RESEARCH AND PRACTICE
Why crypto prices differ across exchanges: an executable comparison
Compare crypto prices across venues using matching assets, quote currencies, timestamps and trade sizes, then account for costs and settlement constraints.
A difference between two exchange prices is a question to investigate, not a profit that can already be collected. The useful comparison is the price at which a defined amount could be bought and sold at the same time, with the required balances in the required places. This workflow separates a genuine venue difference from a stale screen, a different instrument or an incomplete cost calculation.
Define the comparison before subtracting prices
Start with the asset identity, market type, quote currency and intended quantity. A spot token, a perpetual contract and a wrapped representation can share a familiar ticker while creating different claims and obligations. Confirm the contract address where relevant and check whether deposits and withdrawals support the same network. Record bid and ask prices separately. The last trade is a historical event; it is not a standing offer to complete your next transaction. A comparison built from two last prices can therefore show a difference that neither order book currently offers.
Make the two observations contemporaneous
Capture the exchange event time, your receipt time and the time at which the comparison becomes visible. Reject a pair when one side is too old for the intended decision. Match the quote currencies through an executable conversion rather than assuming every dollar-labelled token is worth exactly one dollar. Finally, walk through the available levels for the same base-asset quantity on each venue. The economic comparison is between weighted execution estimates, not merely the smallest ask on one screen and the largest bid on another. Retain the original snapshots so the calculation can be checked.
A hypothetical spread that shrinks before any transfer
Suppose venue A offers one coin at 100.00 quote units and venue B bids 100.60 for one coin. With hypothetical fees of 0.10% on each trade, the purchase costs 100.10 and the sale produces 100.4994, leaving 0.3994 quote units before transfers, conversions or rebalancing. These are invented teaching inputs, not exchange fee quotations. The calculation also assumes the whole coin is available at both displayed prices. If only a small fraction is offered at the attractive level, the result must be recomputed using the deeper book.
Separate inventory from the transfer problem
Selling immediately on venue B requires an available balance there or a separately specified borrowing arrangement; a purchase on A does not instantly create that balance. Transferring the coin later introduces confirmation time, withdrawal processing, network selection and an interval during which the price can change. Keeping inventory on both venues changes the problem but adds custody exposure and a need to rebalance. Write down which path the analysis assumes. Do not report the simultaneous-price calculation as the outcome of a sequential transfer strategy with different timing and constraints.
Design a repeatable observation sheet
For each comparison, record product identifiers, market status, quantity, both book snapshots, quote-conversion prices, account-specific fee assumptions and the oldest input age. Add a separate field for whether transfers are available; missing status is unknown, not permission. Recalculate at several sizes and during different market conditions. A narrow apparent difference that vanishes when size increases has a different explanation from a persistent premium with transfer restrictions. Include rejected observations in the log so the review can distinguish lack of opportunity from a failure to collect usable data.
Interpret the difference without promising arbitrage
Academic work by Makarov and Schoar documents segmentation and trading frictions in cryptocurrency markets. It does not establish that any particular displayed difference is available to a particular user today. The practical conclusion here is an analytical one: keep market observations, operational feasibility and realised account changes as separate records. A good comparison explains why a price differs and what prevents that difference from being captured. It may end with no actionable result. That is a more informative outcome than treating every positive subtraction as a risk-free return.
Sources and example scope
Sources support the definitions and mechanisms. Numerical scenarios are hypothetical teaching examples, not live prices, forecasts or reported HOSTuvo returns. Images are editorial illustrations.