Why two venues quote different prices
Two venues, the same nominal asset, two different numbers on screen at the same moment. And a third number on a data aggregator that matches neither.
None of these is wrong. There is no single place where the price of anything is determined, no mechanism that requires venues to agree, and no consolidated book. What exists is a set of separate markets, linked by participants who can act on both, and separated by frictions that limit how tightly that linkage binds.
There is no single book
An order book is a private data structure belonging to one venue. It contains the orders sent to that venue by the participants who have accounts there and assets positioned there.
Which means “the price” of an asset is not a property of the asset. It is a property of a book, and there are as many prices as there are books. Two venues listing the same pair are running two separate markets that happen to reference the same thing, with different participants, different depth, different tick sizes and different rules.
Equity markets in some places have a consolidated feed and rules about routing between venues. Crypto has neither as a structural feature. Nothing obliges a venue to know what another venue’s price is, and nothing obliges an order sent to one to be filled at another’s better quote.
What links them
The linkage is entirely behavioural: participants who can transact on more than one venue, and who act when the difference between two prices exceeds their cost of acting. If one venue’s ask is below another’s bid, someone able to buy on the first and sell on the second finds it worth doing, and the act of doing it consumes depth on both sides and closes the gap. The prices converge because someone traded them together, not because any rule made them.
This site describes that as the mechanism it is and takes no view on it as an activity — the frictions below are real, the capital requirements are real, and nothing here suggests anyone do it. What matters for reading a price is the consequence: convergence is as fast, as complete and as reliable as somebody’s willingness and ability to act, and no more.
What keeps them apart
Every friction between two venues is a width inside which prices can differ indefinitely without anything being exploitable.
Assets have to already be in the right place. Acting on a difference between two venues requires balances on both, before the difference appears. Moving assets takes time — network confirmation, internal crediting, withdrawal processing — and during that time the difference may be gone.
Withdrawals and deposits are not instantaneous or free. Any transfer cost is a direct addition to the width, and any transfer delay converts a certain difference into an uncertain one.
Fees on both legs. Two taker fees plus two spreads crossed plus any impact on either book. The gap has to exceed all of it.
Different assets that look the same. A pair quoted against different quote currencies is not the same pair, and the conversion between quote currencies has its own price and its own spread. Two venues quoting what appears to be one asset may be quoting it against different things.
Access is not uniform. Not every participant can transact on every venue, for reasons ranging from account status to operational capacity. A difference nobody can act on is a difference that stays.
The consequence: the width is not zero, it is not constant, and it grows exactly when transfers are slow or depth is thin — which is when the differences are largest.
Index prices are a construction
Because no venue’s price is authoritative, anything needing a single reference builds one. An index price is a composite of prices observed across several venues, computed by whoever needs it.
The construction involves choices, and every choice changes the number: which venues are included, how they are weighted, whether outliers are excluded and by what test, what happens when a source stops responding, and how often the whole thing is recomputed.
None of those choices is discoverable from the resulting figure. Two indexes on the same asset, both correctly computed, will differ — and the one that matters to you is whichever one the venue holding your position uses for marking and liquidation, regardless of what any other source says.
The mechanism
THE MECHANISM — many books, many prices
· Each venue
→ its own book, participants, tick
size and rules. Its own price.
· A price difference between venues
→ closes only if a participant with
balances on BOTH chooses to act.
· Transfer time and cost between venues
→ sets a width inside which prices
can differ indefinitely.
· An order sent to one venue
→ NO ROUTING. It cannot be filled
at another venue's better quote.
· Thin depth or slow transfers
→ the width grows. Differences are
largest when they are hardest to
act on.
· An index price
→ a construction. Venue choice,
weighting and outlier rules all
change the number.
· Which venues an index includes, its
weights and its failure handling
→ VENUE-SPECIFIC. The index that
governs your position is the one
your venue publishes.Worked example
Illustrative figures, synthetic throughout. Suppose Venue A quotes 40,000 bid / 40,004 ask, and Venue B quotes 40,030 bid / 40,036 ask, on the same nominal pair at the same moment.
Read as one market, that looks like an obvious inconsistency: B’s bid is 26 above A’s ask.
Now cost the round trip for someone who already holds balances on both. Buy 1 unit on A at 40,004, sell 1 unit on B at 40,030 — a gross difference of 26. Apply illustrative taker fees of 0.05% on each leg, about 20 per leg, for 40 in fees. The 26 gross is already a loss of about 14 before anything else, and before any impact if the quantity at either touch is smaller than 1 unit.
So the difference is not a gap in a single market; it is smaller than the width, and the two prices can sit like that for as long as they like. A visible difference between two venues is not an inconsistency until it exceeds the cost of the two trades that would close it.
Now change one input: suppose transfers of the asset between venues are slow at that moment. Someone without a balance already on A cannot buy there and sell on B without a transfer, and the difference may be gone by the time it completes. The width for them is much larger than for someone pre-positioned, and it is larger for everyone at exactly the moments when transfers back up. Note too what an index built from both venues would print: something between the two, depending entirely on the weighting — not either venue’s price, and not tradeable anywhere.
Which price is “the” price
For any specific purpose, exactly one, and it is not usually the one people quote. For your execution, the book you are sending to, since another venue’s better quote is not available to your order. For your margin and liquidation, the venue’s mark or index price, which may be a composite that never appeared as a trade anywhere. For your accounting, whatever your fills actually were: the prints, not the averages, not the display. A single number described as the price of an asset is a summary of several inconsistent measurements, and asking “which venue, at what moment, on which side of the spread” turns it back into something well-defined.
The failure mode
The characteristic error is comparing a price you saw to a price you could have got, across venues, and concluding that something went wrong. Nothing routed anywhere. Your order met the book you sent it to, at that book’s prices, and the other venue’s number was never available to it.
The quieter failure is the index one. A position can be liquidated on a reference computed from venues you have never used, at a level that never printed on the venue you are watching, and the chart in front of you will show a market that never reached the price the decision was made at. That behaviour is by design and it is well documented by whichever venue holds the position — but it is only discoverable in advance, by reading which reference they use, not afterwards from the chart.