New here? Plain-language guide: what each panel means, money needed, how many clients → Can't take weekend risk? Step-by-step zero-risk plan → Hedge everything on CME →
24/7 Hedge Router — backtest and risk desk
—
Strategy comparison
Three ways to run the same book
Averages over — simulated years of — weeks, — symbols, broker capital —. Rejected = client volume turned away by exposure caps.
| Strategy | Mean P&L | 5th pct | 95th pct | Max drawdown | Worst weekend | Weekend P&L | Spread revenue | Stale-quote loss | Hedge cost | Option premium | Option payoff | Rejected volume | Sharpe |
|---|
Showcase year: cumulative P&L by strategy
One simulated path at 4-hour resolution. The dip in week — is a forced "news weekend": crypto −15% to −25%, gold +4.5%, Brent +12%, all while weekday venues are closed.
Where the money comes from and goes
Annual P&L components, mean across paths. Stale-quote loss is clients hitting a frozen weekend price that the real market has already left.
Monte Carlo band: cumulative P&L, 5th–95th percentile
Shaded band is the spread of outcomes across paths; line is the median. A narrow band is what the hedging buys.
Weekend P&L distribution
Every weekend of every path (Friday 21:00 to Sunday 22:00 UTC, including Friday sweep and Sunday roll). Bins of $50k.
Risk desk — Friday close, stress weekend
What the dealer sees before the weekend
Client net position before and after the Friday sweep, the hedge in place, the weekend cap and option cover, assumed gap (3.5 weekend σ), VaR before and after insurance, then what the weekend actually did. Positions in USD notional; positive = clients net long.
| Symbol | Weekend mode | Client pos (pre-sweep) | Hedge after sweep | Unhedged | Weekend cap | Option cover | Gap assumed | VaR gross | VaR after options | Unhedged at reopen | Realised gap | Weekend P&L |
|---|
Total weekend P&L for the selected strategy: —. "Unhedged at reopen" shows how much new weekend flow the weekend legs could not absorb (capacity limits), which is what the options cover.
Stress weekend, hour by hour
Quote vs true price and unhedged exposure
Three weeks around the stress weekend. Top: the price clients can trade at (solid) against the true market (dashed). Bottom: unhedged exposure. Naive B-book freezes commodity quotes over the weekend; the unified router quotes the CME 1-oz gold and 10-bbl WTI proxies, keeps copper close-only, and caps exposure.
A. Naive B-book
C. Unified router
By symbol
Annual P&L per symbol, by strategy
Mean across paths. The unified router gives up spread revenue on crypto (hedge cost, caps) and gains it back on commodities (no stale-quote losses).
Weekend P&L per symbol per year
Weekends only. Where the naive book loses on frozen commodity quotes, the router earns the widened weekend spread.
Router configuration
Hedge legs, caps and insurance per symbol (strategy C)
The whole system is this table: one row per symbol, one code path. Cap = max net client exposure the router accepts; mode = how the symbol is quoted when the weekday venue is closed; premium = cost of a ±1σ weekend strangle on the weekend cap.
| Symbol | Class | Ann. vol | Weekend σ | Markup bps | Direct leg | Cost bps | Weekend leg | Cost bps | Capacity | Weekend mode | Spread × | Cap weekday | Cap weekend | Options | Premium % | Premium / wk |
|---|
How the simulation works
- Prices: hourly lognormal paths per symbol at the stated annual volatility; weekend hours at a reduced volatility plus a random jump (news) with the stated probability. Commodity venues are closed Fri 22:00 – Sun 22:00 UTC, so the weekend move arrives as a Sunday gap.
- Client flow: persistent random net flow in USD, lower activity on weekends, positions decay as clients close. On a frozen (stale) quote, part of the flow is informed: it trades toward the true price. Wider spreads reduce flow (square-root elasticity).
- Broker P&L = spread captured − stale-quote fills − market move on net client exposure + hedge P&L − hedge costs − option premium + option payoff.
- Hedge router (C): hedge to flat when |unhedged| exceeds 30% of the cap, on the first open leg; Friday 21:00 sweep to flat and strangle purchase on gold, silver, Brent; weekend legs capped by their capacity; Sunday 22:00 roll back to the direct leg.
- Simulated data, not exchange history: swap the price and flow inputs for your own CSVs to re-run.
All figures are from a simulated backtest (300 Monte Carlo paths of 52 weeks), not live trading data.