Circuit limits, and the exit that is not there

A cap on the price, not a floor under the position

A price band is a limit the exchange places on how far a security's price may move during one session, measured from a reference price fixed for that session. Reaching the edge of the band stops the price from moving. It does not stop your position from needing a buyer.

Orders priced outside the band are refused by the matching system, and that refusal is the whole of how a band binds. Notice what it does and does not do. The band constrains prices — the numbers at which an order may be entered. It says nothing about how many people want to trade at those prices, and it cannot conjure anyone into existence at the edge. That distinction is the whole of this article, and it is where almost every practical surprise comes from.

A band is a property of the security, not of your order or your account. The exchange sets it, publishes it before the session, and applies it to everyone at once. Not every security carries the same arrangement either — the width differs, and some instruments operate under a different kind of band altogether, which is an exchange decision published per security.

Keep three actors apart here, because reporting routinely merges them and a reader who merges them cannot check anything. The regulator sets the framework the exchanges work within. The exchange sets and publishes the band, refuses orders outside it and halts trading. Your broker may then impose limits of its own, refusing an order the exchange would have accepted or demanding more margin on a security near its band — that is the broker's commercial decision and not a market rule, however it appears in the error message.

The vocabulary is loose in ordinary use, so fix it once. A stock at the top edge of its band is at its upper circuit; at the bottom edge, its lower circuit. When no further trading happens there because everyone left is on the same side, the security is locked. A halt that stops the entire market rather than one security is a different object with a different trigger, and the two get the same nickname in conversation.

The loop a pause is meant to interrupt

Start with why anyone would want the price to stop, because the case is a mechanical one and it is more interesting than the usual talk of panic.

In an ordinary market a falling price is self-correcting. It gets cheaper, some buyer who thought it expensive now thinks it fair, and the buying slows the fall. The price move produces a force pointing the other way, which is what keeps a market orderly on most days without anyone intervening.

Selling that is forced breaks that. A leveraged holder whose collateral has fallen below what the loan requires must sell, and must sell now, regardless of what they think the security is worth. A cluster of stop-loss orders below the market becomes, on being triggered, a wave of sell orders that nobody chose to send that morning. In both cases the sale is a consequence of the price rather than an opinion about it.

Now the price move produces a force pointing the same way. Lower prices trigger selling, the selling lowers prices, and the lower prices trigger more selling. That is a feedback loop in the strict sense, and the reason it is dangerous is not that people are frightened — a margin call is arithmetic, and it arrives whether or not anyone is calm. The buyers who would ordinarily absorb the supply are slower than the loop: they have to notice, raise cash, get approval, form a view.

Stated that way, the case for a pause is a case about clock speed. Halting trading does not make anyone want the security more, and nobody claims it does. It buys the slow side time to arrive, turning a race into a queue, and it forces the forced sellers to wait alongside everyone else.

The cost is exact and falls on the innocent. The same pause that stops a panicked seller stops the person selling for entirely ordinary reasons — a fee due on Monday, a rebalance, a decision made weeks ago. A market cannot suspend one motive for trading, so it suspends all of them. That is the trade-off, and no design avoids it.

Two different things wear the same name

A security-level band and a market-wide halt are both called circuit breakers and they are not the same object. They differ in what they apply to, what sets them off and what keeps running while they bite.

The market-wide version is tiered: bigger moves buy longer pauses, and the thresholds, the durations and which index is watched are exchange rules, published in advance and revised from time to time. Look them up rather than remembering them, and look them up on the exchange's own site — this is exactly the kind of figure that circulates in secondary coverage years after it changed.

Restarting a market that has stopped is a problem in its own right, and the reason is mechanical. Continuous matching needs a running price to resume from, and a halt has destroyed exactly that, so an exchange may reopen through a single auction instead — orders collected over a window, no trades struck while they arrive, then one clearing price set for everybody at the same instant. Which arrangement applies sits in the same published rulebook as the thresholds.

One asymmetry is worth carrying away. A security-level band affects your position in one name, and the rest of your portfolio remains tradeable, so a diversified holder can usually raise cash elsewhere. A market-wide halt removes that option for everybody simultaneously, which is why the two are governed so differently even though the headline reads the same.

Locked is a fact about liquidity, not about value

Here is the part that catches people, and it is worth going through the order book slowly, because the conclusion is counterintuitive and completely mechanical.

Suppose a security has fallen to its lower band. Anyone who still wants out can only offer at the band price or above, because a lower price cannot be entered. Anyone who might have bought had it fallen further cannot express that either, for the same reason. So the book fills with sellers at one price and the buy side stands empty.

Both conditions are true at once: the price is at its limit, and nothing is trading. A limit on the price at which an order may be entered does not create a counterparty at that price. It removes the prices at which the remaining buyers were willing to act, which is a different thing from finding them.

Which gives the sentence to remember. A locked circuit is a statement about liquidity and not about value — it tells you that nobody will trade at the only price permitted, and it tells you nothing whatever about what the security is worth. The market has not decided the price is right. It has been prevented from saying otherwise.

Read the last traded price on the screen in that light. It is the price at which the last pair of participants still disagreed enough to trade with each other. Once the security is locked, everyone remaining agrees about direction, and agreement is precisely the condition under which nothing trades. The number is a historical fact rather than an offer.

The mirror image traps people in the other direction. A long queue of buy orders at the upper circuit looks like overwhelming demand, and it is largely an artefact of the cap: those buyers cannot bid higher, so they pile up at the one price they are allowed to name. An unexecuted order is also not a commitment. It can be withdrawn the instant the band releases, and queue length measures neither conviction nor money.

One consequence is easy to miss. The band binds an instrument, not the news — so if a derivative on the same underlying is still trading, or the same company is listed elsewhere, opinion keeps forming there while the locked security says nothing. A halt suspends the venue, not the information, and the information is still there when trading resumes.

What a lock does to a position you already hold

It removes your exit and leaves everything else running. The stop you set does not fire you out, the margin on the position is still collected on schedule, and tomorrow's band is measured afresh so the same thing can happen again. Three consequences, all hard.

A stop-loss does not protect you. A stop is an instruction that produces an order when a price is reached. It is not an exit — the order it produces still has to meet somebody, and an order needs a counterparty. In a locked security the triggered stop joins the same queue as everyone else's and waits. This is the same class of failure as a gap through a stop overnight, and it surprises people every time because the word "stop" sounds like a guarantee.

Leverage keeps its schedule. The exit may be unavailable, but the loss is still recognised and, where the position is funded with borrowed money or carries margin, it can be collected in cash on the usual timetable. The uncomfortable shape of that is a demand for money against a position you are not permitted to close. Whether extra margin is charged on a security sitting at its band is a separate question, and partly a matter for your broker rather than the exchange.

The band bounds the session, never the sequence. Each session's band is measured from that session's own reference price, so a security can lock at its limit and then do it again the next day, and the moves compound. A band does not cap a loss. It rations the loss into instalments, and removes the moments in between at which you might have acted.

Put those three together and the honest summary of what a band gives a holder is narrow. It gives you time you did not ask for, at the cost of an exit you may badly want. Whether that trade is worth making is the argument in the next section — but the trade itself is not in doubt, and anybody describing a circuit limit as a protection for holders is describing one half of it.

Does the pause help, or only postpone?

It does both, and which effect dominates is unsettled. Two mechanisms operate at once and they point in opposite directions, which is why the disagreement here is not between sensible people and foolish ones.

The case for a pause is the one already made. Forced selling runs faster than considered buying, and a halt lets the slow side catch up. It also lets clearing systems, risk desks and regulators look at what is happening before it is finished happening, which matters most in exactly the situation the mechanism is designed for.

The case against is subtler and it is not the usual complaint about interference. When a limit is close, the ability to trade becomes a scarce resource with a visible deadline — and scarcity with a deadline creates urgency. A holder who wanted out eventually now wants out before the door closes, and their selling moves the price nearer to the limit, which sharpens the deadline further. The argument is that the band can pull the price toward itself, manufacturing some of the very rush it exists to calm. It is a real mechanism, and it is not proof that bands are harmful.

So which dominates? That is an empirical question, and it is far harder to answer than it looks, for a reason worth understanding rather than taking on trust.

To judge a halt you need the path the price would have taken without it, and that path does not exist. You cannot observe the counterfactual. Comparing halted days with ordinary days does not fix it either: a halt is triggered by an extreme move, so the halted sample is selected precisely for being unusual, and any difference you find may simply be the difference that caused the halt.

What can be said is narrower and more useful. That a band interrupts the mechanical feedback loop is not really arguable — forced sellers cannot transact during a halt, by construction. Whether interrupting it changes where the price finishes, that day or that week, is a judgement, and it is one the evidence has not settled in either direction. A reader who wants certainty here should know that the certainty on offer elsewhere is manufactured rather than measured.

What a locked session does to your numbers

It erases the day from them. A session in which a security opens at its band and never leaves it prints an open, a high, a low and a close that are all the same number. Every range-based measure computed from that session — the day's range, an average true range, a volatility estimate, a candle on a chart — therefore reads the worst day in the security's recent history as a calm, zero-range day. The instrument that was designed to signal distress is invisible to the tools that measure distress.

The check takes a moment. Where a chart shows a run of sessions with no range at all in a security that was moving violently before them, you are not looking at quiet; you are looking at a sequence of locked sessions, and any indicator averaging across that stretch has been fed silence in place of the largest moves.

The same problem reaches valuations. A fund or a portfolio holding a security it cannot sell still has to put a number on it, and the last traded price is not automatically that number — a valuation policy decides, and it exists precisely because a stale print is not a realisable value. If a scheme's NAV is calculated with locked holdings in it, the figure is an estimate of something unobservable rather than a quotation.

Two habits cover most of the exposure, and neither requires a view about anything. Check the band applying to a security before you need it rather than after, since it is published in advance and it differs between securities. And read how a trade actually gets matched once, properly, because every consequence in this article follows from the order book rather than from any rule about percentages.

FNOTrader's Market Pulse scanner charts NSE and BSE price history alongside breadth and relative-strength measures across the listed universe, and the flat-candle check applies to it exactly as to any other chart. Nothing in this article, and nothing in that scanner, says whether a security at its band is cheap. It says that the price on the screen is not a price you can transact at, which is a different and more immediately useful thing to know.

Common questions

Continue reading

More in Investing Basics · App: Market Pulse Stocks Scanner · Definitions: glossary · Free tools: calculators · All: every article