Crypto markets and data — course checkpoint

You came in knowing how markets work. You leave able to read a market that broke most of those assumptions — and to say which broke and what replaced them.

Unit 1 — the clock

Crypto trades 8,760 hours a year against roughly 1,638 for a US equity market. There is no closing auction, so "the daily close" is a convention — two cut-offs produce two return series from identical tape. Annualisation uses √365, not √252; the wrong one is a silent 20% error.

There are no market-wide circuit breakers, because there is no single market to halt. Weekends are the thinnest part of the week — fiat rails, stablecoin issuance and the regulated futures venue all shut — so a $5m order that moves price 0.6% midweek moves it about 1.6% on a Saturday, the 2.67× the depth ratio implies. Depth is a denominator.

Volatility, factually: daily σ of 3.0% annualises to about 57%, against roughly 16% for a broad equity index and 7% for EUR/USD. Bitcoin has recorded several drawdowns exceeding 70%.

Unit 2 — venues and constructed prices

A CEX is a matching engine over a database — fast, deep, a claim on the exchange rather than the asset. A DEX is a formula over inventory: x × y = k, where buying 1 ETH from a 100 ETH / 300,000 USDC pool costs 1% of impact and buying 10 costs 11.1%.

There is no consolidated tape and no NBBO, so prices differ by venue, quote currency and jurisdiction. Arbitrage cannot close the gaps: 8.3 bps of spread against 10 bps of round-trip taker fees is a losing trade, and capture requires pre-funded balances at venues carrying real counterparty risk.

So every "price of bitcoin" is a recipe — venue eligibility, window, weighting, outlier rules. Reported volume is an unverified claim, cross-checked against depth and price impact; and any statistic built on today's listed tokens has already deleted every failure.

Unit 3 — the perpetual and its tether

A perpetual future has no expiry, hence no convergence, hence a replacement anchor: the funding rate. Positive, longs pay shorts; negative, shorts pay longs; paid between traders on notional, only if you hold at the snapshot.

The arithmetic to carry: +0.01% per 8 hours = 0.03% a day = 10.95% a year on notional, and 54.75% a year against capital at five times leverage. At a stressed +0.10% per interval, 109.5% on notional.

Funding says the perp traded above or below its index: evidence about crowding and fragility, read with open interest, silent on timing and direction. Annualised basis = (futures/index − 1) × (365/days) — 2.0% over 90 days is 8.1% a year.

And the mechanism that ends positions: a 10× long at $60,000 with a 0.5% maintenance rate liquidates near $54,271 — a 9.5% move. Forced selling drives price into the next cluster of levels, with no halt, in a thin book; auto-deleveraging can then close profitable opposite positions when the insurance fund falls short.

Unit 4 — the ledger, and what it cannot see

A public chain is a public settlement dataset, which is genuinely new. Active addresses counts addresses, not people — exchange clusters, bots, change addresses and Layer 2 all sit inside it. Exchange flows rest on heuristic labels nobody publishes, producing identical charts for selling, custody migration and mislabelling. Realised cap values each coin at its last on-chain move — MVRV 1.5 in our three-coin economy — a basis that self-transfers reset and exchange trading freezes.

Above all: the chain sees the doorway, not the room.

The risk statement, stated plainly

Perpetual futures and margin products in crypto are leveraged instruments, and liquidation is a routine daily occurrence here, not an edge case. Losses can be total — isolated-margin liquidation loses the entire margin posted, cross-margin can consume the whole account balance, and gaps have historically left negative balances. Auto-deleveraging can close a profitable position without consent. Venues can and do suspend withdrawals, and assets on an exchange are a claim on that company.

Funding rates, open interest, exchange flows and on-chain metrics are observations. Nothing in this course is a signal, a recommendation to trade crypto, or a suggestion to use leverage, and none of it is investment advice.

What you can now do

Interrogate a crypto price for its venue, cut-off and day count. Explain why two datasets disagree. Price a trade's impact against a pool. Convert a funding rate into an annual cost on notional and on capital. Compute a liquidation level. Read an on-chain metric with the definition that limits it.

Before you sit it

Each of these is a minute at your desk. Any one that is not names the lesson to reopen first.

• Say what a market with no halts does to risk that a halted one does not — What happens to risk when nothing can be halted?
• Say why one coin has a different price on every venue at once — Why does the same coin have a different price on every exchange?
• Say what a positive funding rate means and which side is paying — What is a funding rate, and why does it exist?
• Say what an exchange inflow actually measures, and what it does not — What do exchange inflows and outflows actually measure?

Try it now

1. Pick one crypto asset and write five sentences: its venue landscape, your data's cut-off convention, its trailing annualised volatility using √365, its current funding and what that describes, and one on-chain metric with its limitation.
2. Verify every figure. For bitcoin, price is the year of daily closes below, the cut-off convention is in the table under it, and the volatility is the daily σ measured in the volatility lesson of this course, annualised with √365. Everything a price feed does not carry comes from the venues' own public specification and funding pages. Mark each sentence with the source that confirmed it.

3. Then say the sentence this course was built around: in crypto, the price is a construction, the perpetual is a funding stream, and the ledger sees the doorway rather than the room.

Checkpoint quiz next. Nothing in this course was a recommendation to buy, sell, hold or trade any crypto asset — you have learned to read a 24/7, fragmented, heavily leveraged market accurately, which is a literacy skill and not a signal.