. Dated CME futures are on firmer ground; see our page on crypto taxes in the United States.
In the UK, the FCA banned the sale of crypto derivatives to retail consumers from 6 January 2021, and it kept that ban when it reopened crypto ETNs on 8 October 2025. HMRC’s cryptoassets manual treats a company’s derivative as typically a derivative contract under Part 7 CTA 2009 and sends other cases to separate guidance, without a word on funding or perpetuals. Our UK crypto tax guide has the general rules.
Article 2(4) of MiCA, the EU regime, says it does not apply to crypto-assets that qualify as financial instruments, and derivatives on crypto go through MiFID II. ESMA said on 24 February 2026 that products marketed as perpetual futures are likely to fall under the national CFD measures, with leverage limits and negative balance protection for those that meet the definition, and that the commercial name is irrelevant. Treatment of LP fees and staking on the long leg varies again by country. Ask a tax adviser before opening a position, not after, because a hedged trade can show a gain on one leg and a loss on the other, and they may not offset.
Before you open a delta neutral strategy
- Write down who pays you and what happens to your position if that payer stops.
- Compare the yield with the Treasury bill rate on the same day. A spread of one or two points may not pay for the risks above.
- Compute your liquidation price on the short and the rally it would take, then decide how you would top up margin at 3 a.m.
- Add every cost: both legs, in and out, gas, and tax.
- Rehearse on a testnet where one exists. Hyperliquid has one, with 1,000 mock USDC for addresses that have made a mainnet deposit. It teaches the mechanics; it will not teach you funding.
- If you go live, use a small amount and low leverage, on a venue you may legally use, and accept that the gas and fees can take a large share of a small position.
- Decide your exit rule in advance, for example how many days of negative funding you accept.
The figure to watch is funding: when the 90-day mean falls towards the bill rate, as it did for the CME basis in April 2026, these trades stop paying for their risks and the capital tends to leave.
