Crypto Market Structure

2.6.1 One company, four jobs: the centralised exchange

Most crypto trading happens in an app that looks like a brokerage. Behind it, one company does the work that four separately regulated firms do in the markets of Session 2.1.

In Session 2.1 you followed an order through a broker, an exchange, a clearing house and a custodian: four firms, four rulebooks, each checking the others. A centralised exchange (CEX) such as Binance, Coinbase or OKX does all four jobs itself: it takes your order (broker), matches it in its own limit order book (venue), sets your margin and closes your position when it runs out (clearing house), and holds every coin and dollar you deposit (custodian). Many also lend, make markets or issue a token. IOSCO, the global body of securities regulators, lists "conflicts of interest arising from vertical integration of activities and functions" and "custody and client asset protection" among the six key areas of its 2023 crypto recommendations.

Once you deposit, the coins leave your control: your balance is a row in the exchange's database, and you cannot see what stands behind it. In the US, for example, SEC Rule 15c3-3 makes a broker-dealer keep customers' securities and cash segregated from its own, and UK and EU client-asset rules do the same; a crypto exchange's promise of segregation is only as good as whoever checks it.

2.6.2 When the venue is the risk: custody, reserves & FTX

Custody risk and counterparty risk are two faces of one question: if this firm fails tomorrow, what do I get back? In November 2022, FTX's customers found out.

Custody risk is the risk of losing assets because of how, and by whom, they are held: stolen from a hot wallet, locked behind a lost key, or quietly used by the custodian. Counterparty risk is the risk that whoever owes you — a venue, an issuer, a borrower — does not pay. A clearing house caps counterparty risk at about one day's price move (Session 2.4); on a CEX both risks sit in one company, uncapped. Custody can also fail on its own: the FBI attributed the theft of about $1.5 billion of crypto from the exchange Bybit, on or about 21 February 2025, to North Korean hackers. Bybit stayed solvent and, with emergency loans, restored full backing of client assets within about three days (CNBC, 24 February 2025): the loss fell on the firm, not its customers.

Nine days in November

Sources: CoinDesk (2 and 9 Nov 2022), CNBC (9 Nov 2022), and the FTX bankruptcy filings.

What failed was the separation of roles. The SEC alleged in December 2022 that FTX gave Alameda "a virtually unlimited 'line of credit' funded by the platform's customers" and exempted it from key risk controls — the clearing job waived for an insider, paid for by the custody job. John J. Ray III, the new chief executive, told the bankruptcy court he had never seen "such a complete failure of corporate controls". A jury convicted Bankman-Fried on all seven counts on 2 November 2023; he was sentenced to 25 years on 28 March 2024, and prosecutors said he had taken more than $8 billion of customers' money.

Under the plan confirmed on 7 October 2024 (CNBC), customers with claims up to $50,000 — most creditors by number — received about 120% of their allowed claims from 18 February 2025, and larger customer claims had received 105% by the fifth distribution on 31 July 2026 (FTX Recovery Trust). But the claims were fixed in dollars at 11 November 2022 prices, when bitcoin traded near $17,000: a customer who held bitcoin got dollars at the crash price, more than two years later. Counterparty loss is not only whether you are paid, but in what and when.

Proof of reserves, and what it cannot prove

After FTX, exchanges began publishing proof of reserves: their wallet addresses (the assets) plus a Merkle tree of customer balances (the liabilities) in which each customer can find their own. As Buterin (2022) sets out, it is a snapshot, not a stream; coins can be borrowed for the snapshot (exchanges could "shuttle collateral" between each other); other debts never enter the tree; and fiat still needs an auditor. In December 2022 the auditor Mazars paused its proof-of-reserves work for all crypto clients, citing how the public understood the reports.

2.6.3 Decentralised exchanges & the constant-product AMM

A decentralised exchange replaces the company with a program. The most influential design has no order book at all — just two piles of tokens and one line of algebra.

A decentralised exchange (DEX) is a set of smart contracts — programs on a public blockchain — that you trade with straight from your own wallet. Most spot DEX volume runs through automated market makers (AMMs): instead of matching buyers with sellers, an AMM is a pool of two assets, funded by liquidity providers (LPs), that always trades against you at a formula price. Uniswap v2's pool (Adams, Zinsmeister & Robinson 2020) holds $x$ tokens and $y$ dollars of a stablecoin (§2.6.5) and accepts any trade that does not reduce the product of the reserves:

Sell $\Delta x$ tokens and the pool pays out the $\Delta y$ that keeps the product at $k$. The fee $\phi$ — 0.30% on Uniswap v2, split 0.25% to LPs and 0.05% to the protocol since the "UNIfication" governance proposal was executed on 28 December 2025 (Uniswap documentation, checked 2026) — comes off what you pay in: only $(1-\phi)\,\Delta x$ counts toward $k$, but all of it stays in the pool, so $k$ grows with every trade:

Write the trade's size as a share of the pool, $s = \Delta x / x$. Its total cost against the pre-trade price — fee plus market impact — is

That is the rule of thumb to keep: an AMM trade costs the fee plus roughly its own size as a share of the pool. The lab's pool holds 1,000 tokens and $2,000,000, a price of $2,000:

Depth is everything: the same $20,000 sale into a pool 10 times deeper costs 0.40% instead of 1.28%. The swap itself, from the lab:

def swap(amount_in, reserve_in, reserve_out, fee=FEE):
    eff = amount_in * (1.0 - fee)                      # only this part counts toward k
    out = reserve_out * eff / (reserve_in + eff)       # keeps (reserve_in + eff)·(reserve_out − out) = k
    return out, reserve_in + amount_in, reserve_out - out   # the whole input stays: k grows

for s in SIZES:                                  # 0.1%, 1%, 5%, 10%, 25% of the token reserve
    dx = s * X0
    dy, x1, y1 = swap(dx, X0, Y0)
    exec_px = dy / dx
    cost = 1.0 - exec_px / mid0
    assert abs(cost - sell_cost_closed_form(s)) < 1e-12

2.6.4 Liquidity providers & impermanent loss

Someone has to fill the pool. Liquidity providers earn the fees — and pay for them by taking the other side of every arbitrageur.

When the price elsewhere moves, arbitrageurs trade against the stale pool until it matches (Angeris et al. show why a constant-product pool must track the outside price) — which means the LP sells the token as it rises and buys it as it falls. Without fees the pool's state depends only on the price $P$: $x = \sqrt{k/P}$ and $y = \sqrt{kP}$, so the LP holds $xP + y = 2\sqrt{kP}$. If the price moves by the ratio $\theta = P_\tau/P_0$ over $\tau$ years, compare that with holding the deposit:

That gap is impermanent loss — "impermanent" because it vanishes if the price returns, permanent once you withdraw. It is symmetric in the ratio: a halving and a doubling both cost −5.7% against holding, ±20% costs −0.6% or −0.4%, and a fourfold rise −20%. In the lab an arbitrageur makes 720 hourly trades against a pool on each of 40 simulated paths, using only the swap function, and the LP's result matches the formula to within 1e-15 on every path: only the final price matters.

How much does it cost on average? If the price has no drift, holding breaks even in expectation, but $\sqrt{\theta}$ is concave, so the LP does not: for a lognormal price $\E[\sqrt{\theta}] = e^{-\sigma^2 \tau/8}$, with $\tau$ in years, and

At a crypto-like 80% volatility over one year, theory says −7.69%; 200,000 simulated years give −7.72% with a standard error of 0.035%. So fees must earn about 3.1% a year just to match holding at 50% volatility, 7.7% at 80% and 11.8% at 100% ($1 - e^{-\sigma^2/8}$ for one year).

2.6.5 Stablecoins: four designs, four ways to break

Most crypto prices are quoted in a dollar that is not a dollar. How it is built decides how it breaks.

A stablecoin is a token designed to hold a fixed value, usually $1, by being a claim on something: an issuer's reserves, a pile of crypto collateral, a mechanism, or a hedged crypto position.

Read each design as something you already know. A fiat-backed coin is a narrow bank without a banking licence: it carries the credit risk of its reserves and issuer, and few firms can redeem directly — Tether let an average of six agents a month redeem for cash (Ma, Zeng & Zhang 2025). A crypto-collateralised coin is a margin loan with the liquidation rules in code. An algorithmic coin works while people believe it works; Terra's run was visible on-chain, which sped it up, and the wealthier, more sophisticated investors ran first and lost less (Liu, Makarov & Schoar 2023). A synthetic dollar is §2.6.7's basis trade in a wrapper, with that trade's risks.

2.6.6 Perpetual futures & the funding rate

The most traded instrument in crypto is a futures contract that never expires. With no expiry there is no convergence — so the venue builds one in, as a payment between longs and shorts every few hours.

In Session 2.4 a futures contract converges to spot because at expiry it is the asset. A perpetual future (a "perp") has no expiry. The idea is older than crypto — Shiller (1993) proposed perpetual contracts on hard-to-trade indices — but it took off when BitMEX introduced the first crypto perpetual in 2016 (He et al.), long offering up to 100× leverage (250× since April 2024, as reported by The Block). By one estimate, perps traded over $60 trillion in 2024 (Chitra 2025).

What replaces convergence is the funding rate. With perp price $F_t$, spot $S_t$ and premium $\pi_t = F_t/S_t - 1$ (π for premium), at each funding time every long pays every short the funding rate times the position's notional — or receives it when the rate is negative; the money moves between traders, not to the venue (Ackerer, Hugonnier & Jermann 2026 derive the resulting no-arbitrage prices). Binance publishes it for its default 8-hour interval, and Hyperliquid uses the same form paid hourly (checked 2026):

where $\bar{\pi}_j$ is the premium averaged over funding interval $j$ and $\iota$ an "interest" component of 0.01% per 8 hours. While the average premium sits between −0.04% and 0.06%, funding is exactly $\iota$ — a built-in 10.95% a year from longs to shorts when nothing is happening. Outside that band funding follows the premium, so a gap costs the crowded side about the gap itself, every interval.

What funding does to the premium: a toy model

The lab simulates 2,000 years of hourly trading, with spot at 60% volatility. Speculative demand pushes the premium — a steady long bias plus 2-basis-point hourly shocks, correlated 0.5 with spot because buyers chase rallies — and arbitrage, paid by funding, closes a fraction $\psi$ of the gap each hour (a 8-hour half-life):

The counterfactual switches funding off: nobody is paid to close the gap, $\psi = 0$, and the premium is a random walk with drift. The model assumes that funding-paid arbitrage closes $\psi$ of the gap each hour (the funding it computes never feeds back), so it illustrates the mechanism rather than testing it; both regimes match their theory. With funding, the premium after a year averages +0.021% (theory +0.020%) with a standard deviation of 0.049% (theory 0.050%), and 99% of paths sit within 0.14% of spot. Without it, the same shocks leave the perp 14.6% above spot on average, give or take 1.8%.

Funding was exactly $\iota$ in 72% of intervals and negative in 6%. A year of funding averaged 13.9% of notional (5th–95th percentile 12.5% to 15.4%), above the 10.95% that $\iota$ alone pays, because the premium averages 0.02%, above $\iota$, so more intervals break out above the band — where funding follows the premium up — than below it.

2.6.7 Basis trades, liquidation & auto-deleveraging

The funding that longs pay is income for somebody. Collecting it is the crypto basis trade — the cash-and-carry of futures, with two new ways to lose.

The trade

With a dated future it is the cash-and-carry of Session 2.4: buy spot, sell the future, and lock in the futures premium over spot, $F/S - 1$, whatever the price does. Crypto desks call that premium "the basis"; in 2.4's basis convention, $S - F$, the same gap is negative, so check the sign. Crypto futures premiums often sit far above dollar interest rates: Schmeling, Schrimpf and Todorov (2023) document bitcoin and ether carry averaging above 10% a year and at times 60%, left on the table because arbitrage is risky "due to spikes in margins and liquidations amid drawdowns". Whether it is a risk premium is a question for Session 8.4.

With a perp you buy one unit of spot, short one unit of the perp and collect funding. The legs cancel except for the gap between them, so the year's gain is exactly

— funding received, minus any widening of the premium. After 0.20% of fees the trade earned 14.0% of starting notional on average (5th–95th percentile 6.9% to 23.9%, since the notional moves with the price) and 0% of paths lost money, while the spot leg alone had a standard deviation of 64%. With the short leg at 3×, capital is 1.33 times notional and the return on it 10.5% — assuming the short is topped up and never liquidated, and that funding stays as positive as this model makes it; real funding has turned negative for long stretches. It is not free money, for two reasons.

1 · The hedge leg can be liquidated. Hold the spot in cold storage and the short perp at 3× on an exchange, and a 32.7% rally liquidates the short while the offsetting gain sits in a wallet the exchange cannot see. In the lab that happens within a quarter on 29% of paths and within a year on 54%, leaving you long spot at the top of a rally. 2 · The venue can fail (§2.6.2). Assume, purely for illustration, a 2% annual failure probability with nothing recovered. If both legs sit on the failed venue, the expected return falls from 10.5% to 8.3%, and it breaks even at a failure probability of 9.5%; with the spot in cold storage only the margin, 25% of the capital, is at stake, and the expected return is 9.8%. The average is not the problem; the loss tail is.

How a crypto venue liquidates you

Liquidation is the forced closing of a position by the venue when its margin runs out. In Session 2.4 a margin call asked you for cash, and the position was closed only if you could not pay. A crypto venue checks equity against the maintenance margin continuously, on a mark price built from spot indices rather than its own last trade, and there is no call: the engine takes over. For an isolated-margin long with leverage $L$ and maintenance rate $m$ (a fraction of the current notional), that happens when the price falls to

— almost exactly the $d^\star = 1/L - m$ of Session 2.4. Then comes a waterfall (fig 2.6.7). At the bankruptcy price the position's equity is exactly zero. If the engine sells before it, part of the leftover margin feeds an insurance fund as a fee; if the price gaps through it, the fund absorbs the loss; and if the fund runs dry, auto-deleveraging (ADL) closes profitable positions on the other side, ranked by profit and leverage (Binance's auto-deleveraging FAQ).

On 10 October 2025 more than $19 billion of leveraged crypto positions were liquidated within 24 hours, the largest such day on record (CoinDesk Research), and on the on-chain venue Hyperliquid, ADL closed $2.1 billion of positions in 12 minutes (Chitra 2025): traders who had been right had their winners shut.

How long does a leveraged position last?

The lab simulates 20,000 one-year hourly paths in calendar time at 60% volatility, no drift and a 0.5% maintenance margin; the first breach liquidates. Real engines check more often than hourly, so these slightly understate.

These match the barrier-crossing formula of Session 2.4, corrected for hourly checks, within four standard errors (largest gap 1.7) wherever the liquidation price is at least five hourly standard deviations (0.64% each) away: up to 25× here. (Monte Carlo is Session 3.5.)

2.6.8 The risk stack & regulation

Traditional markets hide most of these risks behind law, institutions and a closing bell. In crypto they are all on your side of the screen, stacked under every position you hold.

Regulation, as of September 2026

Rules differ by jurisdiction and are changing fast; these are dated examples, not advice.

For a quant, regulation changes the bottom of the risk stack — an authorised venue must segregate client assets and publish its rules — not the top.

2.6.9 Try it

Open the companion lab, predict before you run, and check your prediction against the asserts.

2.6.10 Glossary · Check yourself · Go deeper

Check yourself

📚 Go deeper

1. Adams, H., Zinsmeister, N. & Robinson, D. (2020). "Uniswap v2 Core." Whitepaper, March 2020. https://app.uniswap.org/whitepaper.pdf
2. Milionis, J., Moallemi, C. C., Roughgarden, T. & Zhang, A. L. (2022). "Automated Market Making and Loss-Versus-Rebalancing." arXiv:2208.06046
3. He, S., Manela, A., Ross, O. & von Wachter, V. (2022, revised 2026; version 7 of 17 September 2026). "Fundamentals of Perpetual Futures." arXiv:2212.06888
4. Schmeling, M., Schrimpf, A. & Todorov, K. (2023). "Crypto Carry." BIS Working Papers No. 1087. https://www.bis.org/publ/work1087.htm
5. Liu, J., Makarov, I. & Schoar, A. (2023). "Anatomy of a Run: The Terra Luna Crash." NBER Working Paper 31160 (April 2023; since published in the Journal of Financial Economics, vol. 184). doi:10.3386/w31160
6. Buterin, V. (2022). "Having a safe CEX: proof of solvency and beyond." 19 November 2022. https://vitalik.eth.limo/general/2022/11/19/proof_of_solvency.html
7. Makarov, I. & Schoar, A. (2020). "Trading and Arbitrage in Cryptocurrency Markets." Journal of Financial Economics 135(2), 293–319. doi:10.1016/j.jfineco.2019.07.001

Sources for the numbers

1. Adams, H., Zinsmeister, N., Salem, M., Keefer, R. & Robinson, D. (2021). "Uniswap v3 Core." Whitepaper, March 2021. https://app.uniswap.org/whitepaper-v3.pdf · Uniswap developer documentation, "Fees" (checked 2026). https://developers.uniswap.org/docs/get-started/concepts/fees · Uniswap governance, "UNIfication" (proposal 93, executed 28 December 2025). https://vote.uniswapfoundation.org/proposals/93
2. Angeris, G., Kao, H.-T., Chiang, R., Noyes, C. & Chitra, T. (2019, revised 2021). "An analysis of Uniswap markets." arXiv:1911.03380
3. Daian, P., Goldfeder, S., Kell, T., Li, Y., Zhao, X., Bentov, I., Breidenbach, L. & Juels, A. (2020). "Flash Boys 2.0: Frontrunning in Decentralized Exchanges, Miner Extractable Value, and Consensus Instability." IEEE Symposium on Security and Privacy. arXiv:1904.05234
4. Shiller, R. J. (1993). "Measuring Asset Values for Cash Settlement in Derivative Markets: Hedonic Repeated Measures Indices and Perpetual Futures." Journal of Finance 48(3), 911–931. doi:10.1111/j.1540-6261.1993.tb04024.x
5. Ackerer, D., Hugonnier, J. & Jermann, U. (2026). "Perpetual Futures Pricing." Mathematical Finance. doi:10.1111/mafi.70018
6. Chitra, T. (2025, revised 2026). "Autodeleveraging: Impossibilities and Optimization." arXiv:2512.01112
7. Cong, L. W., Li, X., Tang, K. & Yang, Y. (2023). "Crypto Wash Trading." Management Science 69(11), 6427–6454. doi:10.1287/mnsc.2021.02709
8. Ma, Y., Zeng, Y. & Zhang, A. L. (2025). "Stablecoin Runs and the Centralization of Arbitrage." NBER Working Paper 33882. https://www.nber.org/papers/w33882
9. IOSCO (2023). "Policy Recommendations for Crypto and Digital Asset Markets: Final Report." FR11/2023, 16 November 2023. https://www.iosco.org/library/pubdocs/pdf/IOSCOPD747.pdf
10. SEC Rule 15c3-3, "Customer protection — reserves and custody of securities." 17 CFR 240.15c3-3. https://www.ecfr.gov/current/title-17/chapter-II/part-240/subpart-A/subject-group-ECFR541343e5c1fa459/section-240.15c3-3
11. FTX: Allison, I. (CoinDesk, 2 Nov 2022) and CoinDesk timeline (9 Nov 2022) https://www.coindesk.com/layer2/2022/11/09/8-days-in-november-what-led-to-ftxs-sudden-collapse · CNBC, "Binance backs out of FTX rescue" (9 Nov 2022) https://www.cnbc.com/2022/11/09/binance-backs-out-of-ftx-rescue-leaving-the-crypto-exchange-on-the-brink-of-collapse.html · Declaration of John J. Ray III, In re FTX Trading Ltd., Case 22-11068 (Bankr. D. Del., 17 Nov 2022) https://www.documentcloud.org/documents/26141228-affidavitdeclaration-in-support-of-first-day-motion-declaration-of-john-j-ray-iii-in-support-of-chapter-11-petitions-and-first-day-pleadings-filed-by-ftx-trading-ltd/
12. SEC press release 2022-219 (13 Dec 2022) https://www.sec.gov/newsroom/press-releases/2022-219 · US Attorney's Office, SDNY, "Samuel Bankman-Fried Sentenced To 25 Years In Prison" (28 Mar 2024) https://www.justice.gov/usao-sdny/pr/samuel-bankman-fried-sentenced-25-years-prison · CNBC, "Sam Bankman-Fried found guilty on all seven criminal fraud counts" (2 Nov 2023) https://www.cnbc.com/2023/11/02/sam-bankman-fried-found-guilty-on-all-seven-criminal-fraud-counts.html
13. CNBC, "FTX bankruptcy judge approves more than $14 billion payback plan" (7 Oct 2024) https://www.cnbc.com/2024/10/07/ftx-bankruptcy-judge-approves-more-than-14-billion-payback-plan.html · FTX Recovery Trust, "FTX Announces Fifth Distribution of Approximately $900 Million to Creditors on July 31, 2026" (PR Newswire, 17 Jul 2026) https://www.prnewswire.com/news-releases/ftx-announces-fifth-distribution-of-approximately-900-million-to-creditors-on-july-31-2026-302828726.html · FTX, "FTX Announces Initial Distribution Date of February 18, 2025 for Convenience Class Creditors" (PR Newswire, 7 Feb 2025) https://www.prnewswire.com/news-releases/ftx-announces-initial-distribution-date-of-february-18-2025-for-convenience-class-creditors-302371626.html · Cointelegraph, "FTX debtors assess value of crypto claims based on petition date market prices" (16 Dec 2023) https://cointelegraph.com/news/ftx-reorganization-plan-crypto-assets-bankruptcy
14. CoinDesk, "Auditing Firm Mazars Pauses Proof-of-Reserves Work for Crypto Clients" (16 Dec 2022) https://www.coindesk.com/business/2022/12/16/binance-proof-of-reserves-auditor-mazars-pauses-all-work-for-crypto-clients · FBI, "North Korea Responsible for $1.5 Billion Bybit Hack" (26 Feb 2025) https://www.ic3.gov/psa/2025/psa250226 · CNBC, "Crypto exchange Bybit says it fully replenished reserves after record $1.5 billion hack" (24 Feb 2025) https://www.cnbc.com/2025/02/24/bybit-replenished-reserves-after-record-breaking-1point5-billion-hack.html
15. USDC, DAI and SVB: CoinDesk, "DAI Depegs to Lifetime Lows as Stablecoin Rout Plagues Crypto" (11 Mar 2023) https://www.coindesk.com/markets/2023/03/11/dai-depegs-as-stablecoin-rout-plagues-crypto · Cointelegraph, "Circle's USDC instability causes domino effect on DAI, USDD stablecoins" (11 Mar 2023) https://cointelegraph.com/news/circle-s-usdc-instability-causes-domino-effect-on-dai-usdd-stablecoins · CoinDesk (11 Mar 2023) https://www.coindesk.com/business/2023/03/11/circle-confirms-33b-of-usdcs-cash-reserves-stuck-at-failed-silicon-valley-bank · CNN (11 Mar 2023) https://www.cnn.com/2023/03/11/business/stablecoin-circle-silicon-valley-bank · Joint statement by the Treasury, Federal Reserve and FDIC (12 Mar 2023) https://www.federalreserve.gov/newsevents/pressreleases/monetary20230312b.htm
16. UST: CNBC, "Cryptocurrency luna crashes to $0 as UST falls further from dollar peg" (13 May 2022) https://www.cnbc.com/2022/05/13/cryptocurrency-luna-crashes-to-0-as-ust-falls-from-peg-bitcoin-rises.html · USDe: Godbole, O., "No, Ethena's USDe Didn't De-peg During Friday's Crash" (CoinDesk, 13 Oct 2025) https://www.coindesk.com/markets/2025/10/13/no-ethena-s-usde-didn-t-de-peg
17. Funding and liquidation rules (checked 2026): Binance, "Introduction to Binance Futures Funding Rates" https://www.binance.com/en/support/faq/360033525031 · Binance, "Binance Futures Liquidation Protocols", "Introduction to Futures Insurance Funds" and "What Is Auto-Deleveraging (ADL) and How Does It Work?" https://www.binance.com/en/support/faq/360033525471 https://www.binance.com/en/support/faq/binance-futures-liquidation-protocols-360033525271 https://www.binance.com/en/support/faq/introduction-to-futures-insurance-funds-360033525371 · Hyperliquid Docs, "Funding" https://hyperliquid.gitbook.io/hyperliquid-docs/trading/funding · BitMEX, "Perpetual Contracts Guide" https://www.services.bitmex.com/app/perpetualContractsGuide · Coinbase, "US Perpetual-Style Futures 101" https://www.coinbase.com/learn/futures/us-perpetual-style-futures-101
18. BitMEX leverage: The Block, "BitMEX's bitcoin perpetual swap now offers 250x leverage ahead of halving" (19 Apr 2024) https://www.theblock.co/post/289766/bitmexs-bitcoin-perpetual-swap-250x-leverage-halving
19. 10 October 2025: CoinDesk Research, "Market Spotlight: Inside Crypto's $19 Billion Liquidation Event" (17 Oct 2025) https://www.coindesk.com/research/market-spotlight-the-19-billion-liquidation-that-shook-crypto
20. Regulation: Regulation (EU) 2023/1114 (MiCA) https://eur-lex.europa.eu/eli/reg/2023/1114/oj · ESMA, "Statement on the end of transitional periods under MiCA" (2026) https://www.esma.europa.eu/sites/default/files/2026-04/ESMA75-113276571-1679_Statement_on_the_end_of_transitional_periods_under_MiCA.pdf · CRS Insight IN12553, "Stablecoin Legislation: An Overview of the GENIUS Act of 2025 (P.L. 119-27)" https://www.congress.gov/crs-product/IN12553 · A&O Shearman, "The GENIUS Act: transforming US stablecoin regulation" (21 Jul 2025; effective date) https://www.aoshearman.com/en/insights/ao-shearman-on-fintech-and-digital-assets/the-genius-act-transforming-us-stablecoin-regulation · CoinDesk, "Crypto Clarity Act flames out in failed U.S. Senate vote" (15 Sep 2026) https://www.coindesk.com/policy/2026/09/15/crypto-clarity-act-flames-out-in-failed-u-s-senate-vote · SEC, Chair Gensler's statement on spot bitcoin ETPs (10 Jan 2024) https://www.sec.gov/newsroom/speeches-statements/gensler-statement-spot-bitcoin-011023 · NYDFS, virtual currency businesses (BitLicense) https://www.dfs.ny.gov/virtual_currency_businesses · FCA, "A new regime for cryptoasset regulation" (updated 22 Sep 2026) https://www.fca.org.uk/firms/new-regime-cryptoasset-regulation