Prices are always relative. Nothing has a price on its own; it has a price in terms of something else, and that something else is chosen, usually by an interface rather than by the person trading. The choice determines what you pay, what depth you reach, and what your performance is measured against, and almost nobody notices making it.
A market does not price an asset, it prices an asset against another asset. What is displayed as a single number is always a ratio, and the denominator is as much a part of the trade as the numerator.
Most venues offer the same asset against several denominators: a national currency, one or more assets designed to track one, and sometimes a major crypto asset. These are separate markets with separate books, separate depth and separate prices, not different views of one thing.
The practical consequence is immediate. Buying the same asset through two different denominators is two different transactions with two different costs, and the difference between them is frequently larger than the trading fee everybody compares.
Nothing has a price. It has a price in terms of something else, and that something else is a decision, usually made by an interface rather than by you.
Not every combination exists. If you hold one asset and want another and no market pairs them directly, the exchange happens in two steps through an intermediate, and the interface may do this without saying so.
Two steps means two crossings. You pay a spread and a fee on the first leg, and again on the second, so the total cost of a routed trade is roughly double that of a direct one at comparable depth. That doubling is the single largest routing effect and it is invisible on any confirmation screen.
The habit worth building is to check whether the pair you want exists before assuming it does. When it does, using it is a saving that requires no skill. When it does not, knowing that the trade is two trades changes how you size it and when you place it.
It is worth being precise about what doubles. The trading fee doubles because it applies to each execution. The spread crossed doubles because there are two books. Slippage on a large order applies twice, and on thin intermediate markets the second leg is frequently the worse of the two.
What does not double is anything charged per transfer rather than per trade, which is why the effect is specific to routing rather than to activity in general. This is a cost of the path, not of the decision to trade.
For small amounts the doubling is a small number and genuinely not worth optimising. For anything substantial or anything repeated, it is one of the larger avoidable costs available, and avoiding it usually means one different click.
On most venues the trading fee is deducted from what you receive rather than billed separately, which means it is taken in one of the two assets of the pair, and which one depends on the direction of the trade. Buying leaves you with slightly less of the asset bought; selling leaves you with slightly less of the denominator.
That detail is why a sequence of trades leaves balances that do not match the arithmetic of the prices alone. The discrepancy is small, systematic and always in the same direction, which makes it invisible on any single transaction and perfectly visible on a reconciliation across many.
Some venues offer a discount for paying the fee in a third asset instead, which is a separate decision with its own consequence: it requires holding a balance of that asset, and that balance is a position taken in exchange for the discount. Whether that is worth it depends on how much you trade and on what the third asset does while you hold it, and both are knowable before opting in.
Holding a balance in the denominator is holding that asset, with everything that implies. If it is designed to track a national currency, you are exposed to whatever keeps it tracking. If it is a major crypto asset, you are exposed to its price while you wait.
That second case is the one that surprises people. A trader who sells one asset for another crypto asset intending to buy back later has not gone to cash, they have swapped one exposure for another, and their result depends on both legs rather than on the one they were thinking about.
Neither is wrong and the choice should be deliberate. The failure is holding a denominator by accident because it happened to be the pair the interface offered, and then measuring performance against something you did not choose.
The same asset can be deep against one denominator and thin against another on the same venue at the same moment. Depth follows habit and habit follows the default, so whichever pair the interface presents first tends to accumulate most of the activity.
This means the theoretically better route is sometimes the worse one in practice. A direct pair with almost nothing resting in it can cost more than two legs through deep markets, and the only way to know is to look at both books rather than reason about the structure.
Checking takes seconds and it is the same check as always: what is resting near the top on each side, and how far does it go. A route with no depth is not a route, whatever the pair listing suggests.
Many interfaces choose the path for you, sometimes across several legs, and present a single result. That is useful and it raises a question worth asking: what is being optimised, and is it the same thing you would optimise.
A router aiming for the best displayed rate may take a path with more legs, and one aiming for fewest legs may accept a worse rate. Neither is wrong and they produce different outcomes, and which one you are using is usually documented somewhere and rarely on the screen where you trade.
The check is to compare the routed result against the direct pair when a direct pair exists. If the router is doing well the numbers will be close, and if they are not, that is worth knowing before the next hundred transactions rather than after.
A router optimises something. Whether it optimises the thing you care about is documented somewhere, and almost never on the screen where you trade.
Quoting in an asset designed to hold a fixed value makes arithmetic easy: gains and losses are expressed in something whose value is intended not to move, and comparisons across time mean what they appear to mean.
That convenience rests on an assumption about the asset holding its value, which is a property of its design and its backing rather than a law. The assumption is usually sound and it is an assumption, and a denominator that stops tracking makes every price quoted in it wrong at once.
The practical implication is not to avoid such denominators, which would be impractical. It is to know which one you are using, because they are not interchangeable and the differences between them are in how they are backed rather than in how they trade.
The denominator you trade in becomes the unit your results are reported in, and that choice changes the numbers without changing anything real. A position that gained against one asset may have lost against another over the same period, and both statements are true.
This matters most for anybody quoting in a volatile denominator. Results expressed in a crypto asset are a comparison against that asset, so a position that rose in absolute terms can show a loss simply because the denominator rose more.
Deciding once what you are measuring against, and keeping it, is worth more than it sounds. Most confusion about whether a period went well comes from an unstated change of denominator rather than from any disagreement about what happened.
Four things, all visible in under a minute. Whether a direct pair exists for what you want. What depth is actually resting on each candidate route. Whether the interface is routing for you and along what path. And what you end up holding in the denominator once the trade is done.
The fourth is the one that gets skipped, and it is the one with a tail. Finishing a trade holding a balance in an asset you never chose, in an amount you never sized, is a position taken by default, and default positions are the ones nobody reviews.
None of this requires any view about prices. It describes the mechanics of getting from one holding to another, which is the part that is entirely knowable before acting and entirely fixed afterwards.
Routing is one of the few costs a trader controls completely. It requires no forecast, no timing and no skill beyond looking at two books instead of one, and the saving is available on every transaction rather than occasionally.
It is also one of the least discussed, because it is unglamorous and because the interfaces that would explain it are the same interfaces that benefit from the default. That is not a conspiracy, it is a consequence of designing for the majority who want one button.
For anybody exchanging more than occasionally, the minute spent understanding which pairs exist on the venue they use is repaid many times over, and it is repaid on every transaction from then on rather than once.
The denominator of a pair: the thing an asset is priced in. A displayed price is always a ratio, and most venues offer the same asset against several denominators, which are separate markets with separate books, separate depth and separate prices rather than views of one thing.
Because two steps means two crossings. The trading fee applies to each execution, the spread is crossed in two books, and slippage applies twice. The total cost of a routed trade is roughly double a direct one at comparable depth, and none of it appears on a confirmation screen.
No. Depth follows habit, so a direct pair with almost nothing resting in it can cost more than two legs through deep markets. The only way to know is to look at what is resting near the top of each book rather than to reason about the structure.
Yes, with everything that implies. Selling one asset for another crypto asset is swapping one exposure for another rather than going to cash, and the result then depends on both legs. The failure is holding a denominator by accident because it was the pair the interface offered.
It depends on the router, and it is usually documented somewhere other than the screen where you trade. One aiming for the best displayed rate may take more legs; one aiming for fewest legs may accept a worse rate. Compare a routed result against the direct pair to find out.
It makes arithmetic easy because gains and losses are expressed in something intended not to move. That rests on an assumption about the asset holding its value, which is a property of its design and backing rather than a law, and such assets are not interchangeable with each other.
Because the denominator is the unit your results are reported in. A position that gained against one asset may have lost against another over the same period, and both are true. Most confusion about whether a period went well is an unstated change of denominator.
Whether a direct pair exists, what depth is actually resting on each candidate route, whether the interface is routing for you and along what path, and what you end up holding in the denominator afterwards. The last is the one that gets skipped and the one with a tail.