Basis strategies

What the basis is

Perpetual futures track spot prices through a funding mechanism. Longs and shorts exchange a funding payment at set intervals, often every eight hours, which pulls the perpetual's price towards spot. When the perpetual trades above the spot index, funding is positive: longs pay shorts. When it trades below, funding is negative: shorts pay longs.

The basis trade captures this spread. Holding the correct side of the futures position — delta-hedged against the spot asset — earns the funding rate while hedged against moves in the underlying asset. Venue failure, liquidation, and auto-deleveraging can still break the hedge. Any return comes from funding payments, not from predicting price direction.

Where the perpetual came from

Futures with no expiry date are older than crypto. Ackerer, Hugonnier and Jermann trace the first undated futures market to the Chinese Gold and Silver Exchange of Hong Kong. Its contracts settled against spot every day with an interest payment much like funding, so in effect they were one-day futures rolled over automatically.

Robert Shiller formally proposed the perpetual future in a 1992 discussion paper, published in the Journal of Finance in 1993. He designed it for assets whose prices are hard to measure, such as real estate, with a daily payment between the two sides linked to an index of rents or dividends. Trading in such contracts would help reveal the prices of the assets behind them.

BitMEX brought the design to crypto in 2016, and in the first half of 2023 perpetuals accounted for 75% of listed bitcoin futures volume. The crypto version turns Shiller's purpose around. The spot price is visible every second, and funding exists to pull the contract back towards it. BitMEX itself stopped trading on 23 September 2026. It had said in July that any positions still open at that time would be force-closed, so a contract with no expiry still ended on a date the venue chose.

How funding is actually set

Binance, OKX and Hyperliquid use one formula. Funding equals the average premium of the perpetual over its spot index, plus the gap between a fixed interest term and that premium, with the gap clamped at 0.05% either way. The interest term is 0.01% per eight hours on all three, according to the OKX and Hyperliquid documentation and He and co-authors for Binance. Hyperliquid states that its formula is designed to closely match the centralised venues.

The clamp has a side effect. Whenever the premium is within 0.05% of the interest term, funding prints that term exactly. A long run of 0.01% prints shows only that the premium stayed inside that band, and hides how far demand for leverage moved within it. Hyperliquid describes the constant as standing in for the gap between the cost of borrowing dollars and borrowing the crypto asset, yet it stays fixed when central bank rates move.

Other venues use other rules. Kraken settles funding every hour from the time-weighted premium and spreads each premium over 24 hours, where the formula above spreads it over eight. Its published method has no separate interest term and caps funding at 0.25% per hour. Hyperliquid also pays every hour, one eighth of its eight-hour rate each time, with a cap of 4% per hour.

Intervals can change as well. When an OKX contract settles at its funding cap or floor, settlement steps from every eight hours to four, then two, then one. It returns to the default only when every settlement in the previous twelve hours has been within 0.20% either way. Funding prints from different venues need normalising for interval and formula before they can be compared.

He and co-authors give two reasons why funding differs across venues. Exchanges do not cross-margin with each other, so a trader arbitraging a funding gap must fund and margin each leg separately. Liquidity also differs. From 2020 to 2022 Binance accounted for about 49% of bitcoin perpetual volume across the five venues they study, and Bitfinex for 0.2%. Binance and Bybit funding rates correlate at 0.74 but diverge when payments are unusually large.

Funding and the dated-futures basis are different trades. A dated future sets the spread at entry, and the spread closes at expiry if both legs are held that long, with margin calls along the way. A perpetual pays a floating rate that resets every interval and has no expiry to converge to, so its carry is known only after it has been paid.

Running it systematically

Raw basis exposure — holding the receiving side of funding — has often been positive in crypto. But the rate varies significantly, and it can reverse sharply.

Running it systematically means conditioning the trade every day on three inputs: the current funding rate across exchanges, the persistence of that rate, and cross-exchange basis comparison to identify the best execution venue.

When funding is high and persistent, the position is sized up. When funding compresses or reverses, the position is reduced. The point is not just holding the trade, but holding it when the conditions support it.

What the evidence says

He, Manela, Ross and von Wachter compare hourly Binance perpetual prices with their no-arbitrage values for five large coins from 2020 to March 2024. The gaps are considerably larger than those documented in currency markets and move together across coins. They widen after strong spot gains, consistent with extrapolative demand for leveraged long exposure, and have narrowed over the sample, consistent with growing arbitrage capital and competition. The authors point to limited arbitrage capital as an important cause of the gaps.

Margin rules can move the basis. When Binance introduced portfolio margining in April 2022, letting arbitrageurs net collateral across spot and futures, the gaps compressed significantly. The authors also link the narrowing to the fall of Alameda Research and Three Arrows Capital. Citing news reports, they say both firms appear to have moved from this arbitrage into directional bets around late 2021 and early 2022, before the 2022 price falls exhausted their capital.

Ackerer, Hugonnier and Jermann supply the pricing theory. A perpetual's fair price is the expected spot price at a random future date, and the funding rules set how far off that date tends to be. In their model the perpetual trades at spot when the interest term equals the gap between the two currencies' interest rates. A fixed 0.01% term matches that gap only by coincidence, so a perpetual can sit slightly above or below spot at its fair price.

He and co-authors add the clamp to the theory. When the cash interest rate is below the fixed 0.01% term, their fair price sits slightly below spot, at the point where funding equals the cash rate. The cash rate they use was below that term in 93% of the hours in their sample, and prices in their data converge close to this clamp-adjusted level.

A BIS working paper by Schmeling, Schrimpf and Todorov explains why crypto carry gets so large. Demand from smaller, trend-chasing investors for leveraged upside meets scarce arbitrage capital, and interest rate differentials explain very little of the variation. High carry predicts future price crashes and tends to rise with the price of crash insurance. The authors describe the arbitrage side as risky because margins spike and liquidations cluster in drawdowns, and they find some evidence of segmentation between crypto exchanges and the traditional financial system.

Spot ETFs gave the dated trade a new long leg. After the SEC approved the listing and trading of US spot bitcoin exchange-traded products on 10 January 2024, hedge funds could hold the long leg in a listed product and sell CME futures against it. The Banque de France reports that hedge funds and asset managers were the main institutional holders of these ETFs in 2024, with nearly $18bn and $7bn respectively, and that hedge funds held net short futures as part of this arbitrage. It puts the basis on a one-month futures contract at between 0.3% and 0.8%. Some of the institutional holding visible in ETF filings is therefore one leg of a hedged trade, and it may shrink when the basis does.

A lesson from the Treasury market

The closest traditional analogue is the Treasury cash-futures basis trade. Hedge funds buy notes and bonds, finance them in repo and sell futures, with the other side largely asset managers holding long futures. Barth and Kahn find the trade grew after 2016 to as much as half of hedge funds' Treasury positions and about a quarter of dealers' repo lending.

For about two weeks in mid-March 2020, government bond markets were in turmoil. Volatility and futures margins jumped, and the New York Fed's Lorie Logan said changes in margin requirements likely contributed to a partial unwind. The BIS described forced selling by investors who had used leverage to exploit small yield differences. The Fed bought around $1.6 trillion of Treasuries between 13 March and 10 June 2020.

Barth and Kahn judge that the unwind was likely a consequence of the stress rather than its main cause, and that prompt Fed action may have kept it from accelerating. The trade lives on small price gaps, so margin calls and financing decided who had to sell. The trade has grown since. The Dallas Fed reports that hedge funds' net repo borrowing reached about $1.8 trillion by the end of 2025, and that the basis trade has become the primary driver of their Treasury holdings.

Crypto has the same shape with a different backstop. The other side is leveraged demand for long exposure, and the financing is margin posted on the exchange. No central bank stands behind crypto markets in a crash. The backstops are the venue's insurance fund and, behind it, auto-deleveraging.

The risks

Basis compression. Funding rates can collapse quickly in a deleveraging event. Many desks hold the same trade and exit at the same time. The exit price may be worse than the prevailing rate suggests.

Exchange risk. The capital sits on the exchange hosting the perpetual. Running the trade across multiple exchanges reduces concentration but does not eliminate exchange counterparty risk.

Regime reversal. In a sustained bear market, funding can be persistently negative. The strategy would then pay funding rather than receive it. Regime signals are used to reduce or close the position when they point toward a negative-funding environment.

Traps that are easy to miss

Auto-deleveraging can close the hedge. When a liquidated account falls below zero and the venue's backstop cannot absorb the loss, the venue closes positions on the winning side. Hyperliquid ranks profitable positions by unrealised profit and leverage and closes them at the previous mark price. In a crash the winning side includes the short legs of cash-and-carry trades, and low leverage lowers a position's place in the queue without removing it. The spot leg is then left unhedged while prices are still falling.

On 10 and 11 October 2025, liquidations exhausted insurance funds and several venues invoked auto-deleveraging at once. On Hyperliquid, Tarun Chitra describes a 12-minute cascade in which about $2.1bn of positions were closed. Hyperliquid's on-chain log for the day records 34,983 auto-deleveraging executions across 19,337 wallets.

Chitra estimates that Hyperliquid's mechanism took $45m to $51.7m more profit from winning traders than the shortfall required. His method also suggests that Binance overused auto-deleveraging far more than Hyperliquid. He notes that most such episodes, including March 2020 and May 2021, closed out highly profitable but lower-leverage traders. For scale, he cites mid-2025 estimates of about $300m and $150m for the OKX and Bybit insurance funds.

The risk engine may not apply to everyone. The CFTC alleged that FTX marketed its auto-liquidation engine and backstop liquidity programme as features that limited risk, while Alameda, a leading backstop provider, was exempt from that engine and could not be liquidated under any conditions. It also alleged that over $8bn of customer assets were missing when FTX collapsed in November 2022. In the bankruptcy, claims were valued as of the 11 November 2022 petition date. A basis trader's short perpetual on FTX became a fixed dollar claim, while the spot leg held elsewhere stayed exposed to the price.

The failure also showed up in funding. As FTX collapsed, bitcoin funding turned sharply negative on all the solvent exchanges, while perpetuals on FTX traded well above spot, according to He and co-authors. They read this as FTX users closing short futures, either by choice or through forced liquidation. A short perpetual on a healthy venue then paid funding instead of receiving it.

Stablecoin margin carries issuer risk. On 10 March 2023 Circle said it could not wire out $3.3bn of USDC reserves held at Silicon Valley Bank, out of about $40bn. USDC fell below 90 cents and recovered over three days, according to a Federal Reserve note. A book margined in stablecoins carries credit exposure to the issuer and its banks.

Some of the crowding is visible. Ethena's USDe, a synthetic dollar, has held its backing mainly in spot crypto hedged with short perpetual futures, with the rest in unhedged stablecoins and short-dated real-world assets, according to its own documentation. Its supply therefore gives a public view of this trade held at scale, and the New York Fed notes that it grew to nearly $15bn between early 2024 and mid-2025. The New York Fed adds that redemptions force the protocol to buy back futures and sell ether, and that USDe's market value fell by more than 13% after a sharp funding reversal in October 2025. On 10 October it briefly traded at $0.65 on Binance while its on-chain peg held elsewhere, a gap the New York Fed links to Binance's price oracle.