When Funding Rate Arbitrage Loses Money
The pitch for funding rate arbitrage is that it is market-neutral: you hold spot, you short the perpetual, price direction cancels out, and you collect the funding payment. All of that is true, and none of it makes the strategy safe.
What a delta-neutral position removes is directional risk. What it leaves — and in some cases amplifies — is a set of risks that have nothing to do with whether the asset goes up or down: funding inverting, the spread compressing, execution costs, margin mechanics, and the venues holding your capital. Each of these has produced real losses at scale within the last four years, and the largest of them destroyed positions that were hedged correctly.
This piece walks through five ways the trade loses money, with the episode or the arithmetic behind each. If you want the mechanics of the strategy itself first, the funding rate arbitrage guide covers construction, sizing and worked returns; this is the companion piece on what goes wrong. Terms are defined in the glossary.
Throughout, the illustrative position is the one from that guide: $100,000 of capital, deployed as $60,000 of spot BTC against a $60,000 notional short perpetual with $40,000 of margin, earning $18 a day at baseline funding of 0.01% per eight hours. Figures are illustrative, not projections. Capital deployed in this strategy is at risk.
1. Negative funding regimes
When perpetuals trade below spot, the funding rate inverts: shorts pay longs. A long-spot/short-perp position stops collecting and starts paying.
The frequency of this is the part that gets understated. Bitcoin funding was negative for 46 consecutive days through mid-April 2026 — the longest sustained streak since November 2022 (Phemex). On 16 April 2026 the seven-day average reached roughly −0.005%, the most negative reading in about three years (CoinDesk, citing Glassnode). The comparable episode after the FTX collapse ran around 50 days, from November into December 2022.
On the illustrative position, −0.005% per interval is about $9 a day paid out. Across 46 days that is roughly $414, against a book that earns $18 a day when funding is at baseline. The arithmetic that matters is not the size of the loss — it is that the strategy needs about 23 days of normal positive funding to recover 46 days of mild negative funding. A month and a half of inverted funding costs a quarter of a year.
And these streaks arrive when they are least convenient. Deeply negative funding means the market is crowded short, which means it usually coincides with a drawdown, redemption pressure and stressed margin. A strategy without a written answer for this regime — reduce exposure, rotate into assets with positive funding, or hold and absorb — does not have a risk control. It has a hope.
2. Basis compression: the slow one
The failure above is visible. This one is not, and it has done more damage to realised returns.
Funding is a price. It exists because traders want leveraged long exposure and are willing to pay for it. When more capital arrives to supply the other side, the price of that leverage falls — and every dollar allocated to funding arbitrage is a dollar supplying the other side. The trade competes with itself.
Two things made this concrete rather than theoretical.
First, exchanges industrialised the short side. BitMEX's review of 2025 attributes the collapse in funding to an oversupply of structural shorts from exchange-native delta-neutral products such as Binance's BFUSD, with rates trending consistently below the 0.01%-per-eight-hours baseline for the first time in a bull cycle, and mid-2025 carry sitting sub-4% — "killing the funding rate trade" (BitMEX). When the venue itself manufactures the short side at scale, the independent arbitrageur is competing with the house.
Second, the largest operator left. Ethena held roughly 5% of global ether perpetual open interest in February 2024, with its founder estimating capacity constraints somewhere around 30–40% (The Block). It never reached that. By 2026, funding had cooled from 2024 highs to the point that "the trade that once paid double digits now pays single digits," and Ethena had rotated its reserves into DeFi and institutional lending, retaining only a small slice of open interest (Altitude).
Neither of those is a market accident. They are what a crowded trade looks like from the inside. The practical consequence for an allocator is simple: any return assumption calibrated on 2024 funding levels is describing a market that no longer exists. Ask what period a track record covers before reading its numbers.
3. Execution slippage
The spread being harvested is measured in basis points. Execution costs are measured in the same units, which makes them a first-order term rather than a rounding error.
Assume 15bps of combined slippage across the two legs on entry — realistic in size, in a fast market, or on a less liquid pair. On $60,000 of notional that is $90. At baseline funding earning $18 a day, that single entry has consumed five days of funding. On a 30-day hold, one poor fill removes a sixth of the gross return before fees are counted.
Two consequences follow. The strategy scales badly downward: below roughly $25,000 of capital, round-trip costs take enough of the spread that the trade is hard to justify. And it scales badly upward in a different way — large positions move the book, so the slippage rate itself rises with size. There is a capital band in which this works, and it has a floor and a ceiling.
4. Liquidation and auto-deleveraging
This is the mechanism that deserves the most attention and gets the least, because it is the one that kills positions that were hedged correctly.
The ordinary version is familiar: a violent rally drives the short leg's margin below maintenance, the position is liquidated, and you are left holding unhedged spot. The hedge was economically sound — your spot gains matched the short's losses — but the gains were on a different balance sheet from the margin call. Liquidation converts a market-neutral position into a directional one at the worst moment.
The version most allocators have never priced is auto-deleveraging. ADL is the backstop after liquidation and the insurance fund have both failed: when a liquidated position cannot be filled near its bankruptcy price and the insurance fund is exhausted, the exchange force-closes profitable positions on the other side to make the book balance. Exchanges rank candidates by unrealised profit percentage and effective leverage, taking the highest first (Arch).
Read that ranking again in the context of this strategy. A funding-arbitrage short that is working — sitting on unrealised profit during a sell-off, running leverage against dedicated margin — is close to a perfect description of ADL's first target. The better the hedge is performing, the earlier it gets taken.
This is not hypothetical. On 10–11 October 2025, a macro shock triggered over $19 billion of liquidations across 1.6 million accounts, with $3.21 billion liquidated in a single minute at 21:15 UTC — roughly nineteen times the scale of the March 2020 crash (CoinGecko). The event "did not just liquidate retail traders," and ADL force-closed the short legs of market makers running delta-neutral books, leaving them holding unhedged spot into a falling market (Arch). Funding arbitrage across the market fell below 4% afterwards as liquidity providers withdrew.
No venue failed that day. The hedges were correct. The exchange's own risk machinery unwound them anyway.
5. Exchange and counterparty failure
To run this strategy, capital must sit on trading venues. That is structural, not a choice, and it is the largest tail exposure in the trade.
The canonical case is FTX, which halted customer withdrawals on 8 November 2022 (CoinLedger). Every hedged position with a leg on that venue became, instantly, an unhedged position plus a bankruptcy claim — irrespective of how well it had been constructed.
October 2025 supplied a subtler version. During the crash, USDe traded down to $0.65 on Binance specifically — about 35% below peg — between 21:36 and 22:15 UTC, while holding its peg elsewhere, because Binance priced from internal order books rather than external oracles. BNSOL and WBETH lost their pegs on the same venue. dYdX went offline for eight hours; Lighter for four and a half (CoinGecko).
If your margin collateral is priced 35% below its real value for forty minutes, your correctly-hedged position is margin-called on a pricing artifact. If your venue is offline for eight hours, you cannot rebalance a position whose hedge ratio is drifting. Neither of those requires anyone to become insolvent.
This risk is not diversifiable in the way allocators expect. Holding more assets does not reduce it, because the exposure is to the venue, not the asset. Spreading across more venues lowers concentration but raises the number of counterparties. There is no configuration that removes it — only configurations that trade one shape of it for another.
What a realistic return distribution looks like
Put the five together and the shape of the return becomes clearer than any headline number.
The central case is a modest positive carry that has compressed materially since 2024 and now sits far closer to the baseline than to the double digits the strategy was marketed on. Around that central case sit stretches of weeks where the carry is zero or negative and the position pays to stay open. Underneath both is a thin left tail — ADL, a venue failure, a collateral mispricing — where the loss is not a bad month of carry but a step change in capital, and where it arrives precisely when correlations go to one.
That is not a normal distribution around an advertised yield. It is a small, variable positive return with occasional negative months and a rare large loss — much closer in shape to selling insurance than to earning interest, which is why comparing it to a deposit rate is the single most misleading thing done with this strategy. Any presentation of funding arbitrage as risk-free is wrong, and a "target range" quoted without the distribution behind it tells you nothing about what you are actually holding.
The useful questions for anyone evaluating a manager running this strategy are therefore narrow and answerable: what period does the track record cover, and what were funding rates during it? What is the written policy when funding inverts for six weeks? Which venues hold collateral, in what proportion, and what is the ADL exposure on each? How is the return split between funding and basis convergence?
If you are weighing this strategy against other non-directional crypto yield, funding arbitrage vs stablecoin yield compares the two on where the return comes from and what each one actually risks. If you are evaluating how we handle these specific exposures, our strategy page sets out the position limits and venue policy, and how it works covers the settlement and monitoring cycle.
By VectorTraders Editorial Team · Published SEP 14, 2026
This article is educational and is not investment advice. Funding rate arbitrage involves risk of loss, including total loss of capital deployed on trading venues. Figures shown are illustrative and are not projections of return.