Run the bot yourself, or hand it over

This page is not here to talk you out of building it. Both routes get laid out on the same table — the work each demands, the costs each deducts, the risk each carries. Decide after reading that, rather than after someone tells you it is hard.

Some people genuinely should run it themselves

Funding rate arbitrage is not a secret formula. The mechanism is public, the exchange APIs are public, and there are working open-source implementations on GitHub. If you have enough capital to fill both legs, you can read and change the code, and you do not mind being on call for it, then running it yourself is strictly better economically: no management fee, everything left after costs is yours, and the money never leaves your own accounts. Everything listed below is then workload, not a barrier.

So the question this page actually answers is the other one:if one of those three does not hold, how large is the gap. The next three sections are the work, the costs, and the risks neither route escapes. The alternative route is usually calledmanaged or delegated arbitrage — capital is executed by a desk on the same strategy, and you touch neither the code nor the screen. Its cost gets its own section near the end, not a footnote.

What self-hosting actually puts on your plate

Opening the position is the easy part, and it is where most tutorials stop. The time goes into every day after that. The third column is what happens when something is missed — not a reminder, the consequence.

AreaWhat has to be handledWhat missing it costs
Balancing the two legsSize by notional, not by contract count — and the two sides are not even marked against the same priceGet it wrong and you are left with directional exposure: the funding you collect gets eaten by price moves, and the account will not show you why
The 8-hour settlementWatch the position at each settlement; before rebalancing across one, work out whether you still collect that periodMiss a period and you lose all of that period. There are over a thousand of them in a year
Margin on the futures legWhen spot rises the short leg goes underwater and margin has to follow. In fast markets that is a minutes-level jobFail to top up and you are liquidated. After that only the spot leg remains, and delta-neutral has become an outright long
Funding turning negativeDecide whether it is one noisy period or a regime change, then decide whether to exit — and exiting costs fees on both legsExit early and you paid a round trip for nothing; exit late and you have moved from being paid to paying
The exchange sideAPI rate limits, reconnects, maintenance windows, contract spec changes, delistingsWhile you are disconnected nobody is watching the position, and the market does not wait for you to reconnect
ReconciliationTrack funding per period, fees on both legs, and cost of capital separately — not merged into one NAV figureCounting only the futures-leg fee is the most common version, and it quietly inflates your net figure

The row most often underestimated is the third.Liquidation is not "losing a bit" — it is the structure of the trade being severed.With the short leg gone, what remains is a full outright long in spot, and avoiding directional exposure was the entire reason for choosing this strategy.

Costs: what each route deducts

This table does not compare numbers, it compares which lines get deducted. Exchange fee tiers move with volume and can be re-cut at any time, so a specific percentage printed here would be wrong next year. The structure of who bears what does not move — and if you are working out a break-even, the structure is the part you need.

TopicRun it yourselfHand it over
Management feeNone2% a year
Fees on both legsYours, at whatever tier you sit on at the exchangeAlso paid, and already deducted in the net figures we report
Size of principalWhatever you have, split across two legsscale principal 10x without directional risk, and that scaling capital carries no interest
Cost of capitalScaling up means borrowing, charged hourly, and the interest comes straight out of the fundingCharged as interest on scaling capital — currently zero, but still carried as a cost line
InfrastructureServers, market data, alerting, 24/7 cover — all yoursCarried by the platform
Ongoing timeGoing live is the start: tuning, reconciliation, keeping up with exchange rule changesNone
Who keeps the yieldAfter costs, all of it is yoursSplit by plan; the protected plan pays a 7% floor
Who carries it when it breaksYou do — mis-sized, disconnected, liquidated, nobody reimbursesThe platform executes, but you carry the credit risk of having funds with a platform

No specific fee percentages are quoted here. This project publishes no fee schedule, and printing something like "around 0.06%" would make it the one number on this page with no source behind it.

What handing it over costs you

The fee is the visible part and the easy part.The real cost is that the funds sit in a platform account. That converts a set of operational risks you would have carried yourself into a single counterparty risk you cannot hedge and cannot monitor: if the platform fails, being right about the funding rate will not help you.

There is a second one worth naming: you give up the ability to intervene. When funding inverts, the decision to hold or exit is made by the desk, on its schedule, not yours. For someone who would have wanted to make that call personally, that is a genuine loss, not a convenience.

Both are set out line by line on risk and boundaries, alongside the ones that have been removed. The entry size for the managed route is 5,000 USDT.

What neither route escapes

Both routes are exposed to the same market. Funding can turn negative — across 7,344 BTC settlement periods, 14.05% were not positive, and they are not evenly spread; they cluster. Exchanges can change contract specs, widen maintenance windows, or delist a pair. Neither a bot you wrote nor a desk you hired removes any of that.

And the returns are thin in the bad stretches for both. Across 2021–2025, ETHUSDT in 2022 produced 7.873% for the whole year. That year is in the table onperformance at the same weight as the good ones — because a comparison that only shows the good years is not a comparison.

Six questions, answerable on the spot

None of these asks how you feel about risk — everyone answers that one the same way. Each asks something factual. If you can answer yes to all six, build it yourself; this page has saved you a management fee. If two or three are no, the gap between the routes is larger than the fee.

  • You have enough of your own capital to fund both legs, not just one.
  • You can read and modify order-routing and risk code, and do not need to wait for anyone when it breaks.
  • You have a stable machine to run it on, and you are willing to own the disconnects and the 3am alerts.
  • You know exactly what happens the moment margin cannot be topped up, and you have rehearsed it.
  • You are willing to reconcile every period, not glance at a total once a month.
  • You find the work itself interesting — because for a long stretch, the hourly rate on it is poor.

The last one is not filler. Plenty of people build the bot, run it for two months, and then stop reconciling — and an unreconciled delta-neutral position is just a position nobody is watching.

Next

Plans and termsGet started