Arbitrage bot explained: the edge that's harder than it looks
Arbitrage sounds like free money: buy an asset cheaper on one venue, sell it dearer on another, pocket the difference. Arbitrage bots automate the hunt. But the gaps are tiny, vanish in milliseconds, and are eaten by fees, transfer times and execution risk. Here is how the main types work and why retail arbitrage rarely pays.
The main types of arbitrage
- Cross-exchange — the same asset trades at a higher price on exchange B than exchange A; buy on A, sell on B. See our arbitrage trading bot guide for the mechanics.
- Triangular — exploit pricing inconsistencies between three pairs on one exchange (e.g. BTC→ETH→USDT→BTC) so you end with more than you started.
- Funding-rate — hold spot long and a short perpetual to collect a positive funding rate while staying market-neutral. This is the most accessible form, but it is also a known crowded trade closely tied to statistical arbitrage.
Why the edge is tiny and fleeting
Price gaps between liquid venues exist for a reason — they are small and they close fast, often within milliseconds, because thousands of bots are racing for them. By the time a retail bot detects a gap and sends orders, faster, co-located professional systems have usually already taken it. The remaining gaps tend to appear precisely when execution is hardest: during volatility spikes, withdrawal freezes, or on thin, risky venues.
The costs that erase the edge
A profitable arb must clear: trading fees on both sides, withdrawal and network fees if you move funds, the spread and slippage on each fill, and the time and price risk of moving an asset between exchanges (during which the gap may close or reverse). A "2% arb" can easily net negative after all of these. Add counterparty risk — the cheaper venue may be cheaper because it is in trouble.
Can retail traders actually do it?
Pure latency arbitrage is effectively closed to retail — you are racing firms with co-located servers and fee rebates. The one form that remains viable is the slower, capital-based funding-rate carry, which is market-neutral and earns the funding spread rather than racing for price gaps. Even then, returns are modest and risks (exchange solvency, sudden funding flips) are real. Model every cost in the backtester before assuming an arb is profitable.
Frequently asked questions
Are arbitrage bots profitable?
Rarely for retail. Price gaps between liquid venues are tiny and close within milliseconds as professional, co-located bots take them. After fees, spreads, slippage and transfer risk, most retail cross-exchange arbitrage nets negative.
What is triangular arbitrage?
It exploits pricing inconsistencies between three trading pairs on a single exchange — for example converting BTC→ETH→USDT→BTC — to end with more than you started. Such gaps are rare, small, and contested by fast bots.
Is funding-rate arbitrage safer?
It is the most accessible form: you hold spot long and a short perpetual to collect a positive funding rate while staying market-neutral. Returns are modest and it carries exchange-solvency and funding-flip risk, but it does not require winning a latency race.
Why can't retail traders win latency arbitrage?
Professional firms use co-located servers, direct market data, and fee rebates, letting them detect and take gaps in microseconds. A retail bot on a normal connection is far too slow to compete for those gaps.