Funding arbitrage risks

Funding rate arbitrage is not risk-free. The two legs cancel most price moves, but they sit on separate venues with separate margin. The largest losses come from one leg being closed without the other (liquidation, auto-deleveraging, a venue failure); the more frequent ones from funding turning against the pair and from execution costs that eat a thin spread.

Every figure below is a hypothetical illustration, not a forecast. Funding rates change all the time, and nothing here is investment advice. If the trade itself is new to you, start with how funding rate arbitrage works.

Is funding rate arbitrage risk-free?

No: the hedge holds only while both legs exist. Here is what each risk can cost a pair with $10,000 per leg.

What happens if one leg is liquidated?

Margin is not shared between venues. When the price rises, the long's gain on venue A cannot top up the losing short on venue B until you move money yourself, which takes minutes to hours.

A rough rule: a leg is liquidated after a move against it of about 1 / leverage minus the maintenance margin, which varies by venue and coin. Assuming 3%:

Say the coin rises 30% and the short at 3x is liquidated. The long's $3,000 gain roughly offsets the short's $3,000 loss, though the venue may keep most of the short's remaining $333 of margin as a liquidation fee. Then the long is a naked $13,000 position, and a 10% pullback before you close it costs $1,300. Small coins can move 30% in a day.

What helps: lower leverage, spare collateral on both venues, price and margin alerts on each leg, and a plan for the losing side: add collateral there or cut both legs equally, well before the liquidation price.

What if funding flips?

A spread measures the past. Spikes tend to fade, and a pair can stay negative for days. A −30% APR spread for 3 days on $10,000 per leg costs about $24.66 (10,000 × 30% × 3 / 365).

Suppose a pair paid +15% APR for 11 days and then −30% for 3 days. Funding earns $45.21 and gives back $24.66, and a round trip of taker fees (4.5 and 4.0 bps, four fills) costs $17. Two weeks leave about $3.55.

What helps: pick pairs on the 7-day spread and its stability, not on today's rate (see how to pick a funding pair), and set an exit rule before you open: for example, close after a full day of negative spread, or when a better pair would pay back the fees of switching within a few days.

How much can legging and slippage cost?

The two orders never fill at the same instant. On $10,000 per leg, a 0.5% move between the fills costs $50, and that gap is locked in: if your long filled 0.5% above your short and both venues trade at the same price when you close, the 0.5% is lost; you get it back only if the long's venue trades at a similar premium at exit. A week of a 20% APR spread pays about $38.36, so one bad entry can erase a week of funding.

Thin books add slippage, and some venues add latency on purpose: Lighter's Standard accounts carry a taker delay of about 300 ms. The live price spreads show how far apart venues trade right now.

What helps: open the leg on the thinner or slower venue first and hedge on the deeper one, split large orders, and never assume the price on screen is your fill.

What is ADL, and why does it hit hedged traders?

Auto-deleveraging (ADL) is a venue's last line of defence. When a liquidated account's losses exceed what the venue's backstop can absorb, it closes positions on the opposite side, ranked by profit and leverage, at a recent mark price. Hyperliquid's docs describe this ranking, and CEXs run ADL too.

In a crash, the hedge is often the profitable leg: your short is deep in profit, ADL closes it, and your long on another venue keeps falling. On 10 October 2025, Hyperliquid's ADL closed about $2.1 billion of positions in 12 minutes, according to a study of that day.

What helps: lower leverage moves you down the queue on venues that rank by it; many CEXs show an ADL indicator; and your alerts should check that both legs still exist, not only their margin.

What if a venue goes down or fails?

Outages, exploits and closures happen. Drift lost about $285M to an exploit in April 2026. Ethereal delisted all its markets on 1 September 2026. Withdrawals get paused during incidents. If the venue holding one leg stops responding, you cannot close that leg, and closing the other one leaves the stuck leg unhedged for as long as the venue is down.

What helps: keep no more on a venue than you can afford to lose, withdraw idle collateral, and spread pairs across several venues.

Can one ticker be two different assets?

Yes. In one real case a coin traded at $11.65 on one venue and $0.106 on another under the same name. In one of our own pairs, equal coin counts on both venues meant a $1,541 long "hedged" by a $14 short. Contract multipliers do the same thing quietly: 1000PEPE and kPEPE are contracts on 1,000 coins, and a stock perp can share a ticker with a token.

What helps: compare mark prices and dollar notional, not coin counts, and read both contract specs. Our screener warns when a pair's mark prices differ by more than 2% and hides pairs that differ by more than 10%.

Do settlement times and stale data matter?

You pay or receive funding only if your position is open at the funding time, so an 8-hour leg pays in larger, rarer lumps than an hourly one, and a short hold can catch or miss a payment by timing alone. On CEXs the current rate only predicts the next payment and can change before settlement. A venue can also go quiet: we never pair a market without fresh data, but a frozen number elsewhere can look like an opportunity.

Which costs do people forget?

• Four taker fills, not two: open and close on both legs ($17 in the table above).
• The quote spread on quote-based venues: Variational charges no trading fee, but you pay its quote spread.
• Deposits, withdrawals, bridges and gas, plus idle days while money is in transit.
• USDT vs USDC: the CEX perps we track are USDT-margined and many DEXs settle in USDC, so moving collateral between them costs a conversion spread, and a depeg of either stablecoin moves one leg's collateral.
• Taxes, which depend on where you live.

Our net APR covers only taker fees; the rest is listed under "What we do not model" in the methodology. In our own funding bot, a stretch of frequent re-entries once cost about twice as much in fees as the pairs earned in funding.

Why count profit from equity, not from a model?

Venues book fees and funding differently. One folds the opening fee into position PnL, another shows funding as a separate line, a third rounds or delays settlements. A spreadsheet that adds up "funding received" drifts from reality: on one trade, our own bot's report showed 1.8 times the profit that both accounts' equity confirmed, because the opening fee fell outside the fills it read. Record the total equity of both accounts before you open and after you close, adjust for deposits and withdrawals, and treat the difference as the result.

What about rules and access?

CEXs require KYC; most DEXs do not, though GRVT is widely reported to verify identity. Many venues bar certain countries, ban multiple accounts and trading with yourself, and can change or cut points programs retroactively. A frozen account is also a frozen leg. Check each venue's terms on the exchanges page and use only venues you are allowed to use; we do not help anyone get around restrictions.

A risk checklist before you open a pair

1. Leverage low enough that the move to liquidation (≈ 1 / leverage − maintenance margin) is far beyond the coin's daily range.
2. A collateral buffer on both venues, not just the riskier-looking one.
3. Alerts on price, margin and leg size on each venue.
4. The same asset on both legs: mark prices, multipliers, contract specs.
5. An exit rule written down before entry: when the spread turns negative, or when a better pair would pay back the cost of switching.
6. A size the thinner book can absorb on entry and on exit.
7. A cap on capital per venue.
8. Profit counted from both accounts' equity, not from a model.