Introduction

Two protocols, Liminal and Falcon Finance, have TVLs that are skyrocketing (respectively $70m and $1,2b) thanks to the same strategy: arbitrage of funding rates.

Falcon launched in feb 2025 and received $10m funding from WLFI; Liminal just opened their public beta, and are swiftly getting traction.

I wondered if an individual could replicate and improve upon this strategy. We will therefore in this article study how it works, lay the base principles out, to prepare the implementation in a future article.

How does Funding rate farming / arbitrage work ?

Funding rates

They are fees collected by perpetuals-futures platforms. For a given token, either short or long holders will pay this fee. It’s computed and transferred periodically (hourly on Hyperliquid).

Since perpetual contracts don’t expire, platforms continuously re-anchor the prices of the contracts. Funding rates also keep the price of the contract aligned with spot prices.

Funding Rate = Premium Index + Interest Rate
Premium index measures the deviation between perp and spot price.
Interest Rate represents the cost of holding capital on the margin platform

If the interest rate is 0.01% and the premium index is 0.02%, the funding rate would be 0.03%.

For example, if Bitcoin’s spot price is $30,000 but the futures price is $31,000 due to high demand for long positions, a positive funding rate incentivizes long holders to pay short holders, encouraging a rebalancing of positions.

Real life scenario

Let's take an example :

As of Sunday, 3rd August, the market has fallen quite a bit from the recent ATHs. Fear and greed is back in neutral after almost one month of greed sentiment.

We therefore are in a bearish mood overall. Bears are shorting aggressively. This translates into Hyperliquid balancing out the shorting prices of several coins so shorts pay longs. Using the Hyperliquid SDK, we were able to retrieve some funding rates to illustrate :

Here, people shorting BERA pay 0.0190% of their position value to people longing, each hour. This rate is subject to change in between two payments. The last value at the last second of the last minute of every hour is taken as the funding rate for this hour. We can see that a few minutes later, the funding rate for BERA dropped to -0.0187% and that OMNI funding rate climbed to -0.0236 (previously -0.0224%).

If we were to open a $ 10 000 long on BERA, and the funding rate stayed the same until the next payment, we would receive $1.87. If the rate stayed the same for an entire year, our annualized profit would be 164% (ignoring fees and slippage).

Differences between Liminal and Falcon

Liminal focuses its entire strategy to holding delta neutral strategies on Hyperliquid. This essentially allows them to get paid to hold their positions.

Falcon finance makes funding rates arbitrage a part of their activity, among a bunch of other yield generating strats. On their docs, they also mention Cross-exchange Price Arbitrage, Native Altcoin Staking, Liquidity Pools.

Here is a table comparing the two protocols:

Who else does that ?!

Other major players execute this strategy, onchain hedge funds and major protocols. Namely, Active Digital the "Best Performing Fund over 2 Years" reports using this strat. We don't have much info about it but their site mentions : "Mid frequency, funding rate arbitrage strategy covering liquid cryptocurrencies.". The strategy yielded a cumulative 95% return over 3.5 years (2022 to mid 2025). The metrics look promising, max drawdown -2.31% (really good for crypto), 97% win months, 100% win year.

Although less transparent on their strategy's details and returns, Pythagoras Investments do rates farming on their Arbitrage Fund. They were elected Best Performing Fund over 5 and 10 Years.

Finally, Pendle (7b + TVL), Top 16 in protocol in revenue (07, august 2025) just launched Boros. Boros is a new way to long or short the funding rates of assets. For now, only BTC and ETH are available but we can imagine a lot more tokens available in the future.

Risks

There are 3 issues to using this strategy : volatility, changing in funding rates, fees. Here is how we plan to counter each to have a winning strat.

Token Volatility

Taking back our example, we are long BERA, and capture 0.0187% of our position each hour. However, if BERA completely plummets, the loss on our position might negate the funding rates we receive.

This is why we need to make our position delta neutral (DN). This is a technical term to say that we have a long and a short position at the same time. It allows us to break free from the variations - the delta- of the asset we are trading.

Making our position DN is easy in theory, but actually quite hard to implement in an automatic way. We need to consider several things.

First, if we are shorting, we can't just open a long on the same asset on Hyperliquid, as we would be paying an receiving equal funding rates. One option to counter this is to only focus on positive funding rates, and buy the tokens in spot. This way, we would be short on Hyperliquid and long via the spot token (for which we don't need to pay funding rates).

We can do all the above, but also implement funding rates arbitrage. In our examples, the rates are negative, meaning that we must long on Hyperliquid, to capture the funding rate. Shorting a given token without an exchange is harder to do than buying spot. It requires the asset to be available to borrow, this is not the case for most smaller tokens, that tend to have better funding rates opportunities. We need to find another market where the price of holding a short (funding rate) is lower than on Hyperliquid. We shall therefore lay out some exchanges where we would take the opposite position (short to the one on Hyperliquid). For this system to work, the funding rate on the second market shall be lower than the one on Hyperliquid.

Here’s a scenario matrix:

Funding-rate changes

We will focus on Hyperliquid as a primary way of farming funding rates. However, these rates change a lot. If BERA's funding rate goes back to 0, then our long position won't earn anything. We would however have paid fees to open our position. Those fees can make our earnings null.

The best way to counter this issue is to track the variance of funding rate of each token. This data set would then allow our engine to choose a token with a lower funding rate but lower variance, in order to minimize transactions and keep rates steady.

It will take some tweaking in order to get an algo that does not put too much weight on variance. Otherwise it would almost always prioritize BTC as it tends to have the lowest variance. However, it shall still take it into account to minimize transaction number.

Another important metric to implement will be the distance of the current rate from the median of the token. This combined to the standard deviation will allow us to assess the risk of the rate staying in our favor. It will allow us to reduce the number of transactions

Here is an example to illustrate :

Here, the standard deviation, and the distance to the median is way higher for OMNI than for BERA. We would therefore be taking the same risk of the rate flipping the opposite way for both tokens. We shall introduce some level of risk tolerance into our algo to make it more or less risk averse. Here, the annualized rate is so good that it would be ok (in our opinion) to take the risk if there were only those two rates.

Fees

In order to build the most robust strategy, we need to consider all the fees that it could generate. Fees could erode our returns over time.

Open / Close position fees (Hyperliquid)

Hyperliquid takes 0.045% on every position opening and closing

Bridge fees

In order to be able to DN our position, we will buy the assets in spot on either Base or Solana. It was identified that most assets available on Hyperliquid are mostly available on those blockchains. In order to not be forced to deploy capital simultaneously on three blockchains we will need to bridge the assets. We are currently considering using LI.FI in order to get the best prices and lowest fees for our spot buys

What's next ?

The next step is to actually code this idea out. There is no guarantee that it will work, nor if it will be profitable. We at least know that if someone would farm funding rates alone (without delegating to a protocol) this article is a first step towards having an optimized system.

We'll collect funding rate data and build the token selection algorithm, dev the parts that we can in the mean time, and come back with a part 2, with the full system.

There will of course be some articles published in between. Drop a follow if you found this interesting.

always, 888