Market Makers in Crypto: Who They Are and How They Earn

Open the BTC/USDT order book on any top exchange. The best bid and the best ask sit a couple of ticks apart, every level has orders on it, and the market is ready to swallow a hundred bitcoin without noticeable slippage.

Now open the book of a coin that got listed yesterday. The book is empty, there's a 4% hole between prices, and your own market entry moves the price half a percent.

The difference between those two pictures isn't how popular the asset is. The difference is that in the first case there's an algorithm sitting in the book around the clock, one that is obligated to sit there, and in the second case it either isn't there or it left.

Let's go through who these participants are, where their money comes from, what traces they leave in the order book, and how to tell those traces apart from an ordinary large player.

What a market maker is

A market maker (often written as two words) is a participant who holds limit orders on both the buy and the sell side at the same time and earns on the difference between them. He isn't guessing direction. He's trading flow.

The logic is simple. You post bids at 100.00 and asks at 100.02. Someone sells into you at market, you bought at 100.00. Ten seconds later someone buys at market, you sold at 100.02. Two hundredths of a percent net, the position is flat again, the cycle repeats. A thousand times an hour.

The profit from one cycle is laughable. The profit from a million cycles isn't.

In crypto, one word hides at least four different types of participant, and mixing them up is expensive.

• Contracted exchange MMs. Wintermute, GSR, Cumberland, Amber Group and a dozen smaller firms. They work under an agreement with the venue, get fee breaks and API priority, and in exchange commit to holding quotes of a set width and depth for a set percentage of the time
• Project MMs. Hired not by the exchange but by the token team. Their job is to provide liquidity for one specific coin after launch. The economics of those deals get their own section below, and they're interesting
• Independent algo firms. No obligations, pure arbitrage logic. They quote while it pays and pull everything at the first sign of risk
• Small HFT shops and solo operators. They work on one or two exchanges, often in illiquid names where the spread is 2-5% and there's almost no competition

Only the first two have obligations. The rest disappear exactly when you need them most, and that is precisely why liquidation cascades in crypto print such long wicks.

Their role: liquidity and the spread

Liquidity is the market's ability to fill your order at a price close to what you expected. It's measured by two numbers: spread width and book depth. A detailed breakdown of where it comes from and where it goes is in the article "How Liquidity Works in Crypto and Why It Matters".

The spread, put plainly, is the gap between the best buy order (Bid) and the best sell order (Ask). A hole with no limit orders in it at all.

The relationship is rigid. The more participants quote an instrument, the tighter the queue and the smaller the hole.

The spread between the first and last row runs up to 300x. On a liquid pair turning over around 300 million dollars a day the spread sits near 0.011%, while on a fresh low-liquidity listing four percent counts as normal.

Here's the first non-obvious conclusion. For a scalper trading at market, a tight spread is a blessing. For a scalper collecting the spread with limit orders, a tight spread means no income. The same liquidity works both ways.

The second thing a quoting firm provides is shock absorption. When a large market order lands, someone has to take it. If an algorithm is taking it on both sides of the book, price shifts two ticks and comes back. If there's nobody to take it, price jumps across a dozen levels without a single trade.

That's what it looks like when the book comes apart. Orders were pulled, liquidity evaporated, and the first mid-sized order draws a three percent spike. On the chart it's a powerful move. In reality less money went through it than in an ordinary five-minute candle.

I check this simply. If ATR expanded sharply while the tape is sluggish, the move wasn't bought, it was dropped into a vacuum. Flushes like that come back 80% of the time, but a stop inside them fills with slippage many times larger than what you calculated.

If reading depth is still hard for you, start with the article "What the Order Book Is and How to Read It Properly".

Crypto MMs vs the stock market

On the NYSE a designated market maker (DMM) has written obligations. He holds a quote no wider than a set width, participates in the open and the close, and sits in the book for an agreed percentage of the time. Fail, and you lose the status and the perks.

In crypto there's no regulator over the order book. There's a contract between two private companies whose contents nobody publishes.

That last row explains half of the strange behavior you see in alts.

The standard arrangement with a project is called "loan plus call option." Shortly before the token launch the project hands the quoting firm 1% to 5% of circulating supply. It isn't sold, it's lent, which is why the deal doesn't count as a sale and doesn't require disclosure. The term is usually 12-24 months. By the end of it the firm either returns the same number of tokens or buys them out at a pre-agreed strike, set 25-100% above the launch price.

Let's work out what follows from that. If price has gone well below the strike, exercising the option makes no sense. The rational strategy for the borrower is to sell someone else's tokens at market, buy them back cheaper, and return the same quantity. If price approaches the strike from below, that's unpleasant too: the firm hedges in advance and sells into the market exactly where everyone is waiting for a breakout.

These contracts almost never surface publicly. A rare exception was the leaked agreement on the MOVE token in spring 2025, which made it clear that the persistent selling pressure came from the structure of the deal, not from "market sentiment."

There's one takeaway for a trader. If supply keeps appearing at the same level on a fresh listing and kills every rally, that isn't necessarily a large investor distributing. It may be an option hedge, and it isn't going anywhere until the term ends.

If the theory still sounds abstract, watch the free lesson from our trading-from-scratch course, where a live order book shows how limit orders build liquidity and where entry points come from.

How market makers earn

Three income sources, and only the first one is obvious.

The spread

The basic mechanics are described above, but there's a part below the surface that sometimes makes the algorithm simply leave the book.

Arithmetic with numbers. An algorithm quotes a pair with a 0.02% spread, turns over 500 million dollars in a month, and captures half the spread on each side of a trade. Gross revenue is around 50 thousand dollars. That sounds modest until you remember the capital is barely tied up: the position revolves around zero rather than sitting in the asset.

Then the real work starts, inventory management. If the market sells at market for fifteen minutes straight, the algorithm buys all of it on its own bids and finds itself several million long in a falling asset.

Hence the two defensive mechanisms you see in the book every day without knowing that's what they are.

Quote skew. Having accumulated a long, the algorithm shifts both sides down: bids further from price, asks closer. The position starts unloading itself. From the outside it looks like "a seller pressing from above," though nobody is pressing anywhere, it's just rebalancing.

Spread widening. When uncertainty rises (a data release, a liquidation cascade, an outage on another venue), the algorithm spreads its quotes wider. That's the price of risk. In those moments the coin looks normal by volume, but a market fill suddenly costs you half a percent.

The second risk is called adverse selection, and it can't be removed. When your limit order gets filled, it was filled by someone who at that moment considered your price good for himself. Which means on average you're always slightly on the losing side. The spread is the compensation for that asymmetry.

By the way, a retail scalper collecting the spread in illiquid names runs exactly the same mechanics. The only difference is scale, and the fact that a private trader has no obligation to stand in the book when things get scary. The practical capital limit here is 300-500 dollars per coin: above that you become the density level other people close their trades against.

Exchange rebates

The second income source has nothing to do with price movement at all.

Exchanges use a maker-taker model. Whoever posts a limit order and adds liquidity to the book pays less or gets paid. Whoever takes liquidity at market pays more. The difference between the two rates is the price the venue is willing to pay for a deep book.

On Binance futures the base rates for VIP 0 are around 0.02% for the maker and 0.05% for the taker. At VIP 9 the maker fee drops to zero with a taker fee of 0.017%. There's a separate liquidity provider program with negative maker fees, and for new listings the exchange periodically runs promos with a 0.005% rebate on maker trades.

Let's see what that means in money.

Twenty-five thousand dollars earned without a single correct call on direction. Purely for having orders sitting in the book.

Two observations follow that are useful in practice.

First. Yesterday the book on a coin was full of holes, today there's an even, deep queue on both sides? Most likely the exchange switched on a liquidity provider promo. The spread will collapse and the spread-collecting strategy will stop working there. Not because you're doing something wrong.

Second. Rebates explain why some orders in the book aren't interested in getting a good fill at all. What matters to the algorithm is volume and status, not two ticks of profit. Price doesn't bounce off orders like that.

Informational advantage

The most underrated income source and at the same time the most mythologized.

Let me start with what doesn't exist. A market maker does not see your specific stop. A stop order sits on the exchange's servers until it triggers and isn't displayed in the book, and if it's a stop set in your terminal it's purely local.

Now here's what does exist.

• Their own flow. A large firm quotes one pair on eight venues plus runs OTC deals. It's the first to see that someone is methodically lifting offers across the entire market at once. An ordinary trader will see that on the chart about twenty minutes later
• Level clustering. Stops aren't scattered randomly, they sit behind round numbers, behind the extremes of recent bars, behind the edges of a range. You don't need to know any specific stop, it's enough to understand where they pile up
• Liquidation map. On futures there's no guessing at all: leverage, entry price and open interest give you the calculated forced-close price. A ready-made map of fuel for the move, computed publicly
• Speed. While your order travels over a normal API, the quote has already changed four times

The picture for retail is unpleasant. When price approaches a cluster of stops and the density level in the way is suddenly pulled, a move into that zone becomes very likely. Not because "the whale is out to get you," but because that's where the liquidity is, and you can only fill large size where there's someone on the other side.

Telling a quoting algorithm apart from an ordinary large buyer is covered in the article "How Large Players Move the Crypto Market": their traces look similar, their motives are different.

How to see what MMs do in the order book

Now the practical part. The algorithm leaves three types of trace, and all three are visible in an ordinary order book if you know where to look.

An important caveat before we start. None of these traces proves that what you're looking at is a market maker. Anyone with money can spoof. But the mechanics of the price reaction don't change because of that, and what you trade is the mechanics, not the identity of the counterparty.

Spoofing

Spoofing is posting a large limit order with no intention of filling it. The goal is to scare the market and push price in the desired direction, then pull the order before price reaches it.

The tell is simple: an order that appears or disappears instantly as price approaches is an attempted manipulation.

How to tell a fake from a real density level in the book:

• Lifetime. A real participant posts size and waits. A spoofer flickers. A timer showing how long an order has been sitting at a level answers most of the question faster than any other filter
• Behavior as price approaches. A real density level meets price and starts getting eaten into. A fake one dissolves a few ticks before contact
• Round number. A genuine large player wants to get size done and picks psychologically obvious levels like 1.500 or 2.000
• The trace in the cluster. The main after-the-fact check. Price went through the level and there was no volume in the cluster there? Then there was nothing to eat, the order was painted on

A numerical example from practice. A top-50 coin, price 2.145. At 2.160 a sell order for 1.8 million dollars appears, while the average order in the book is around forty thousand. It looks like an impassable density level.

I check the timer: 4 seconds. I open the density map, the level isn't there, nothing has been sitting at 2.160 for the past hour. Price reaches 2.157 and the order disappears entirely. The cluster at that level shows 60 thousand of volume, ordinary background.

Conclusion. There was no wall, there was a set decorator. Someone wanted you not to buy.

A full breakdown of the technique, including layered spoofing and layering, is in the article "Spoofing in Trading: How to Spot Fake Orders".

Iceberg orders

An iceberg works in exactly the opposite way. A spoofer shows size that isn't there. An iceberg hides size that is.

The mechanics work like this. A small visible portion of the order is displayed in the book, say 5,000 coins out of 200,000. As soon as the visible portion is filled, the exchange automatically posts the next slice. The book shows 5,000 the whole time while hundreds of thousands go through the level.

Signs that let you identify an iceberg:

• The order at the level doesn't shrink even though trades at that price are printing on the tape one after another
• Cluster volume at the level is many times larger than what's visible in the book. You see 5,000, the cluster at that price shows 180,000, conversation over
• Price hits the level five, ten, fifteen times and pulls back a couple of ticks each time
• Once the order is pulled, price goes through the level instantly, because there's nobody left to defend it

Why do it this way. A large participant needs to get done size the book can't digest in one go. Show the whole size at once and the market sees it, and price never even reaches that level.

An iceberg, unlike spoofing, is real money at a real level. Price bounces off it. Catch a working iceberg on the bid and you have a level that's being defended and an obvious stop placement right underneath it.

Absorption

Absorption is a situation where aggressive volume on the tape is large but price doesn't move at all.

The difference from an iceberg is subtle but matters. An iceberg is an order construct, a technical thing. Absorption is a result you observe on the chart. The market hits and the level holds.

How to read it on your tools:

• The tape shows a stream of large buys, speed has picked up noticeably
• Cluster volume at the level is huge, POC is highlighted right there
• Price has covered one or two ticks in that time, or nothing at all
• Delta is strongly positive while the candle stands still

That means someone is taking all that aggression with limit orders and not letting price through.

Absorption is the strongest reversal signal microstructure gives you. The logic is that the buyer spent his ammunition and got no move for it. When the flow of aggression ends, and it will end, there'll be nobody left holding price up, and it goes in the direction of whoever was absorbing.

The reverse situation reads the same way. Cluster volume is low while price flies, which means the move is running across an empty book, and the reversal can be just as sharp.

Summary table: what you see and what you do.

How to use this in trading

Four working scenarios. Each is built on being able to tell real liquidity from painted liquidity.

Scenario 1. A bounce off a confirmed density level.

Scalping classic. Find a density level in the book, verify it's genuine, enter with a limit in front of it, and hide the stop behind it.

The check before entry:

• Order lifetime over 30-60 seconds, better still several minutes
• The level is round or matches a previous extreme
• The density map shows the order has been sitting there a long time and across the whole market
• On the tape, aggression against my side is fading rather than building
• There's somewhere to put a stop behind the density level, meaning the book isn't empty

If even two points fail, there's no trade.

Scenario 2. Front-running a density level.

Post a limit buy one tick in front of a large density level on the bid, become first in the queue, and let the density level work as a shield behind you. Post the sell in front of the opposing density level on the ask.

The golden rule that comes with it. If the density level you're leaning on gets pulled and doesn't come back within 15-20 seconds, exit at market immediately. No waiting, no hoping, no averaging down.

Scenario 3. Don't enter against absorption.

The cheapest advice in this article, because it saves money rather than making it. You've seen a level break and you want to join the move. Look at the cluster at the breakout level. If volume there is huge while price barely moved, there's no breakout, someone is taking it.

Scenario 4. Trading a book that has come apart.

When the quoting firm leaves, price flies into a vacuum on laughable volume. Moves like that revert more often than they continue. But you can only trade against them on two conditions. The move happened with no rise in cluster volume, and orders on the opposing side have already appeared in the book. Without the second one it's catching a falling knife.

A case study. Everything done right and the trade still loses.

BTC/USDT five-minute chart, price 67,180, approaching 67,200 from below. At 67,200 there's an order for 2.4 BTC in the book, the timer shows 20 minutes, the density map flags it as long-lived.

The order is genuine by every sign. I build a short off it. Limit entry at 67,195.

I size the stop off ATR(14), which equals 95, with a 1.5 multiplier, giving 143 points, i.e. 67,338. I look higher: at 67,350 there's a second density level, so I put the stop behind it at 67,360. That's 165 points. With $50 of risk, size comes out at 0.30 BTC.

Then the tape accelerates, about 5 BTC of aggressive buying goes through in a minute, the order gets eaten into, the density level is pulled, price runs up. The stop triggers, minus $50.

There's no mistake in that trade. The order was real, it simply got eaten. In my experience a confirmed density level works out roughly two times out of three, and the rest are eaten levels like this one. The mistake would have been moving the stop or re-entering the same idea on emotion.

The mirror version of the same case. The timer shows 3 seconds, the density map doesn't know the level, so I don't take the short at all. And I don't put a stop there either, because a stop behind a fake order is a stop sitting in a vacuum.

Secret Terminal tools for this job.

• Order lifetime timer. Next to large size it shows how long it has been sitting at the level. Sitting a long time means the interest is real. Flashed and vanished means spoofing
• Multi-exchange density map. Aggregates large limit orders from all connected venues into one table with a size filter. A one-sided density level is almost always decoration
• Clusters with POC highlighting. They show at which prices inside a candle volume actually went through. That's your iceberg and absorption detector, and without it you can't tell the two apart
• Tape with a volume filter. Cuts out the small stuff, leaving only the trades that move price. Absorption can't be read on a dirty tape
• Spread line and preset auto-tuning. The width of the Bid-Ask gap is visible immediately, and the "large volume" thresholds adjust to a specific coin's volatility in five seconds

Three mistakes I see most often.

First, treating any large order as a level. Half the large orders in alt order books live less than ten seconds.

Second, putting a stop right behind a visible density level without checking the timer. The order gets pulled, the stop ends up in a vacuum, and the fill goes half a percent away.

Third, and most expensive, explaining every loss as manipulation. Price went for your stop not because someone was personally hunting it. It went where the liquidity was, and your stop was sitting in the same pile as a thousand others. The conclusion isn't "the market is against me," it's "don't put your stop where everyone else does."

If you want to work through the order book plus clusters plus tape combination step by step, watch the free lesson from the full course "Trading From Scratch," which is about exactly how professionals read the market through those three tools.

FAQ

• What is a market maker in simple terms?

  A participant who holds buy and sell orders at the same time and earns on the difference between them rather than on guessing direction. His profit per trade is a fraction of a percent, and the number of such trades runs into the thousands per hour. The position, meanwhile, almost always returns to flat.

• Market maker: what is it on an exchange and how does it differ from a large player?

  A large player came to build or unload a position, he has a direction. A market maker stands in the book on both sides and has no direction, his goal is volume and spread. Their traces in the book look similar, so you have to tell them apart by behavior: a quoting firm reposts its orders after they fill, while a large player leaves once he's done.

• Do market makers in crypto manipulate price?

  Some of what they do falls under the definition of manipulation, spoofing for instance, and exchange rules prohibit it. But most of the moves retail calls manipulation have a simpler explanation: inventory rebalancing, a hedge against an option agreement with a project, or the ordinary search for liquidity to fill large size.

• Does a market maker see my stop-losses?

  A specific stop order isn't displayed in the book until it triggers, so it can't be seen. What is perfectly visible are the zones where stops pile up: round numbers, the extremes of recent bars, the edges of a range. On futures you can add the calculated liquidation map to that, which anyone can compute publicly from open interest.

• How do I tell a real density level from spoofing?

  By order lifetime and by the trace in the cluster. A real density level sits for minutes and meets price, spoofing appears and disappears within seconds as price approaches. If price went through the level and there was no volume in the cluster there, the order was fake.

• Why is the spread 0.01% on some coins and 4% on others?

  Because of how many participants quote the instrument. On BTC hundreds of algorithms close any gap instantly, while on a fresh low-liquidity listing there may be no quoting firms at all. For spread collecting as a strategy the second case is the useful one, for trading at market it's the first.

• What happens when a market maker leaves a coin?

  The spread widens several times over, book depth drops, and an ordinary-sized order starts moving price by whole percentage points. From the outside it looks like a strong move, though very little money went through it. The tell for this situation is a sharp expansion in ATR while the tape is sluggish.

Briefly, if you take one idea away. Spoofing shows size that isn't there. An iceberg hides size that is. Absorption shows that size was taken and let nowhere. Checking any large order comes down to three questions: how long has it been sitting at the level, is it there on other exchanges, and did volume go through it in the cluster.

See what market makers do. Secret Terminal shows the lifetime timer on every large order, builds a density map across all connected exchanges at once, highlights POC inside clusters, and filters the small stuff off the tape with one scroll of the wheel. Spoofing gets separated from a real density level before you enter the trade, not after your stop fills.