CircuitGet started

Funding rate arbitrage scanner

Updated by the Circuit team

What a funding scanner does

A funding scanner reads the funding rate of every perp on every venue you trade, ranks the gaps, and flags the ones worth acting on. It answers one question each hour: where does holding a hedged position pay more than it costs?

The funding rate arbitrage guide explains the trades themselves. This page covers the scanner that finds them: how each venue computes funding, how to annualize it, what to check before trusting a rate, and how to run the scan as a bot on Circuit.

How each venue sets funding

Hyperliquid and Lighter both settle funding every hour, and both use an interest rate of 0.01% per 8 hours on most markets. They differ in how they sample the premium and where they cap it. As of October 2026, from each venue's own docs:

Hyperliquid Lighter Papertrade
Settlement Every hour Every hour No funding
Interest rate component 0.01% per 8 hours (0.00125% per hour) 0.01% per 8 hours for most markets n/a
Premium sampling Every 5 seconds, averaged over the hour Once a minute at a random time, averaged over the hour n/a
Cap 4% per hour 4% per 8 hours (0.5% per hour) for most markets n/a
Notional for payment Size × oracle price Size × index price n/a
Exchange fee on funding None, peer to peer None, peer to peer n/a

On Hyperliquid the formula is F = P + clamp(interest − P, −0.05%, 0.05%) on an 8-hour basis, and each hourly payment is one eighth of it (Hyperliquid funding docs). Lighter clamps the premium toward the interest rate within 0.05%, then clamps again at 4% and divides by 8 (Lighter funding docs). The practical result is the same on both: when the premium sits within 5 basis points of the interest rate, funding settles at the baseline of 0.00125% per hour, paid by longs to shorts.

Papertrade settles its synthetic perps at Hyperliquid's mid with no funding at all, so it has no place in a funding scan. Its economics are covered in the Papertrade guide.

Annualizing hourly rates

Venues quote funding per settlement. To compare it with a lending rate or a staking yield, annualize it: multiply the hourly rate by 24 hours and by 365 days.

annualized = hourly rate × 24 × 365 = hourly rate × 8,760

Hourly rate Math Annualized
0.00125% (baseline) 0.00125% × 24 × 365 10.95%
0.0025% 0.0025% × 24 × 365 21.9%
0.005% 0.005% × 24 × 365 43.8%
0.01% 0.01% × 24 × 365 87.6%
0.03% 0.03% × 24 × 365 262.8%
0.5% (Lighter cap) 0.5% × 24 × 365 4,380%

This is simple annualization with no compounding. Hyperliquid's docs quote the baseline as 11.6% APR, which is the compounded figure. Use simple annualization for ranking. It keeps the comparison honest when rates change every hour.

Two conversions trip people up. A rate quoted per 8 hours is eight times the hourly rate, so 0.01% per 8 hours is 0.00125% per hour. And the payment uses the oracle or index price, so a position's funding in dollars moves with price even when the rate holds still.

What the scanner checks

A high rate on its own is a weak signal. Rates spike for an hour and revert. A good scanner checks six things before it flags a market.

Check Why it matters Circuit data field
Rate now The next settlement's predicted rate fundingRate, settlesAt
Persistence A rate that held for days is more likely to hold for hours cumulativeFundingRate and fundingSettlementCount over a lookback
Open interest Thin open interest means one trader can move funding openInterestUsd
Liquidity You need to enter and exit both legs without slippage eating the carry volumeUsd, medianDailyVolumeUsd
Fees Entry and exit fees must be paid back by funding Venue fee schedule
Leverage limit Sets the maintenance margin on the short leg maximumLeverage

Persistence is the most useful filter. Divide the cumulative rate by the settlement count to get the mean hourly rate. If a market shows a cumulative rate of 0.84% over the last 7 days across 168 settlements, the mean is 0.84% ÷ 168 = 0.005% per hour, or 0.005% × 8,760 = 43.8% annualized. A market that printed 0.03% last hour but averages 0.001% over the week is noise.

Fees decide the minimum holding time. As of October 2026, Hyperliquid's base perps fees are 0.045% taker and 0.015% maker (fee schedule). Lighter's default Standard account pays no maker or taker fees (Lighter fees). Divide round-trip fees by the hourly carry to get the hours needed to break even.

A worked example

Say the scanner reads ETH funding at +0.0040% per hour on Hyperliquid and +0.0010% per hour on Lighter, and both have held near those levels all week. A bot goes short $5,000 of ETH on Hyperliquid, where longs pay more, and long $5,000 of ETH on Lighter. Price moves cancel. The bot earns the spread.

Item Math Result
Hyperliquid short receives $5,000 × 0.0040% $0.20 per hour
Lighter long pays $5,000 × 0.0010% $0.05 per hour
Net carry $0.20 − $0.05 $0.15 per hour
Per day $0.15 × 24 $3.60
Spread annualized 0.0030% × 24 × 365 26.28% on the hedged notional
Hyperliquid taker fees, open and close 2 × $5,000 × 0.045% $4.50
Lighter fees on a Standard account 2 × $5,000 × 0% $0.00
Hours to break even $4.50 ÷ $0.15 30 hours

The bot needs the spread to hold for 30 hours before it makes money at all. After 30 days at the same rates it collects 30 × $3.60 = $108, or $103.50 after fees. Each leg also needs margin. At 3x, each leg posts about $1,667, so $3,333 of collateral earns that $103.50, about 3.1% in a month.

The risk is in the assumption. If Hyperliquid funding drops to match Lighter, the carry goes to zero and the bot still owes the exit fees. If it flips, the bot pays on both legs.

Reading funding on Circuit

Circuit's data commands return funding for both venues in one shape, so a bot never parses two APIs. Rates are fractions per settlement, never percents, and positive means longs pay shorts.

  • markets lists every perp on a venue with its current fundingRate, open interest, volume and maximum leverage. With a lookback it adds cumulativeFundingRate and fundingSettlementCount, and it can sort by either rate.
  • funding_rates returns the predicted next rate or every settled rate over a window, keyed by venue and symbol.
  • open_interest and candles give the history behind the size and liquidity checks.

These are display reads. When the bot places an order, the venue's own live price decides the fill. The data commands reference lists every field.

Build the scanner as a bot

A scanner can report or trade. Start with the report. A prompt for the Terminal:

"Every hour, right after funding settles, read 7-day funding for every perp on Hyperliquid and Lighter. Keep markets with at least $5 million of open interest and $20 million of 24-hour volume. For each asset listed on both venues, compute the 7-day mean hourly rate on each venue and the spread between them, annualized. Log the top 10 spreads with both rates, open interest and the hours needed to cover Hyperliquid taker fees. Don't trade."

Once the log looks right, add the trade: "When the top spread has been above 20% annualized for 24 straight hours and covers fees within 48 hours, short the higher-funding venue and long the lower one, $2,000 per leg at 3x. Close both legs when the spread falls below 5% annualized or after 14 days."

Schedules can align to the hour, so the scan runs right after each settlement. The funding rate short strategy is a simpler one-venue version you can start from.

Test it on recorded funding

Circuit backtests replay recorded Hyperliquid or Lighter markets, and positions pay funding at each hour's opening price. A one-venue funding bot can be backtested over the last seven days by default, or a window you choose. A backtest replays one venue at a time, so a cross-venue spread bot is tested in paper trading, where both venues fill against the live book with their real fees and funding.

Where scanners go wrong

  • Ranking by one hour. The top rate on any given hour is usually a spike. Rank by the mean over days.
  • Ignoring the cap. Hyperliquid can charge up to 4% per hour, Lighter up to 0.5%. In a squeeze the paying side bleeds fast.
  • Mismatched notionals. Funding pays on oracle or index price times size. Hedge by size, and rebalance when price moves the dollar values apart.
  • Forgetting margin. Each leg can be liquidated on its own. A cross-venue hedge has no shared margin, so a 30% move against one leg needs collateral on that venue.
  • Chasing thin markets. A small perp with $500,000 of open interest can show a huge rate that disappears when you trade it.

The delta-neutral strategies guide works through liquidation prices on the short leg.

Keys and permits

Bots run on a wallet you create in Circuit or import by private key. Circuit keeps the signing key in a hardware enclave, and bots never see it. A bot requests typed actions through Circuit's SDK, and Circuit signs only under the permit you sign with your passkey when you start the bot.

You can export the private key from Settings at any time. Pause, resume or stop any bot whenever you want. Stopping leaves your positions and resting orders in your account, and you close them from Terminal.

What it costs

Circuit charges no fees today and adds no builder fee to Hyperliquid or Polymarket orders. You pay the venue's trading fees, network gas and any bridge or routing fees, and each prepared action's review shows the costs it reports before you confirm. Today's terms make no commitment about future fees.

FAQ

How do you annualize an hourly funding rate?

Multiply it by 24 and by 365, which is the same as multiplying by 8,760. An hourly rate of 0.01% annualizes to 0.01% × 24 × 365 = 87.6%. Hyperliquid's baseline of 0.00125% per hour annualizes to 10.95%.

How often is funding paid on Hyperliquid and Lighter?

Both settle funding every hour as of October 2026. Both compute the rate on an 8-hour basis and pay one eighth of it each hour.

What is the maximum funding rate on Hyperliquid?

Hyperliquid caps funding at 4% per hour, the same for every asset. Lighter caps most markets at 4% per 8 hours, which is 0.5% per hour.

Is funding rate arbitrage risk-free?

No. Funding can flip, fees have to be earned back before the trade pays, and each leg can be liquidated on its own if price moves far enough. A hedge across two venues has separate margin on each.

Does Papertrade have funding rates?

No. Papertrade settles synthetic perps against its own pool at Hyperliquid's mid with no funding, order book, spread or slippage.

Can I backtest a funding strategy?

Yes, on one venue at a time. Circuit replays recorded Hyperliquid or Lighter markets and pays funding at each hour's opening price. Cross-venue strategies are tested in paper trading on live markets.

Bots you can run now

Start trading