4 / 4 · 7 min
Arbitrage: why a 17% spread usually yields nothing
The difference between exchanges is visible and measurable. But between it and money stands a long list of deductions, and most often it eats everything — and eats the largest spreads whole.
The word 'arbitrage' sounds like riskless profit: buy cheaper there, sell dearer here. The mechanics really are simple, which is exactly why such opportunities do not linger. Everything you see as an open gap has already been seen by thousands of people with faster access. If the gap is still there, it has a reason, and that reason is usually cost.
What gets subtracted from a spread
- Fees on both sides: the buy and the sell, each with its own percentage.
- Slippage: the quoted price refers to the top of the book, while your size goes deeper.
- Transfer between exchanges: network time and network fee, and on some coins withdrawals are closed entirely.
- Funding, if one leg is a perpetual: holding costs money every few hours.
- Borrow, if you are selling something you do not own.
- Execution risk: one leg fills, the other does not — and you are no longer an arbitrageur but the holder of a directional position.
The last term has no exact number, but it cannot be dropped: it is precisely what turns an 'arithmetically profitable' operation into a losing practice. The whole line must be computed, and computed BEFORE the trade.
A live spread and what is wrong with it
Measured on our spreads page: the coin NIGHT, spot on HTX at 0.01744, futures on KuCoin at 0.0204269 — a difference of 17.13%. We checked both prices directly against the exchanges and they matched to the digit: the spread is real, not a data error. Yet earning from it is nearly impossible, and here is why. This is spot against a future, so not 'buy there, sell here' but a two-legged trade with a carrying cost. The coin's turnover is small, so your size will dig into the book. Moving the coin between venues takes time during which the gap may close. And above all: a gap of that size persists precisely because it is hard to take — otherwise it would already have been taken.
Why we discard spreads above twenty percent
Our spreads page has three honesty filters, and one of them looks strange: differences above 20% are not shown at all. The reason is simple — almost always that is not a market but a fault: a wrong contract multiplier, a namesake coin, a frozen price of a delisted instrument. Showing them would mean promising an opportunity where there is a data defect. The other two filters: both sides must have a live feed no older than a minute and turnover of at least a hundred thousand dollars. Without them the list fills with dead pairs you cannot execute on.
Kinds of arbitrage and what is actually hard in each
| Kind | Essence | Where it breaks |
|---|---|---|
| Cross-exchange | Buy on one, sell on another | Coin transfer, speed, withdrawals may be closed |
| Spot-perp | Spot against a perpetual | Funding eats the basis, margin needed on both legs |
| Cash-and-carry | Spot against a dated future | Money locked until expiry, margin requirements |
| Triangular | Three pairs inside one exchange | Fees three times, the gap lives for seconds |
| Statistical | Related assets diverge | This is no longer arbitrage but a bet: the relationship can break |
Treating basis as arbitrage
A future priced above spot is not a market error. It is the cost of carry: longs pay shorts through funding, and the gap reflects exactly that. You can earn on convergence, but you take on funding, margin requirements on both legs, and the risk that the basis widens further before it narrows. None of those words go with 'riskless'.
Live differences between venues with three honesty filters: a fresh feed, real turnover on both sides, and impossible spreads filtered out.
Open the spreadsCompute it to the end
Take the top row of our spreads page and carry the calculation through to a net number: find both venues' fees, look at book depth for the size you want, find the network fee for the transfer and its confirmation time. Subtract everything from the spread. Do this three times on different rows. Most likely nothing positive will remain — and that is the most useful result this exercise can give.
You see a 4% spread on a coin with $200k of daily turnover. What stops you taking it?
Depth above all: at that turnover the book is thin and your size will fill at a noticeably worse price than the top row. Then the transfer time between exchanges and the risk that the gap closes during it. Plus fees at two venues and the network. A 4% spread on an illiquid coin is usually 4% on paper and zero or negative in execution, not an opportunity everyone overlooked.