Circuit breakers

Pre-set rules that pause trading when prices move too far too fast. US equities use single-stock LULD bands plus market-wide breakers; they were tightened after the 2010 Flash Crash. Most venues, including crypto, now run an equivalent.

The idea

Why does a continuous market need a circuit breaker?

A continuous market matches orders instantly, which is efficient until liquidity vanishes; then prices can spiral as automated sellers hit a thinning book and trigger more selling. A circuit breaker interrupts that feedback loop: it pauses trading so participants can assess, market makers can re-quote, and the market can reopen via an auction at a sane price rather than free-fall.

The intuition first: in continuous trading there is no natural "stop and breathe", because every order matches the instant it can. Normally that is fine. But when liquidity evaporates (market makers widen or pull quotes as toxicity spikes), a wave of automated selling can sweep through a near-empty book, printing absurd prices, which trips stop-losses and feed-driven algos into more selling: a liquidity-driven crash, not a fundamental one. The circuit breaker is the deliberate "stop" the continuous market otherwise lacks.

There are two failure modes a circuit breaker guards against. A single-stock dislocation (one name craters or spikes on a liquidity gap or an erroneous order) is caught by LULD (below). A market-wide cascade (a broad, correlated plunge across the whole market) is caught by market-wide circuit breakers (below).

The design tension: a halt prevents a free-fall but also removes liquidity and certainty at the worst moment (you cannot trade out during a halt). Circuit breakers are a deliberate trade-off, a brief loss of immediacy in exchange for breaking the spiral, which is why their parameters are tuned, debated, and not universally loved.

The 6 May 2010 Flash Crash: the event that built the modern breakers

On 6 May 2010, US equity indices plunged about 9% and recovered within minutes; some stocks traded at absurd prices (cents, or \$100,000) before snapping back. A large automated sell program met thinning liquidity and triggered a cascade across the interconnected, fragmented market. The regulatory response built today's LULD and modernised circuit-breaker regime.

What happened, in brief (source: CFTC–SEC joint report, Findings Regarding the Market Events of May 6, 2010, 30 September 2010): a large automated sell order in E-mini S&P 500 futures, executed aggressively into already-stressed liquidity, helped trigger a rapid, self-reinforcing decline. As prices fell, automated liquidity providers widened or withdrew, the book thinned, and the same selling pressure swept across futures and equities and their many venues. Some stocks printed at \$0.01 or \$100,000 as orders swept to absurd levels in an empty book, before the market recovered most of the move within roughly 20 minutes.

Why it mattered structurally: it was the canonical demonstration that a fast, automated, fragmented market can dislocate violently without any fundamental cause, purely from a liquidity-and-feedback failure. The pre-2010 single-stock circuit-breaker patchwork was inadequate to it.

The response: regulators introduced single-stock circuit breakers, then the Limit-Up/Limit-Down (LULD) mechanism (piloted from 2012, made permanent), the clearly-erroneous-trade break rules, and revised market-wide circuit breakers to be index-based and faster. The Flash Crash is to circuit breakers what Knight Capital is to kill switches: the cautionary event that made the control mandatory.

Limit-Up/Limit-Down (LULD): the single-stock band

LULD prevents a single US stock from trading outside a price band set around a rolling reference price (a recent average). If the best bid or offer would move beyond the band, the stock enters a limit state; if it stays there about 15 seconds without re-entering, trading pauses briefly and reopens via an auction. It stops one stock free-falling on a liquidity gap.

The mechanics (illustrative; exact percentages depend on the stock's tier and time of day): a reference price is computed as a rolling average (e.g. over the last 5 minutes); price bands are set at plus-or-minus a percentage of the reference (e.g. ±5% for the most liquid "Tier 1" names in core hours; wider for less liquid stocks, and widened near the open and close), and quotes may not trade through the band; and if the market sits at a band edge (a limit state) for about 15 seconds without re-entering, a LULD trading pause of about 5 minutes follows, and the stock reopens via a re-opening auction.

What it does and does not do: LULD does not stop a stock moving; it stops it moving faster than the band allows, channelling a violent move into a series of band-limited steps and pauses rather than one free-fall. It directly prevents the cents/\$100,000 prints the Flash Crash produced. The strategy consequence: a strategy must model the band. A market maker or momentum strategy that ignores LULD can find its orders rejected at the band, or be caught with a position when a pause freezes the market and it cannot trade out, exactly the kind of event a risk system must anticipate.

Market-wide circuit breakers (MWCBs)

Market-wide circuit breakers halt all US equity trading when the S&P 500 falls by set thresholds from the prior close. Level 1 (7%) and Level 2 (13%) each trigger a 15-minute halt (before 15:25 ET), and Level 3 (20%) halts trading for the rest of the day. They are the broad-cascade backstop, distinct from single-stock LULD.

The three levels (as of 2026; verify against the current exchange rules). Level 1, a 7% decline: a 15-minute market-wide halt (if before 15:25 ET; no halt for a Level 1 or 2 after 15:25). Level 2, a 13% decline: a further 15-minute halt (same time caveat). Level 3, a 20% decline: trading halts for the remainder of the day, at any time.

The history: MWCBs date back to the response to the 1987 crash but were recalibrated to be index-based (S&P 500) and percentage-based after the Flash Crash era, replacing the older point-based Dow triggers that had become too coarse. They have rarely fired; the most recent prominent triggers were the Level 1 halts during the March 2020 COVID volatility. The strategy consequence: a market-wide halt is a correlated, total liquidity event, with every position frozen at once. A risk system and a capacity and sizing framework must treat an MWCB as a tail scenario you cannot trade through, not a normal market condition.

The EU and other equivalents

The EU does not use a single US-style LULD/MWCB regime; instead, MiFID II requires trading venues to run their own volatility-management mechanisms, typically volatility interruptions (price-collar checks that pause a stock and reopen it via an auction when a move exceeds a venue-set threshold). The effect is similar (pause a runaway move, reopen via auction) but it is venue-level and less uniform than the US system.

The MiFID II requirement: venues must have mechanisms to manage volatility (preventing "disorderly trading conditions"), which in practice means static and dynamic price collars. If the next trade would move the price more than a set amount from a reference (the last auction price, or the last trade), the venue triggers a brief volatility interruption and reopens via a call auction. (This ties to MiFID II's market-integrity obligations.)

The contrast with the US: the US imposes a harmonised, market-wide LULD/MWCB regime; the EU delegates volatility controls to each venue, so the bands and triggers vary by exchange. Both achieve the same goal (channel a violent move into pause-and-auction rather than free-fall) by different architectures.

The shared principle worth holding: every regulated equity market answers a runaway move the same way, by stopping continuous trading and reopening with an auction. The auction aggregates resting interest and finds a single clearing price, which is exactly why continuous-vs-auction is the microstructure idea underneath all circuit breakers.

Crypto's inconsistent halts

Crypto has no market-wide circuit breaker and no uniform halt regime. Some centralised exchanges run their own price bands or brief halts (and some derivatives venues have limit mechanisms), but many do not, and there is no cross-venue equivalent of LULD or an MWCB. A coin can move 30% in seconds with nothing to pause it: a genuine structural risk, not just a missing feature.

The reality (as of 2026): crypto is natively fragmented across dozens of venues with no consolidated tape and no harmonised circuit breaker. Individual venues may impose their own protections (some run price-band rejections, brief auto-halts on extreme moves, or futures price limits) but they are inconsistent across venues, and a halt on one exchange does not pause the others.

What this means for a crypto strategy: there is no backstop. A liquidity-gap cascade (the exact failure the Flash Crash exposed in equities) can run unchecked, because no rule stops continuous trading. The protection the equity reader takes for granted is your problem to manage, through your own risk limits and kill switch, defensive quoting, and position sizing that assumes a 30%-in-seconds move is possible.

The honest framing, consistent with the regulation hub's cross-market theme: crypto's missing circuit breakers are the same coin as its missing market-abuse rules. Fewer rules means fewer constraints and fewer protections; the unregulated venue hands you both the freedom and the tail risk the regulated one removed. Prediction markets sit similarly outside the halt regime.

Worked example

Tracing the controls on concrete, dated, illustrative numbers (as of 2026; verify thresholds against current exchange rules). LULD on a single stock: a Tier 1 stock trades at \$100 with a 5-minute reference of \$100 and a ±5% band, so it may not trade outside \$95.00–\$105.00. Bad news hits; sellers push it to \$95.00, where it enters a limit state: quotes cannot trade below \$95.00. If it sits at the \$95.00 limit for about 15 seconds without bids re-entering above it, a 5-minute LULD pause triggers, and it reopens via an auction. Without LULD, the same selling into a thin book could have printed \$80 or worse in milliseconds, the Flash Crash failure mode.

A market-wide cascade: the S&P 500 closed yesterday at 5,000. Today it falls 7% (to 4,650) at 11:00 ET, triggering MWCB Level 1: all US equity trading halts for 15 minutes. If, after reopening, it reaches a 13% decline (4,350), Level 2 brings another 15-minute halt. A 20% decline (4,000) at any time triggers Level 3: trading closes for the day. The thresholds are percentage-based and index-referenced precisely so the trigger scales with the market level.

The EU contrast: the same stock on an EU venue has no harmonised band; instead the venue's volatility interruption triggers if the next trade would move price more than the venue's set collar from the last reference (say ±3% dynamic), pausing it into a short call auction. Different number, different architecture, same outcome: pause-and-auction instead of free-fall. The crypto gap: the equivalent coin on a crypto venue with no halt sees the same 7% (or 30%) move simply happen, continuously, with nothing to pause it. Your only "circuit breaker" is your own kill switch and position limits, which is the whole point of the contrast. Thresholds (LULD band percentages, MWCB levels, EU collars) are specific, dated, tier- and venue-dependent parameters; reverify against the current SEC/exchange and venue rules before relying on them (as of 2026). This is educational only and not investment advice.