Bitcoin is offered at 100 on one venue and bid at 101 on another. The screen appears to show a one-unit profit—until order depth, two sets of fees and the time needed to rebalance capital enter the calculation.
Crypto arbitrage trades price differences for the same or closely related exposure. Professional execution often uses inventory already positioned at several venues because transferring an asset after spotting the gap can take longer than the opportunity lasts.
Why price differences exist
Crypto trading is fragmented across centralised exchanges, decentralised protocols, networks and currency pairs. Each venue has its own participants, liquidity, settlement path and access restrictions. Prices can diverge when information reaches markets at different speeds, withdrawals are constrained, demand is regional or capital cannot move freely.
Research by Igor Makarov and Antoinette Schoar on trading and arbitrage in cryptocurrency markets documented persistent cross-market price differences and the role of capital controls and settlement frictions. The exact gaps change, but the underlying lesson remains useful: a difference can persist because exploiting it is difficult or risky.
Main forms of crypto arbitrage
Cross-venue arbitrage. The trader buys on one venue and sells on another. To avoid transfer delay, inventory may be held on both venues and rebalanced later. That reduces timing risk but increases custody and counterparty exposure.
Triangular arbitrage. Three pairs on one venue imply an inconsistent exchange rate—for example, moving from a stablecoin to BTC, BTC to ETH and ETH back to the stablecoin. All three fills, minimum sizes, rounding rules and fees determine whether the cycle is profitable.
Spot–derivative basis trades. A trader combines spot exposure with a futures or perpetual futures position. This is not risk-free: funding can change, collateral can be liquidated and the two markets may use different prices or settlement rules.
Onchain arbitrage. A transaction or bundle trades across pools when their implied prices diverge. It competes for block inclusion and ordering. Ethereum’s explanation of maximal extractable value shows why arbitrage is closely connected to transaction ordering, searchers and block builders.
Calculate executable profit, not the headline spread
Suppose the best displayed ask is $100 on venue A and the best bid is $101 on venue B. The apparent spread is $1, but a realistic calculation is:
net result = sell proceeds − purchase cost − trading fees − spread and slippage − funding or borrowing − transfer and network costs − rebalancing cost.
The quoted prices may apply to only a small quantity. A larger order consumes several levels, changing the volume-weighted average fill. The separate calculations for slippage and complete trading costs show how quickly the headline spread can shrink.
Both legs must be modelled independently. Positive execution on one side does not compensate automatically for a failed or delayed second leg.
Risks that make a spread persist
- Execution risk: one leg fills while the hedge is rejected, partially filled or repriced.
- Latency risk: prices move before orders reach both venues.
- Inventory risk: capital held across assets and exchanges changes value while waiting to rebalance.
- Transfer risk: a network, bridge or exchange can delay or suspend deposits and withdrawals.
- Counterparty risk: a venue can restrict accounts, become insolvent or fail to honour balances.
- Leverage risk: borrowed capital introduces interest, margin calls and liquidation.
- Model and data risk: stale quotes, bad symbols or an API outage can produce a false signal.
The CFTC’s virtual-currency advisory highlights volatility, cyber and platform risks that remain relevant even when a strategy is designed to be market-neutral.
Bots do not remove the difficult parts
Automation can monitor more markets and submit orders faster, but it introduces software, API-key and operational risk. Backtests often assume fills at prices and sizes that were never simultaneously available. Fees, rate limits, queue position and outages must be reproduced realistically.
A bot also needs rules for partial fills, stale data, maximum inventory, disabled withdrawals and emergency shutdown. A strategy that works only while every dependency behaves normally is not robust.
A pre-trade checklist
- Confirm that instruments, contract terms and settlement assets are genuinely comparable.
- Use executable depth for the intended size, not the last traded price.
- Calculate every entry, exit and rebalancing cost.
- Define what happens if only one leg fills.
- Verify withdrawal status, network, limits and account eligibility.
- Cap exposure to each venue and API credential.
- Model severe volatility and complete loss of access to one side.
Large, persistent gaps usually carry information about a blocked transfer, restricted account, weak market or mismatched instrument. Identify that obstacle before treating the difference as available profit.
Editorial note: this guide was fully reviewed and rewritten on September 3, 2026. It provides general educational information, not a trading strategy or profit claim.
