Open interest and funding rates, and the limits of reading them

A perpetual contract has no expiry, so it has no final settlement to pull its price back toward the underlying asset. The funding rate is the mechanism that replaces that pull. It is a periodic payment exchanged directly between holders of long and short positions, and it is set so that when the contract trades above the underlying, the long side pays the short side, and when it trades below, the short side pays the long side. The payment creates an incentive to take the other side of the trade, which pushes the contract price back toward the reference price.

The rate is calculated from a premium index rather than set by hand. The published methodology at the major venues is consistent in shape: the premium index compares an impact bid and an impact ask price — the average prices at which a fixed notional size could be executed — against an index price drawn from spot venues, and expresses the difference as a proportion of the index. That premium is then averaged over the funding interval, combined with a small fixed interest-rate component, and clamped between an upper and a lower bound. The result is charged at fixed intervals, most commonly every eight hours, though venues shorten the interval when the rate presses against its cap.

Two features of that construction matter for interpretation. The first is that the payment is peer-to-peer: the exchange facilitates the transfer but does not collect the fee, so funding is not a revenue line and its sign is not a statement about the venue's view. The second is that the rate is bounded. A clamp means the published figure cannot exceed a stated maximum, so an extreme reading is evidence that the mechanism has saturated rather than evidence of how extreme the underlying imbalance actually is.

The common reading is that rising open interest means new money is entering and falling open interest means positions are being closed, and that persistently positive funding means the market is crowded long. Both readings are directionally reasonable and both are routinely overstated. Open interest cannot distinguish between a new leveraged long and a new hedged short, because both sides of the contract are counted. Funding cannot distinguish between genuine directional conviction and a basis trade in which the funding payment is simply the cost of carrying a hedged position. In both cases the figure is consistent with more than one story, and the data does not choose between them.

There is a second limitation that is structural rather than interpretive. Both measures are reported by the venues that operate the contracts, and the definitions are not harmonised across them. Basket sizes, funding intervals, index composition, impact notional sizes and clamp bounds all differ. A funding rate of a given magnitude on one venue is not the same quantity as the same number on another, and a comparison across venues that ignores the methodology is comparing labels rather than measurements. The same caution applies to open interest, where the contract multiplier decides how much exposure a single contract represents.

The honest use of these measures is as context rather than as signal. Open interest tells a reader how much exposure is outstanding and whether the book is growing or shrinking. Funding tells a reader which side is currently paying to hold its position, and therefore which side the mechanism is leaning against. Neither tells a reader what will happen next, and neither is a substitute for the spot record, which is the only series that shows what was actually bought and sold. The spot and futures page sets out why that distinction matters, and the liquidity page explains why a large book is not the same thing as a deep market.

One further caution is worth stating plainly. Leverage amplifies moves in both directions, and a market with high open interest and stretched funding is one where a price move can force liquidations, which produce further price moves. That dynamic is real and it is visible in the record. It is also not forecastable from the two figures alone, because the same configuration can resolve quietly or violently depending on what happens next. A reader who treats a funding extreme as a timing signal is reading a mechanism as though it were a prediction.