A delta-neutral funding-arbitrage trade looks like it has no directional risk because the long and short legs cancel out. They do — for the asset price. They don't for everything else. Three categories of risk persist, and one of them is the most common cause of account blow-ups in this strategy.
Risk 1: Per-Leg Liquidation
Each leg has its own margin requirement at its own exchange. If BTC moves 10% against the leg you're long on (i.e., BTC dumps), that leg loses money fast even though your short leg on the other exchange profits the equivalent. The two exchanges don't know about each other.
If you ran your long leg at 5× leverage with tight margin, a 10–15% adverse move can liquidate it. The short leg profit sits on the other venue, untouched and unreachable. You're effectively closed out at the worst possible moment.
Risk 2: Exchange Outage
CEXs and DEXs both occasionally have downtime or withdrawal freezes. If one of your two legs becomes inaccessible during a market move, you're effectively running a directional position on the other side. Hyperliquid had a brief halt during the 2024 launch; Binance has had multiple 1–2h outages over the years.
Mitigation: don't put more than 30–40% of your capital into a single pair. If you have $50K, run several pairs of $5–10K each rather than one $50K megaposition.
Risk 3: Basis Risk on Synthetic Markets
This is the subtle one. If both venues use the same underlying spot index (e.g., both Binance and Bybit BTC perps reference the same major spot exchanges), they'll converge tightly. If one uses an oracle and the other uses RFQ, or if it's a synthetic asset (gold, equities) with no shared spot reference, the "two prices" of the same nominal asset can drift apart for days.
Most extreme example: pre-IPO equity synthetics. ANTHROPIC perp on one venue and ANTHROPIC perp on another might both quote "Anthropic" but use different oracle methodologies — their mark prices can sit 20%+ apart with no convergence force. Funding arb on those is not really arbitrage; it's a bet on a specific oracle going somewhere.
Sizing Framework
Three rules that have kept this strategy profitable for retail-size accounts ($5K–$500K):
- Per-pair size ≤ 25% of total capital. Diversify across 3–5 active pairs.
- Per-leg leverage ≤ 3×. Buffer above maintenance margin ≥ 25%.
- Skip any pair where one leg has under $1M open interest or under $500K daily volume. These thin markets are where slippage and basis blowups happen.
When to Close a Position Even at a Loss
Funding spread compresses below break-even? Close. Spread flips sign (long leg now pays you)? Close and re-open in the opposite direction. One leg is on an exchange undergoing rumored insolvency? Close immediately, even at adverse spread — the discount you take is cheaper than chasing the receivable through bankruptcy court.
The unsexy answer to "how do you not get liquidated running delta-neutral?" is: by sizing conservatively, diversifying venues, and being willing to exit small losses before they become big ones. The math edge is real; the discipline to capture it is what separates profitable arbitrageurs from blown accounts.