Every time an order executes immediately, somebody had already placed the order it executed against. That somebody is usually not expressing a view on the price, and understanding what they are doing instead explains a long list of things that otherwise look arbitrary, starting with why the market gets more expensive precisely when it gets interesting.
A market maker quotes both a price to buy and a price to sell, continuously, and stands ready to transact at either. The product being sold is immediacy: the ability for somebody else to trade right now instead of waiting for a counterparty to appear.
The compensation for providing it is the spread, the gap between the two quotes. If a maker buys at the lower price and sells at the higher one, the difference is revenue, and doing that many times is the business. Direction is not the point; turnover is.
This is why describing a maker as betting against you is a misunderstanding. They would generally prefer to hold nothing at all and simply capture the spread on a matched pair of trades. Any position they end up holding is an unwanted by-product, and most of their engineering exists to get rid of it.
The product is immediacy. The maker would rather hold nothing and capture the spread twice; anything they end up holding is an unwanted by-product.
The first is inventory risk. Having bought from somebody, the maker holds an asset whose price can move before they sell it, and that exposure is real money. The wider the spread, the more the position can move before the trade turns unprofitable.
The second is adverse selection, and it is the deeper one. Some of the people trading against a quote know something the quote does not reflect, and those trades are systematically unprofitable for the maker. The spread has to be wide enough to cover the losses to informed traders out of the gains from uninformed ones.
Every observable behaviour of a quoting system follows from balancing these two. A spread is not a fee somebody chose, it is a price computed continuously from an estimate of how much inventory risk and how much adverse selection the current conditions carry.
Consider the position from the maker's side. When somebody buys from them and the price then rises, they have lost; when somebody buys and the price then falls, they have gained. The difference is whether the buyer knew something, and makers cannot tell which is which at the moment of the trade.
What they can do is measure it afterwards, in aggregate, and adjust. If trades in an instrument are followed by moves against the maker more often than chance would suggest, the spread widens until the remaining flow is profitable. That adjustment is automatic and it is not personal.
This explains a pattern that traders notice and rarely explain correctly. Instruments where information arrives suddenly have wider spreads than instruments where it does not, independently of volume, because the risk of being on the wrong side of somebody who knows is higher.
A maker who has accumulated a large position in one direction is no longer neutral, and their quotes will show it. They will quote more attractively on the side that reduces their position and less attractively on the side that increases it, because their priority has shifted from capturing spread to getting flat.
That is why the book is often visibly lopsided without anybody having a view on the price. More depth on one side can simply mean that whoever is quoting wants to trade in that direction, which is a statement about their inventory rather than about the asset.
Reading it as a signal about future prices is the common error. The skew is real information about who wants what right now, and it says nothing about where the price goes, because the party creating it is explicitly not trying to predict that.
Both risks rise together in fast conditions. Inventory becomes more dangerous because prices move further before a position can be unwound, and adverse selection becomes more likely because that is when informed participants are most active. A maker facing both raises the spread or withdraws entirely.
This produces the pattern every trader has experienced and most experience as unfairness: the market is cheapest to trade when nothing is happening and most expensive at exactly the moment they want to act. It is not a policy and nobody decided it, it is the price of the risk being taken.
The practical implication is that the cost of trading is a function of when, not only of what. Choosing the moment is one of the few levers available to a participant, and it is available at no cost to anybody who can wait.
In many traditional markets, designated makers have formal obligations: a maximum spread, a minimum size, a percentage of the session during which they must be present. Those obligations are the counterpart to privileges granted by the venue.
In most crypto markets there is no such obligation. Participants quote because it is profitable and stop when it is not, which means the depth visible on a screen is a description of current conditions rather than a commitment about future ones.
That distinction is worth carrying. Depth that exists because somebody chose to provide it can disappear in the second it stops being attractive, and the moment it stops being attractive is the moment everybody wants it. Nothing about the display distinguishes committed depth from voluntary depth.
Depth that exists because it is profitable disappears when it stops being profitable, which is exactly when everybody wants it. The screen does not distinguish the two.
Venues want depth because depth attracts traders, so many of them pay for it. The common arrangement rewards resting orders relative to aggressive ones, which pays participants to sit in the book rather than take from it.
Some venues go further with formal programmes: better terms in exchange for measured obligations on spread, size and uptime. These arrangements are usually documented and their existence tells you something about how the depth on that venue is produced.
For a trader, the useful consequence is that some visible depth is there because it is subsidised. That does not make it fake, it makes it conditional on the subsidy, and knowing whether the depth you rely on is voluntary, subsidised or obligated is a real distinction between venues.
Where several makers compete for the same flow, spreads compress towards the level that just covers the two risks. That compression is a genuine benefit to everybody crossing, and it is the strongest argument that this activity is useful rather than extractive.
The compression has a floor. A maker cannot quote inside the cost of adverse selection and inventory risk for long without losing money, so there is a spread below which nobody quotes regardless of competition. Observing a market at that floor is observing a competitive market, not a broken one.
Competition also concentrates. The participants best at estimating the two risks quote tighter, capture more flow, and improve their estimates further, which is why this activity ends up dominated by a small number of firms in most markets. That concentration is a consequence of the mechanism rather than a failure of it.
That concentration carries a consequence worth naming for anybody relying on the depth it produces. A market served by very few quoting participants is tight while they are all present and thin the moment one of them steps away, which is a fragility no average measure of spread will show. The measurement that would reveal it is how depth behaves during stress, and that is observed by watching rather than by reading a statistic.
Three things become readable once you know what the quoting party is doing. A wide spread means the two risks are estimated as high right now. A lopsided book usually means somebody is managing inventory. And depth that vanishes as price approaches means quotes were being pulled rather than filled.
None of these predict direction, and the temptation to read them as prediction is the main way this knowledge gets misused. They describe the current cost and current willingness to transact, which is a description of conditions rather than a forecast.
What they do support is a decision about whether to act now or wait, which is a smaller question and one that this information actually answers. That is the honest use of it and it is worth more than the version people wish were true.
The spread you cross is the price of not waiting, set by somebody else's estimate of two risks. It is not arbitrary and it is not negotiable, and the only lever available is whether to pay it at all, which is the difference between an aggressive order and a resting one.
A trader who rests rather than crosses is standing on the other side of that trade, accepting the same two risks in exchange for not paying the spread. That is a real trade-off rather than a free improvement, and the risk being accepted is precisely the one described above.
Understanding what the quoting party is doing does not make anybody better at predicting prices. It makes them better at knowing what they are paying for and when it is most expensive, and that is a smaller claim that happens to be true.
Generally no. They quote both sides and their business is capturing the spread on turnover, not taking a view on direction. Any position they end up holding is an unwanted by-product, and most of their engineering exists to get rid of it as quickly as possible.
Two risks. Inventory risk, because a position acquired can move before it is unwound. And adverse selection, because some counterparties know something the quote does not reflect, and the spread must cover those losses out of the gains from everybody else.
The systematic tendency for informed traders to trade against a quote precisely when it is wrong. A maker cannot tell who is informed at the moment of a trade, only measure it afterwards in aggregate and widen the spread until the remaining flow is profitable.
Usually because whoever is quoting has accumulated a position and wants to reduce it. They quote attractively on the side that flattens them and unattractively on the side that adds. It is information about their inventory, not about where the price is going.
Both risks rise together. Inventory becomes more dangerous because prices travel further before a position can be unwound, and informed participants are most active at exactly those moments. The maker either widens or withdraws, which is the price of the risk rather than a policy.
In many traditional markets designated makers have formal obligations on spread, size and presence, granted in exchange for privileges. In most crypto markets there is no such obligation: participants quote while it is profitable and stop when it is not, and the screen does not distinguish the two.
Frequently. The common arrangement rewards resting orders relative to aggressive ones, and some venues run formal programmes offering better terms for measured obligations. Some visible depth is therefore conditional on a subsidy, which is not fake but is worth knowing about.
No, and that is the main way the knowledge gets misused. A wide spread, a lopsided book and vanishing depth describe current cost and current willingness to transact. They support a decision about whether to act now or wait, which is a smaller question they actually answer.