Weekend Forex Perpetuals: Design Decisions for Continuous FX Venues

Spot forex closes for roughly 48 hours a week. From Friday evening in New York to Sunday evening in Sydney, the distributed network of about a dozen tier-1 bank dealing desks that make the interbank market goes home. Prime brokers unwind risk, credit lines reset, and daily FX turnover — over $7.5 trillion on a weekday — is effectively zero on Saturday.

A venue running continuous forex perpetuals does not get to close. The order book stays open, the matching engine keeps matching, funding keeps accruing, and the mark-price feed has to produce a defensible number every second for 48 hours without an interbank reference. Whether that continuous window is a competitive moat or a slow-motion incident report depends entirely on decisions made at the platform layer months before the first weekend goes live.

This is a playbook for operators — the design choices a venue takes when it commits to hosting continuous FX perpetuals through the interbank blackout.

The Mark-Price Problem During the Blackout

Weekday FX mark price is easy. Blend a handful of independent spot sources, median-filter, throttle, publish. During the Friday-to-Sunday window there are no spot sources to blend. The venue's own order book is the only thing producing prints, and that book is thinner than the weekday book by design — market makers who quote tight during London hours are not all at their desks at midnight Saturday.

Three architectural options and their trade-offs:

• Book-only mark price. Simple, but every fill on a thin weekend book moves the mark. A single sweep of shallow depth can print a wick that cascades liquidations across every leveraged account on the pair. This is the failure mode that produces incident reports
• Synthetic fair-value anchor with book blending. The mark is anchored to the last valid interbank print, adjusted by a moving fair-value proxy (interest-rate differential drift, correlated-asset moves, index-level FX proxies), and blended with the venue book at a weighted ratio. The venue's own prints influence the mark but cannot dominate it
• Defensive-mode mark. The mark widens its own throttle bands during the blackout — outlier rejection tightens, update rate slows, and single-print moves are dampened. Liquidations still trigger on real market-wide moves, but transient book noise does not cascade

The pattern that ships on venues that survive multiple weekend cycles without incidents is a combination of all three: a synthetic anchor as the fair-value floor, book blending for genuine price discovery, and defensive-mode throttling as the safety layer. Getting the weights and the throttle bands right is the work.

Session-Boundary Funding for FX

Crypto perps settle funding every 8 hours because the underlying trades 24/7 and the funding curve should track continuously. Forex is different. The interest-rate differential between two currencies is a daily-carry number set by central-bank policy, not a minute-by-minute market print. An 8-hour cycle on an FX perp forces the venue to make up two funding numbers per weekend day when there is no fresh interbank basis to reference.

The cleaner design is a 24-hour funding cycle anchored at 22:00 UTC — the Sydney open, and the exact moment the interbank market comes back online after the weekend. That places the settlement boundary at the session pivot, so:

• Each Friday's final settlement happens with the interbank market still open and a clean reference for the day's funding rate
• Saturday and Sunday settlements happen with the venue on synthetic-anchor pricing, but only one settlement per day rather than three
• The Sunday-evening 22:00 UTC settlement lands at the exact moment interbank flow returns, so the first post-blackout settlement uses fresh reference data

The engine has to support per-symbol funding cadence — a mistake we have seen operators make is to hard-code an 8h cycle globally, then discover after launch that weekend FX settlements are computed against stale references. The BP engine's `SymbolFunding` config is per-instrument for exactly this reason: FX at 24h anchored 22:00 UTC, crypto at 8h anchored 00:00/08:00/16:00 UTC, equity perps aligned to their own session model.

The Sunday Reopen Is a HOUSE-Inventory Event

Whichever entity is running as the venue's market-maker of last resort — HOUSE, the operator's own book, or a designated internal maker — is carrying inventory into the 22:00 UTC Sunday reopen. That inventory gets marked against the first interbank prints, and if the interbank market opens 30-80 pips away from the venue's Sunday-close print (which is common after a real weekend news event), the HOUSE book takes an immediate mark-to-market move.

The operator has three levers:

• Position caps on HOUSE by symbol. Hard limits on how much inventory the internal maker is allowed to accumulate during the blackout. Wide enough to keep the book quotable, tight enough to survive a gap
• Skew the HOUSE quote schedule. Weekend spreads should be wider than weekday spreads not just because depth is thinner, but because the maker is pricing in the gap risk it will absorb at reopen. Weekend spreads on major pairs typically sit 3-6x weekday levels for exactly this reason
• Reopen hedging path. As soon as the interbank market prints post-22:00 UTC Sunday, the HOUSE book needs a fast path to hedge out accumulated weekend inventory into the returning liquidity. Hedging desks that are not staffed for the Sydney open leave that risk on the table

Related reading: funding-rate arbitrage across asset classes covers the flow patterns that concentrate in HOUSE inventory during off-hours, and liquidation prices on multi-asset margin covers the mark-price interactions that decide whether the Sunday reopen is quiet or loud.

Weekend-Gap Liquidation Risk on Aggressive Leverage

Even with a well-engineered mark price, a real weekend event — an unscheduled central-bank action, a geopolitical escalation, an EM-currency shock — will move the venue book. High-leverage positions taken into the weekend on the assumption that spreads stay tight and depth stays weekday-normal are the accounts that liquidate hardest, and the ones most likely to touch the insurance fund.

Design decisions that limit the operator's downside:

• Weekend leverage caps. Some venues drop max leverage on FX pairs by one tier from Friday 22:00 UTC to Sunday 22:00 UTC. Retail-messaging-friendly it is not; operator-book-friendly it is
• Maintenance-margin uplift during the blackout. A 0.3% weekday maintenance margin on EUR/USD moving to 0.6% during weekend hours doubles the buffer between mark and bankruptcy price without changing the leverage the trader sees at entry
• Pre-weekend margin-call cadence. The risk engine should push margin-ratio warnings more aggressively on Friday afternoon so participants have the option to trim before the interbank close, not after

Whether an operator ships all three is a commercial choice as much as a risk one. Operators on the BP platform get them as configurable defaults, tunable per tier.

Ops Shift Model — Does the Venue Actually Run 24/7?

A continuous order book is a technical claim. A continuous ops team is an organisational one. Venues that market weekend FX and then run reduced ops cover discover the gap the first time a mark-price feed misbehaves on a Saturday.

Minimum ops posture for a serious weekend-FX venue:

• On-call risk engineer with authority to switch mark-price mode (book-blended to synthetic-anchor-only) without waiting for a Monday standup
• On-call trading-ops with visibility into HOUSE inventory and the ability to widen spreads or pull quotes on specific symbols
• Escalation path to the CTO or head of risk within 15 minutes at any hour, including Sunday 03:00 UTC
• A pre-agreed decision tree for weekend halt scenarios — which pairs pause, which mark modes are triggered, which participants are notified — so decisions get made in seconds rather than being invented on the incident call

The alternative — 24/7 order book with 40/5 ops cover — is the design pattern that produces the incidents.

Why Weekend Forex Is a Moat (or a Liability)

Operators who ship continuous FX perpetuals correctly get three things a spot-hours-only venue cannot offer: cross-asset traders running FX alongside crypto on one account and one margin pool, hedgers with real-economy exposure who cannot tolerate a 48h risk window, and algorithmic desks whose systems run 168 hours a week regardless of whether the venue is open. That flow does not migrate back to weekday-only venues once it lands.

Operators who ship it badly get the inverse. A single cascade of liquidations at the Sunday reopen, a mark-price feed spike on a thin Saturday book, a HOUSE inventory event that eats a quarter's revenue — any of these produces a public incident, a trust cost that takes months to earn back, and a migration event in the wrong direction.

The difference is entirely at the platform layer. The traders do not choose whether the venue survives the weekend. The design does.

What Basis Points Ships

Every default in the BP platform reflects the team's ~30 years of combined experience shipping matching engines, hedging stacks and venue infrastructure — including running weekend FX on prior venues and watching how the failure modes actually manifest.

Specifically: the mark-price engine ships with defensive-mode weekend behaviour built in — synthetic fair-value anchor, book blending at configurable weights, tightened outlier rejection during the blackout window. Funding cadence is per-symbol in `SymbolFunding` config, so FX runs 24h anchored 22:00 UTC while crypto stays on 8h. Weekend margin uplift and leverage caps are shipped as configurable defaults. HOUSE inventory and mark-freshness surfaces are exposed on the ops console so the on-call engineer can see the shape of the book at any hour. The ops-shift model, mark-mode switching authority, and pre-agreed weekend decision tree are documented in the platform runbook that ships with the licence.

Operators inheriting the platform get all of it as the starting point. Continuous FX perpetuals are one of the harder pieces of infrastructure a multi-asset venue can ship. Getting the design right at the platform layer is what turns them into the moat.

Frequently Asked Questions

Evaluating the platform?

Related articles