Trading halts & LULD

Definition

US equities operate under a layered halt regime: LIMIT UP-LIMIT DOWN price bands on every stock (trading pauses if the price can’t return inside a moving band), MARKET-WIDE circuit breakers on the S&P 500 (7%, 13%, 20% decline tiers), and discretionary NEWS/regulatory halts. All of it is post-crisis engineering — 1987 built the market-wide breakers, 2010 built the single-stock bands, 2015 refined the reopenings — and a position’s realizable exit path runs through these rules, so the engine carries them as mechanics, not trivia.

How it works / structure

• LULD bands (the always-on layer): each stock gets a band around a rolling 5-minute reference price — ±5% for liquid Tier 1 names, ±10% for others, doubled in the open/close windows and for low-priced stocks; a quote stuck at the band for 15 seconds triggers a 5-minute pause, reopened by auction; repeated at the new level if imbalance persists (the 2015 cascade pattern — episode-2015-etf-dislocation).
• Market-wide breakers: S&P 500 −7% (Level 1) and −13% (Level 2) each halt everything 15 minutes (once per day each, not after 3:25pm); −20% (Level 3) ends the session — fired four times in March 2020 (episode-covid-2020), the only Level-1 uses since the 2013 redesign.
• News halts (T1/T2) and regulatory halts (SEC suspensions): dated, discretionary, and — for pending-news halts around binary events — the state a position can sit in while its value is being decided (event-fda-approvals decisions frequently land inside halts).
• Engine-relevant consequences: halted positions can’t be exited or hedged in the name itself (options quoting also pauses; correlated instruments become the only hedge); reopening auctions concentrate discovery (ms-sessions-auctions); stop orders don’t work inside a halt — they queue for the reopen print.

When it applies

Dislocation execution (band-aware order placement; the reopen auction as the liquidity event); binary-event positions (the halt-then-gap path is the normal case for FDA/M&A news); short positions in squeeze conditions (limit-up sequences remove the exit); volatile small caps (where bands fire routinely and reopen prints define the session).

Risk profile & failure modes

• The exit illusion: risk plans assuming continuous exit meet a regime designed to interrupt continuity — halt-aware scenarios (position stuck N minutes, reopens X% away) belong in sizing for event names.
• Band-cascade dynamics: 2015 documented reopenings re-triggering bands — pauses can prolong dislocation; the fix (amended reopening procedures) helped but the pattern remains possible.
• Halt asymmetry for shorts: repeated limit-up pauses in a squeeze leave shorts watching losses compound between prints — the meme-squeeze record (episode-meme-squeeze-2021).
• Misreading news halts: a halt is not information about direction — resumption gaps go both ways; pre-positioning “for the halt” is event risk, not edge.

Evidence & limits

The LULD plan, breaker tiers, and halt codes are SEC/exchange-documented; the firing history (2020 breakers, 2015 cascade counts) is public record. Band parameters and procedures are amended periodically — the entry carries the architecture; current thresholds are verifiable against the live rule set.

Falsifiable-thesis examples

Illustrations only, not signals:

• “This event position’s plan survives a halt-then-reopen −30% path within the account’s drawdown budget (halt-aware sizing audit)” — falsified by the scenario computation.
• “LULD reopening auctions in names halted once print within the next band more than 80% of the time (cascade-rarity check)” — falsified by the halt-data tally.

Cross-references

• The episodes that built it: episode-1987-crash, episode-flash-crash-2010, episode-2015-etf-dislocation
• The mechanics it interrupts: ms-order-types, ms-sessions-auctions
• The positions it binds: event-fda-approvals, episode-meme-squeeze-2021 (short-side asymmetry)