Perpetual Premium Index vs Basis: What the Numbers Mean

A market screen says Bitcoin perpetuals trade above spot. Before calling that difference an opportunity, identify what was subtracted from what. A last trade can be above a spot index while the best bid is lower. An average sale price for a larger order can be lower still. Each number answers a different question.

All three figures can be correct at the same moment. The first describes a quotation relative to a benchmark. The second measures one ingredient of funding. The third uses the sides of the two markets required to open a short-perp/long-spot pair. The difference between them is not an arithmetic error; it comes from changing the measurement.

Throughout this guide, basis means futures price minus spot price, with percentage basis divided by the stated spot reference. Some research uses the opposite sign convention. When a chart labels a discount as positive, read its definition before assuming that it contradicts the exchange.

An impact price asks what average price a defined order value would obtain from the displayed book. For an impact bid, imagine selling into bids. For an impact ask, imagine buying from asks. The exchange chooses the order value used by its formula; your own intended order may be much larger or smaller.

Here is a deliberately small book. The hypothetical impact value is 20,000 USDT. There are 0.2 BTC available to sell at 50,050 and another 0.2 BTC at 49,950. Selling the first portion raises 10,010 USDT; the second raises 9,990 USDT. Together, 0.4 BTC produces exactly 20,000 USDT, so the impact bid is 50,000.

The best bid is 50,050, but that price only covers half of the required BTC. Using it for the entire notional would overstate the average sale price. In an uneven book, do not simply average the printed price levels either: a level offering 0.01 BTC should not have the same influence as a level supplying 0.39 BTC.

For the ask side, suppose a corresponding walk through the offers produces an impact ask of 50,120. With the index at 49,900, both impact prices sit above the reference. The positive bid-side excess is 100 USDT. There is no negative ask-side contribution, because the index is not above that impact ask.

The max operation means “keep the amount if it is positive; otherwise use zero.” It prevents an ask that is simply above the reference from being counted as a negative premium. This is the impact formula documented by OKX and Bybit; it should not be substituted into every perpetual product on the market.

What happens when the index sits inside the impact spread?

With an index of 50,060 between the impact bid of 50,000 and ask of 50,120, both formula contributions are zero. The premium index is then zero while the executable bid–ask gap remains.

Move the reference above both prices, to 50,200. The bid contribution stays zero, while the ask-side contribution becomes 80. The premium is −80 ÷ 50,200 × 100, about −0.1594%. The same book now represents a discount relative to a higher reference.

To open a conventional long-spot/short-perp pair, you buy spot from asks and sell the perpetual into bids. In our example, the 0.4 BTC perpetual sale averages 50,000, while enough spot is offered at 49,920. The gross entry spread is 80 USDT per BTC, or 32 USDT on 0.4 BTC.

Now include illustrative opening fees: 0.08% on the spot purchase and 0.05% on the perpetual sale. The spot leg costs 19,968 USDT and incurs 15.9744 USDT in fees. The futures leg has a 20,000 USDT notional and incurs 10 USDT. Only 6.0256 USDT of the entry spread remains after those opening fees. These are exercise inputs, not a current fee quotation from OKX.

That 6.03 USDT is not a completed arbitrage profit. The perpetual sale opens a derivative position; it does not hand you 20,000 USDT of spendable proceeds as a spot sale would. The pair still needs to be closed or otherwise settled, and its result depends on the exit spread, later funding, commissions and any financing.

If the exit basis is wider than the entry basis, the paired position can lose money even when Bitcoin itself is little changed. If it narrows, the convergence can contribute a gain. At the same time, funding can add to or subtract from the result. A screenshot showing only the entry premium cannot establish the final return.

A larger order changes which comparison is relevant

Suppose you want 2 BTC rather than 0.4 BTC. The impact sample above covers only the first 0.4 BTC. Additional bids and spot asks determine the rest of the execution. Applying the 80 USDT spread to all 2 BTC would invent a fill price for liquidity that has not been inspected.

Even at 0.4 BTC, two independent orders are not one guaranteed transaction. One leg might fill before the other, one limit might remain open or a quote might disappear. This is why a usable spread comparison includes quantities on both sides at the same observation time. It does not rely on two attractive prices captured several minutes apart.

Use the same quote currency and valuation time on both sides. BTCUSDT futures divided by a BTCUSD spot price mixes a contract premium with the USDT/USD conversion. A spot index combines constituents; one exchange’s best bid is a different reference.

For a cash-and-carry entry, buying spot consumes asks while shorting the perp consumes bids. That executable difference, after fees, can be smaller than the midpoint premium shown on a chart.

A premium observation is a measurement of the book against its reference. Funding is the output after the contract applies its averaging window and calculation rules. They may be displayed together, but a fresh premium reading does not erase the observations already included in the window.

Consider a simplified weighted average with three observations: +0.06%, +0.03% and −0.03%, weighted one, two and three. The weighted result is (+0.06 + 0.06 − 0.09) ÷ 6 = +0.005%. The latest observation is negative while the average remains positive. This three-sample exercise demonstrates the arithmetic; it is not an exchange’s full sampling schedule.

A separate interest component or dampening rule can also affect the result. That is why a slightly discounted perpetual can coexist with a positive funding charge. To explain a specific payment, follow the chosen contract’s full formula and observation window instead of substituting the last traded price.

The distinction between products is real. Deribit’s funding specifications use a mark-versus-index premium with a dampener. Hyperliquid documents an impact-price approach for its native perpetuals, with a different midpoint-based premium for HIP-3 markets. A recipe labeled simply “DEX premium formula” misses that difference even within one trading ecosystem.

For an ordinary trader, reconstructing hundreds of premium samples is rarely necessary. The immediate question is whether you are evaluating an order price, estimating the next carrying cost or explaining a past funding entry. The first needs executable quotes, the second needs the current funding estimate and the third needs the settled history. Use the measure that belongs to that decision.

A dated future has a remaining life until its specified settlement. A perpetual normally has no scheduled expiry. That difference matters when someone takes a price gap and labels it an annual return. There is no automatic “days until convergence” for the perpetual.

Imagine a dated future at $51,000, spot at $50,000 and 90 days remaining. The simple basis is 2%. Under a 365-day, non-compounding convention, annualizing that gap gives 2% × 365 ÷ 90, or about 8.11%. This is a rate-comparison convention before financing, execution and settlement costs, not a promised return on the margin deposited.

Put the same $1,000 premium on a perpetual and there is no contract expiry to supply the 90-day denominator. You can choose a planned holding period, but the basis may not have disappeared when that period ends. Dividing by an invented maturity converts an uncertain trading outcome into a deceptively precise yield.

Funding annualization is different again: it projects repeated payments at an assumed rate. A chart might place annualized funding and dated basis next to one another, yet one is extrapolated from a short observation and the other is tied to a future settlement date. The units can be made comparable while their uncertainty remains different.

CME’s cryptocurrency tools separate basis analysis from implied-rate curves. That separation is helpful here: first understand the observed price relationship, then identify the assumptions used to turn it into a rate. If a page reports “BTC basis 12%,” ask whether that means a price premium, an annualized dated spread or annualized perpetual funding.

• Comparing entry prices: use the relevant bid or ask and the average fill for the intended quantity.
• Explaining an upcoming funding estimate: use the contract’s premium averaging and formula, not a last-price gap.
• Reviewing a spot/perp pair: reconcile both executions, the exit basis, actual funding and all fees.
• Comparing dated and perpetual carry: keep maturity-based basis and projected funding explicitly separate.

In the worked example, “BTC trades 0.3206% above the index” is a description of the midpoint. It is not the 0.2004% impact premium and not the 6.03 USDT left from a 0.4 BTC entry spread after assumed opening fees. Keeping those labels attached to the numbers is the essential reading skill.