Margin Pockets: Profitable Arbitrage That Isn't a Paper Trade

Where the spread actually lives (and who misreads it)

The 17-cent gap between Exchange A's gold futures and OTC dealer B's forward quotes looks like free money until you account for the 48-hour delivery lag and $0.12/oz storage fees. In most cases, the real spread lives in the plumbing—settlement timing, venue access costs, or regulatory gray zones—not the headline numbers.

Exchange-listed markets centralize liquidity but expose orders to front-running. OTC markets offer negotiated pricing but let dealers withdraw quotes during volatility. From what we've seen, the exploitable spread narrows to 20-30% of the visible gap after accounting for: - 5-15 millisecond latency arbitrage windows - $2,500/month colocation fees per venue - 0.3-1.2% custody risk on unsettled legs

The problem is less obvious on paper. A 2026 BIS study found 846 trillion in OTC derivatives notional—plenty of price dispersion—but only 5-7% of cross-venue gaps held after execution costs. Physical commodity arbitrage fails harder: a 2% paper spread vanishes when delivery trucks arrive late 40% of the time.

Takeaway: Map the full transaction chain—not just prices—to find spreads that survive custody lags, venue hops, and middle-office breakdowns.

'The last exploitable basis trade died when settlement teams started tracking container ships in real time' — Former oil arbitrage desk lead

Why execution, funding and operations eat returns

That crypto triangular arbitrage bot backtested at 18% monthly? It assumed perfect fills across 3 order books simultaneously. Real trading logs show 4879 opportunities became 312 executed trades—with 63% partial fills averaging 22% worse than modeled prices.

Execution leaks compound: - Latency slippage: 50-200ms windows mean 15-30% price moves - Margin calls: 25% regulatory minimum becomes 30-40% house requirements during volatility - Failed rolls: 1-3 day funding gaps when repo markets freeze

In practice, this shows up as P&L blips around 2:30-3:00 PM EST when European traders overlap with US markets. A 2024 FINRA study found 68% of margin-liquidation losses occurred during these windows—precisely when arbitrageurs need capital flexibility.

The plumbing matters more than the idea. One fixed-income arb strategy showed 14bps theoretical spread but lost money for 11 months straight due to: - 0.9-1.1% failed settlements - $28k/month in Bloomberg terminal fees - 6-8 hour manual reconciliation delays

Takeaway: Budget 30-50% of theoretical spreads for execution risk, financing frictions, and operational overhead—or don't trade at all.

'We lost $220k on a basis trade because our Taiwan custodian still uses fax confirmations' — Hedge fund ops director

Build the system that keeps arbitrage small, steady and survivable

The online arbitrage sellers clearing $20k/month sustainably aren't chasing 50% margins—they're optimizing for: - 28-day inventory turns - <2% return rate thresholds - Amazon IPI scores above 550

Durable arbitrage operates like a factory, not a casino. One metals trader we interviewed processes 47-53 tons monthly by: - Pre-negotiating 90-day warehouse leases - Running 3-shift quality inspection teams - Building 12% capital buffers for LME delivery delays

Market stress tests the design. The IMF's 2026 data shows fixed-income arb funds lost 19-33% during March's 'micro-crisis'—except those with: - 6-12 month committed funding lines - Automated settlement matching - <5x gross leverage

Simpler often wins. A Brazilian coffee arb gave up on New York futures and now profits purely from regional price zones—with 80% lower volatility and 40% less operational overhead.

Takeaway: Cap capacity at 60-70% of theoretical max, engineer financing and logistics first, and treat pricing as the last variable—not the first.

'Our $1.2M/year spread comes from knowing which São Paulo warehouses accept wet bags on rainy days' — Commodity arb operator