Cross-Margin & the Scenario-Margin Model

This doc explains the durable principles behind how Proof decides whether your account is solvent: how one shared pool of collateral backs every position you hold, what initial and maintenance margin mean, how leverage works, and — the part that is specific to Proof — how margin is charged across the possible outcomes of every conditional market you trade. The headline result is that fully hedged, offsetting positions can net to roughly zero margin.

This is a conceptual reference. It describes the principles you can rely on, in plain terms, not wire formats, APIs, or exact parameters.

One collateral pool

Proof is cross-margin on a single collateral pool. You have one account with one collateral balance, denominated in a single stablecoin, and every position you hold shares that one pool. There is no per-position isolated-margin mode and no separate buckets — a gain on one position is available to support another, within the limits described below.

A few quantities follow from this:

• Balance — your settled cash. Deposits, withdrawals, realized profit and loss, fees, and funding payments all move this number.
• Unrealized PnL — the mark-to-market gain or loss on your open positions, valued at the relevant mark price.
• Equity — your balance plus the unrealized PnL of your open positions. This is what actually backs your risk.
• Free margin — the equity left over after covering the margin your open positions and resting orders already require. This is what you can use to open more risk or to withdraw.

Because everything shares one pool, equity is the thing the engine cares about. If your equity stays above your requirement you are solvent; if it drops below, you are at risk of liquidation.

Initial vs maintenance margin

Two requirements govern a position.

• Initial margin (IM) is the collateral required to open or increase a position (and to rest an order — see below). It is the larger, up-front requirement: you cannot put on risk unless your free margin covers it.
• Maintenance margin (MM) is the smaller, ongoing requirement to keep a position open. If your equity falls below your maintenance requirement, the position becomes eligible for liquidation.

Both scale with position size. Each market sets an initial-margin level and a maintenance-margin level, and the maintenance level is always at or below the initial level. The gap between them is your buffer: you open with IM, and you have room to lose down to MM before liquidation.

You can choose to be more conservative than a market's default by setting a tighter personal initial-margin level on a market — this only ever raises your initial requirement (lowers your own leverage). It never loosens anything, and it never changes the maintenance requirement, which always follows the market's setting.

Leverage and the conservative default

Leverage is the inverse of the initial-margin level: a lower margin level means higher leverage and a larger equity swing per unit of price move.

Two points specific to Proof:

• Conditional perpetuals carry leverage too. A conditional perpetual — a perp on the underlying that pays out only if its event resolves its way and otherwise voids — has its own initial- and maintenance-margin levels and is margined like a perp in the branch where it fires. They are margined positions, not full-collateral instruments.
• The listing posture is conservative. Proof lists conditional markets conservatively, and the conservative listing is the reference for the leverage you should expect on them.

Prediction binaries — the $0–$1 instruments whose price is the market-implied probability of an outcome — carry no separate position margin. Their maximum possible loss is already fully reflected in your equity under each outcome (a binary that loses simply settles to $0, a real and already-counted cash loss). They are accounted for through equity, not through a margin add-on.

The worst-case idea: scenario margin

A perp-only account is simple: equity must stay above the sum of each position's margin requirement, valued at one mark. Conditional markets make this richer, because a conditional position's value depends on how its event resolves. A gain on a "YES" conditional perpetual is real if the event resolves YES — and disappears entirely if it resolves NO. You should not be able to borrow against a gain that can vanish.

Proof handles this with scenario margining: portfolio, worst-case margining. Rather than valuing your account at a single point, the engine considers the possible outcomes across the conditional markets your account touches and requires that your account be solvent — equity at or above the requirement — in every one of those outcomes. The binding constraint is the worst one.

Some properties that follow:

• No credit travels between outcomes. A gain that exists in one outcome cannot offset a requirement in a different outcome. You can never lean on a profit that some outcome erases. The published equity figure for an account is the equity in its worst outcome.
• Perp-only accounts collapse to the simple form. With no conditional markets there is one trivial outcome, and the check reduces to the plain "equity ≥ sum of margin" with no overhead.
• A non-firing conditional perpetual contributes value zero in an outcome — it voids, its PnL is zero, and its margin is returned. It is not a loss to $0. A non-firing prediction binary, by contrast, contributes its real loss (it settles to $0) directly to equity in that outcome — which is why binaries need no separate margin charge.
• The check is deliberately conservative. It evaluates each outcome on its own terms and holds you to the worst one, which keeps the requirement on the safe side of your true risk.

The check that gates opening an order, the check that gates a withdrawal, and the check that triggers liquidation all use this same worst-outcome predicate, so they agree on what "solvent" means.

The relationships between fair conditional prices — for example, that the underlying forward equals the probability-weighted blend of the two conditional prices — are no-arbitrage relationships that hold between prices. The margin check does not depend on the books being coherent; it simply requires you to survive each outcome.

How hedged positions net to near-zero margin

The payoff comes from how scenario margining treats offsetting positions that share the same underlying.

Within each outcome, positions on the same underlying offset before margin is charged. Legs that cancel each other's exposure in an outcome — a long and a short of equal size — net to zero exposure there, and zero exposure means roughly zero margin. You are charged on your net directional risk in each outcome, not on each leg gross.

The clean case is a fully matched box that pairs conditional perpetuals with the underlying perpetual so that, whichever way the event resolves, the surviving legs cancel. Because such a box nets to zero in every outcome, it charges close to zero initial and maintenance margin, while the same legs held un-hedged would each charge full margin. A bare directional leg pays full margin; a properly offset position pays almost none. That capital efficiency is the core reason to build conditional markets out of perp-style legs in the first place.

One point of operating practice belongs with this benefit. A conditional-perp-plus-perp hedge cancels in the outcome where the conditional fires; in the outcome where it voids, the conditional disappears and the perp leg stands on its own. Scenario margining prices that outcome in too, so you are margined honestly throughout. Standard practice is to keep your conditional inventory flat or fully boxed before an event resolves, so your book is squared in whichever post-resolution world you land in.

Resting orders also reserve margin

Margin is not only about positions you already hold. Every resting order reserves the margin it would require if it filled, and keeps reserving it until the order fills or is cancelled. The reservation is folded into the same scenario check, in the outcomes where the order would carry risk.

Two consequences worth internalizing:

• Your requirement is your positions' maintenance margin plus the reserved margin of your resting orders. Stale or oversized resting orders consume free margin even when you have no position.
• Cancelling resting orders lowers your requirement. Because reserved order margin is part of the requirement, freeing it reduces what you must post. The liquidation process reflects this: before closing any positions, it cancels the account's resting orders and re-checks — and if the account is solvent without those reservations, nothing is closed.

The practical rule: cancel or flatten before you walk away. An order left resting keeps reserving margin until it fills or is cancelled.

An illustration

A short illustration ties the principles together; it shows the mechanism, not any live values.

Suppose your account holds collateral plus a single long conditional perpetual that is showing a gain at the current underlying price. You want to add a sizable long on the underlying perpetual itself.

Scenario margining asks how each outcome looks. In the outcome where the conditional fires, the conditional's gain is real and your equity is high. In the outcome where it voids, the conditional disappears — PnL zero, margin returned — and your equity is back to roughly your starting collateral. Because a perp carries its full requirement in both outcomes, the new perp is sized against the worst outcome — the one where the conditional contributes nothing. The conditional gain was real-looking but outcome-contingent, and the worst-outcome check declines to let you borrow against a value that some outcome erases. Size the perp so you clear margin in that worst outcome and the order goes through.

The same logic runs in reverse to your advantage. Add the offsetting legs so that, in every outcome, your net exposure on that underlying collapses toward zero. Your requirement collapses with it, and the capital you must post falls to near nothing. You are charged for the risk you are actually carrying in your worst outcome, and no more.

Liquidation and the backstop

When an account's equity falls below its maintenance requirement, the position becomes eligible for liquidation. Liquidation first cancels the account's resting orders and re-checks; if freeing those reservations restores solvency, nothing is closed. Otherwise the engine closes positions to bring the account back within its requirement, valuing closes at a reference price drawn from the market rather than at an arbitrary print.

Behind the per-account checks sits a backstop reserve that absorbs any shortfall left if an account cannot be unwound for its full requirement, keeping the rest of the exchange whole. The scenario check is the first line: by requiring solvency in every outcome, it keeps accounts on the safe side of their true risk before liquidation is ever needed.

Summary

• One account, one collateral pool; every position shares it. Equity = balance + unrealized PnL; free margin is what's left after requirements.
• Initial margin opens a position; maintenance margin keeps it open; breaching maintenance triggers liquidation, which closes positions at a reference price. Leverage is the inverse of the IM level, and Proof lists conditional markets conservatively.
• Conditional perpetuals carry real leverage; prediction binaries carry no separate margin (their loss is already in equity).
• The binding gate is scenario margining: solvency is required in every outcome across your conditional markets, with no credit travelling between outcomes.
• Offsetting positions on the same underlying net within each outcome, so a fully matched box charges roughly zero margin while the un-hedged legs would each pay full margin.
• Resting orders reserve margin too; cancelling them lowers your requirement. A backstop reserve stands behind the per-account checks.