On-chain derivatives: perpetual futures, funding and margin

A derivative is a contract whose value depends on a price it does not hold. On chain, the clearing house that would normally stand between two traders is replaced by collateral rules written in code, an oracle that supplies the reference price, and a procedure for closing positions that can no longer support themselves. The dominant instrument is the perpetual future, which has no expiry date and therefore needs a mechanism to stop it drifting away from the asset it tracks.

Funding: the tether that replaces expiry

A dated future converges on spot because it settles on a known day. A perpetual never settles, so it uses a periodic payment between the two sides instead. When the contract trades above the index price, holders of long positions pay holders of short positions; when it trades below, the payment reverses. The funding rate is typically a function of the premium, computed and exchanged at fixed intervals.

The arithmetic is simple and the consequences are not. A payment equals the position's notional multiplied by the rate for the interval, and it is charged on notional rather than on margin. A position of 100,000 notional at an illustrative rate of 0.01 percent per eight-hour interval pays 10 units three times a day, which is around 11 percent of notional per year if the rate persisted. Against margin of 10,000, that same flow is over 100 percent a year. Funding is the price of holding a directional position when most of the market wants the same side, and at high leverage it can dominate the outcome regardless of what the underlying price does.

Funding is not a yield paid by anyone's protocol. It is a transfer between traders, and receiving it means holding the unpopular side of a position with its own directional exposure. Related concepts from dated markets still apply: basis is the gap between a future and spot, contango and backwardation describe which way it points, and both are visible on chain in a way they rarely are in traditional venues.

Two market structures, and who is the counterparty

The first structure is a familiar order book, with matching handled by a fast off-chain or application-specific sequencer and settlement enforced on chain. Traders face other traders. The compromise is that a sequencer is a privileged party that sees orders before they are final, which puts front-running and the general problem of MEV at the center of the design rather than at the edge.

The second structure pools capital and lets it act as the counterparty to every trader, either through a virtual market maker curve or through a pool that quotes off an oracle price. Depositors receive the fees and, necessarily, the aggregate profit or loss of the traders. This is a coherent business and it is not a fee-earning position: a pool of this kind is short whatever the trader base is long, and its worst results arrive precisely when a large directional move is underway. Where the pool quotes at an oracle price with a small fee and no price impact, it is also the natural target for anyone who can act on a price update before the oracle reflects it.

Margin, liquidation and the socialization of losses

Positions are marked continuously against a mark price, usually an index of several venues rather than the venue's own last trade, specifically so that a thin book cannot be pushed into triggering liquidations. When margin falls below the maintenance requirement, the position is closed by a liquidator or by the protocol itself, and a penalty is charged.

What happens if closing the position does not recover the debt is the part that distinguishes designs. An insurance fund, built from liquidation penalties and fees, absorbs the shortfall first. If it is exhausted, the remaining loss has to fall somewhere: on the pool acting as counterparty, on profitable traders through automatic deleveraging that force-closes winning positions at the bankruptcy price, or on the protocol's own reserves. Automatic deleveraging is the honest disclosure that a fully collateralized system with no external capital cannot always pay winners in full.

Two structural details are worth adding. Collateral is usually posted in a single asset, so a trader holding a volatile asset as margin for a position in that same asset has correlated exposure on both sides of the account and the margin shrinks precisely when the position needs it. And because settlement is on chain, positions cannot be netted against an outside prime broker, which means capital sits fragmented across venues and the total collateral supporting the market is smaller than the sum of the positions suggests.

What to read, and its limits

Open positions and leverage across a market are described by open interest, positioning skew and the funding rate, and cascades are visible as clusters of forced closes at particular price levels. What deserves care is treating derivative volume as evidence of adoption: derivatives turnover can exceed spot turnover by a wide margin without any change in the underlying holder base, so turnover should be read against the market capitalization it is turning over, and on-chain venue volume against the capital supporting it. A large part of reported volume across the industry has historically been wash trading on venues with incentives tied to volume, which is a measurement problem before it is anything else.

The exchange activity pages carry venue-level volumes, the risk section covers liquidation and insurance fund design, and the metric catalog defines each volume series and its source. Because these systems price from oracles and settle in collateral borrowed elsewhere, the next two articles, on oracles and on bridges, describe most of the ways they break.