Crypto market making

The "you can actually start here" venue: open APIs, free L2/L3 data, 24/7 books, no prime broker. The canonical maths transplants directly; you pay for it in counterparty risk, wash volume and books that vanish in a gap.

See it move

How does market making work in crypto?

Identically to anywhere else: you post a resting bid and ask on a venue's order book, earn the spread when flow trades against both sides, and manage the inventory and adverse selection you accumulate. The maths is the canonical Avellaneda–Stoikov quoting problem. What is different is the venue, not the model, which is the whole point of the sandbox above, where the same engine re-skins to a 24/7 crypto perpetual.

Intuition first: a crypto market maker on a centralised exchange (CEX) is doing exactly what a currency-bureau stall does. It posts "I'll buy at 99, sell at 101", earns the 2 per round trip, and manages whatever position it is left holding. The reservation-price logic of the A–S model applies unchanged: you skew your quotes away from your inventory so the market flattens you. When you are long, you shade both quotes down so your ask is more likely to lift.

The canonical building blocks all transplant: the limit order book is a CEX matching engine; price-time priority governs your queue position; order-flow imbalance and the microprice read the book exactly as they would on an equity; PIN/VPIN gauges flow toxicity. The common mistake is reinventing these badly; this page is the canonical transplant. What is genuinely new to crypto is the funding rate on perpetual futures (a periodic payment between longs and shorts that tethers the perp to spot, a basis you can earn or pay, and a market-making cost you must account for) and the AMM as an alternative to the order book, below.

CEX order books vs DEX / AMMs: which microstructure?

Crypto has two liquidity-provision paradigms. On centralised exchanges (CEXs) you market-make a classic central limit order book, and the canonical maths applies directly. On decentralised exchanges (DEXs) with automated market makers (AMMs), liquidity is a pool priced by a formula, not a quote: a fundamentally different game with its own risk, impermanent loss.

CEX CLOB market making is the direct transplant: a matching engine with price-time priority, a depth ladder, a spread, signed flow and adverse selection, where every concept in the microstructure and market-making guides applies verbatim. This is where the sandbox's crypto preset lives, and where most of this page focuses.

AMM / DEX liquidity provision is not quoting. You deposit into a pool and the pool prices trades along a deterministic curve, with the constant-product invariant $x\cdot y = k$ being the canonical example. You earn fees on volume but bear impermanent loss: the pool automatically sells the appreciating asset and buys the depreciating one, so a price move leaves you worse off than simply holding. It is structurally closer to a short-volatility position than to a quoting strategy: a different risk profile entirely.

The practical 2026 picture: serious low-latency market making happens on CEX order books (the AMM curve cannot compete on tight, deep, fast quoting), while AMMs dominate on-chain spot where decentralisation matters more than execution quality. For an HFT/quant reader, the CEX order book is the relevant arena and the one this page treats.

Why is crypto the most accessible arena for an independent?

Because the institutional barriers that wall off lit equities are absent or cheap. No prime broker, no exchange membership, no colocation requirement to start, gettable (often free) L2 data over a WebSocket, 24/7 hours, and a self-hosted bot that runs from your own machine. The capital-and-connectivity bar that takes years to clear in equities is days in crypto.

No prime broker, no membership. You open an exchange account, generate API keys, and connect: there is no sponsoring broker, no clearing relationship and no membership gate. The access barrier that defines equities & futures simply is not there.

No colocation barrier to start. Lit-equity market making is a colocation-and-FPGA arms race. Crypto venues do offer colocation and the speed game exists at the top, but the entry bar is far lower: many profitable crypto niches run on software latency, not FPGA. You compete on fair-value quality and discipline, not only nanoseconds.

Gettable data. L2 (and sometimes L3) order-book data streams over public exchange WebSockets, often free and in real time: the single biggest practical difference from the expensive, licensed feeds behind lit equities. 24/7 and fragmented: the market never closes, and the same coin trades on many exchanges with real price dispersion, leaving room for latency arbitrage and cross-exchange statistical arbitrage at a speed bar a small team can clear. This is the practical foundation of the going-independent-in-2026 thesis; the fixed-cost-versus-margin maths is in the economics of HFT.

What are the real risks: counterparty, wash trading, thin books?

Crypto's open access comes with distinctive, first-order risks. They are not footnotes; they are the price you pay for the absent barriers.

Counterparty / venue risk is the dominant risk. On a CEX your funds sit with the exchange; the 2022–2023 collapses (FTX and others) wiped out balances held on-venue. A crypto market maker is structurally a creditor of every exchange it quotes on, so diversify venues, sweep funds off-exchange, and price the risk. There is no equivalent in a cleared, regulated equity market.

Wash trading and fake depth. A meaningful fraction of reported volume on some venues is wash-traded, and displayed depth can be self-dealt. Your fair-value and toxicity estimates are only as good as the tape, and the crypto tape is dirtier than the equity tape, so validate venues and weight by real flow.

Thin, gappy books. Outside the top pairs, books are thin and gap on news, exactly when your inventory matters most. Adverse selection and fat tails are worse in crypto than in liquid equities, not better. The same A–S model that is calm on a deep book blows up faster on a thin one; crank the volatility slider in the sandbox to feel it. And the regulatory patchwork (no Reg NMS or MiFID II equivalent, rules varying by jurisdiction and venue, and thinner market-abuse enforcement) is itself a risk.

The honest summary: crypto is the accessible frontier, not the safe one. The barriers that are absent (prime broker, colo, licensed data) are replaced by risks that are present (counterparty, wash trading, thin books). A clear-eyed small team can still earn real spread here, and that is the genuine opportunity, but only by pricing these risks, not ignoring them.

Which strategies apply in crypto, and which are alive in 2026?

Market making, cross-exchange latency arbitrage, statistical arbitrage and funding-rate basis trades all transplant to crypto, and the field is shallower than equities. Pure speed at the very top is contested by professional firms; but at the small-team scale, market making on second-tier pairs, cross-venue arb and funding-basis capture remain genuinely accessible openings.

Market making is the flagship application and the one the sandbox demonstrates: alive on CEX order books, especially on pairs below the most-contested majors where the competition is shallower. Latency arbitrage exploits the real, persistent price dispersion of the same coin across many exchanges; the speed bar to pick off a stale cross-venue quote is lower than in fragmented equities. Statistical arbitrage (cross-exchange, the same asset on two venues, and cross-pair, cointegrated coins, mean-reversion) keeps corners in crypto that the classic pairs trade lost in liquid equities. And funding-rate / perp-basis trades harvest the basis the perpetual funding rate creates between perp and spot: a largely market-neutral, microstructure-adjacent strategy native to crypto.

The 2026 honesty: the very top of crypto market making is now contested by the same pro firms that dominate equities. The opening for a small team is one rung down (second-tier pairs, smaller venues, the niches the giants ignore) exactly as is HFT still profitable in 2026 describes for the open venues.

Worked example

A simplified crypto market-making round on a synthetic perpetual, mid = 100.00 (USD), variable tick, taker fee 4 bps, no rebate, 24/7; illustrative, as of a 2026 worked snapshot. You quote bid 99.90 / ask 100.10, a 20-bps spread, wider than the equities one-tick example because the book is thinner and adverse selection larger. A taker hits your bid: you buy 1 unit at 99.90, inventory $q = +1$. You pay no fee as the maker; the taker pays the 4 bps.

Balanced case. A buyer lifts your ask: you sell 1 at 100.10, back to $q = 0$. Round-trip gross is $(100.10 - 99.90) = +0.20$ (20 bps), earned as the maker with no rebate to add. Compare the equities example, where a rebate sweetened each fill: in crypto you typically earn the spread alone, which is exactly why you quote wider to compensate.

24/7 inventory case. The second fill does not come, and because there is no session close to flatten into, you carry the long across the night while the mid drifts to 99.50. Your unrealised mark is $(99.50 - 99.90) = -0.40$ on inventory, double the half-spread you earned. The absence of a close is a real cost: equities flatten daily; crypto inventory random-walks around the clock unless you skew it flat.

Thin-book / adverse-selection case. The taker who hit your bid was informed; the mid gaps to 99.20 right after (thin books gap further than deep ones). You bought at 99.90 something now worth 99.20, a $-0.70$ mark before you can react. This is why crypto quotes are wider: the break-even spread must cover a larger expected adverse-selection cost than in liquid equities. The numbers are illustrative and synthetic; real crypto spreads, fees, funding rates, depth and gap behaviour vary by venue and pair, so check the exchange spec, as of 2026. Educational only, not investment advice; no P&L is promised.