Standard stops and guaranteed stops

The instruction and the fill are two separate events 

A stop order rests with the counterparty as a conditional instruction, and it does two things in sequence rather than one thing at once. First it waits. When the market trades at or through the specified level, the condition is satisfied and the order is released. Second, the released order goes to market as an instruction to close the position at the best price available at that moment. The specified level governs the first event completely. It has no authority whatsoever over the second.

The previous lesson established that split while explaining why levels are reached at all. This lesson starts from it, because it is the entire basis on which the two order types differ. Most of the time the two events sit close enough together that the distinction never surfaces, which is exactly what makes it consequential: an instruction behaves as though it were a promise across a long run of ordinary sessions, and then behaves as what it actually is on the one occasion the difference is large.

Stated precisely, a standard stop guarantees exactly one thing: that once the market trades at or through the specified level, an instruction to close is released. It does not guarantee the price that instruction receives. It does not guarantee that the amount lost equals the distance from the opening price to the level multiplied by the size of the position. It does not guarantee that the position closes at the level, near the level, or within any stated distance of it. And where a market has no price at all for a period, it does not guarantee when the position closes either.

Where the two prices come apart 

Between the trigger and the fill sits whatever the market does in the interval. Usually that interval is measured in fractions of a second and nothing happens in it. Occasionally it is a weekend. Three circumstances account for almost all of the distance between a specified level and a realised one, and they are worth separating because a guaranteed stop is priced against all three at once.

• A market that closes and reopens. No trading occurs between the last price of one session and the first of the next, so no price inside that interval exists to be filled at. A level sitting inside it is not traded through, it is passed over.
• A scheduled release. A rate decision, an inflation print or a company's results arrive at a known time and are unknown until they arrive, and the market that reconvenes around the new information can reconvene some distance from where it was.
• A market moving faster than resting orders absorb. The depth at the level is consumed by the orders ahead in the queue, and each order after that is filled against the next price with size behind it. Nothing is broken here. The queue is simply longer than the liquidity at one price.

The difference between the specified level and the realised price is slippage, and it is not directional by nature. A close can be filled worse than the level, which is the case that matters here, and it can be filled better when a fast move overshoots and the first available price sits on the other side. A record of fills across many positions contains instances of each, in unequal proportion.

That is the sentence the rest of this lesson turns on. Because profit and loss are calculated on the full contract value rather than on the margin held against it, a loss carried past the level is measured against the whole position, so it can exhaust the margin posted and is not limited to the amount deposited. A favourable move is measured on exactly the same basis. The order type changes which party carries that arithmetic, and it changes nothing about the arithmetic itself.

What a guaranteed stop undertakes 

A guaranteed stop is not a faster stop, a better placed stop or a stop with a shorter queue. It is a contractual undertaking by the counterparty that the position will be closed at the specified level, whatever price the market actually traded at when the level was passed. If the next available price is well beyond the level, the counterparty absorbs the difference and settles the position as though the level had been traded.

What is guaranteed is a closing price, and nothing else. The position still closes, it still closes for a loss, and the loss is still real money. The undertaking says nothing about whether the level was well chosen, whether the market will continue in the same direction afterwards, or whether the position was well founded in the first place. It converts one specific uncertainty, the price at which a triggered order is filled, into a known quantity, and leaves every other uncertainty in the position untouched.

Read side by side, the two order types are the same instruction with the gap risk allocated to different parties. The discontinuity itself is a property of the market and exists either way. Under a standard stop the holder of the position carries it. Under a guaranteed stop the counterparty carries it, having agreed in advance to do so. Nothing has been removed from the world, and that is the entire explanation for why the second version has a price and the first does not.

What it costs, and how the cost is charged 

The charge for the undertaking is conventionally called the premium, and the word is borrowed from insurance for an accurate reason. Charging conventions differ between firms, and the difference is not cosmetic, because it changes what a long run of positions costs rather than what any single one costs. Three arrangements are commonly reported: a premium charged when the order is placed and retained whether or not the level is ever reached, a premium charged only if the guaranteed stop is actually triggered, and a premium folded into a wider quoted spread on instruments where the order type is offered. Under the first arrangement the cost is paid on every position. Under the second it is paid only on the positions that close at their stop.

The comparison that block invites is the wrong one to stop at. One gap set against one premium makes the undertaking look inexpensive, and one quiet week against the same premium makes it look like a fee for nothing. The calculation is neither: it compares how often discontinuities of a given size occur in a given instrument against how often the premium is charged, and no page can supply the first of those numbers. What can be said with certainty is the direction of the effect. A premium is a cost, deducted from favourable outcomes and added to adverse ones, and it moves the reward to risk of a plan exactly as the round turn cost did in an earlier lesson.

Why it is priced this way, and when it is available 

A counterparty that writes a guaranteed stop has taken on an obligation whose cost to itself is unknown in advance and, in principle, unbounded: it has agreed to settle at a level regardless of how far past that level the market went. That is an underwriting problem, and it is priced with the inputs an underwriter would use. How far the level sits from the current price matters, because a nearer level is passed over more often. So does the volatility of the instrument and its own history of discontinuity, which is why an instrument that regularly reopens away from its previous close is priced differently from one that trades around the clock. So does size, because the counterparty's exposure scales with it.

The availability restrictions that commonly accompany the order type follow from that pricing problem rather than from administrative preference. A minimum distance between the current price and the level is usual, because an undertaking to settle at a level a hair away from the market is an undertaking to absorb almost every move. Availability is typically restricted to a subset of instruments, generally the ones whose exposure the counterparty can hedge. Size limits are usual for the same reason. And the order type is not offered at all by every firm, or under every regulatory regime, and where it is offered the terms are set per instrument.

This page does not state which order types are available on any particular account, and deliberately so. Availability, minimum distances and premiums are per instrument and per firm facts that change, and the place they are stated accurately is the contract specifications for the instrument in question. What each order type does, as a mechanism, is set out in the order types reference.

Assuming one behaves like the other 

Every calculation in this module so far has taken the stop distance as an input and treated the amount it implies as the amount at risk. Position size followed from it. Portfolio heat was the sum of those amounts across open positions. A daily limit was a count of those amounts. The arithmetic is sound, and it rests on one assumption that a standard stop does not support: that the amount the level implies is the amount that will actually be lost.

The block shows the failure mode plainly. Under standard stops the aggregate the sizing arithmetic reports is a figure for ordinary conditions and a floor rather than a ceiling, and the ceiling does not exist, because there is no bound on how far a market can reopen from where it closed. Under guaranteed stops the aggregate is a real bound, and the premiums are a real cost paid on the positions that never needed it. Neither column is the safe one and neither is free. They differ in which of the two is known in advance, and a plan that treats the first as though it were the second has substituted a wish for a calculation.

Where practitioners disagree 

Whether the premium is worth paying is genuinely unsettled. One tradition treats it as an ordinary expense on exposures with obvious discontinuity, positions held across a weekend or through a scheduled release, on the reasoning that a known small cost in exchange for an unknown large one is the trade insurance has always been. The other points out that the premium is charged against every position while the discontinuity is rare, so the cost is continuous and the event is not, and holds that closing before the discontinuity or carrying a smaller position bounds the same exposure at no charge. Both are internally coherent. Settling between them requires knowing how often gaps of a given size occur in a given instrument, an empirical question neither side tends to answer with evidence.

A second disagreement is behavioural rather than arithmetical, and is reported here as an argument rather than a finding. Some practitioners hold that a bounded worst case invites larger positions than the same trader would otherwise carry, so the undertaking is spent on size rather than kept as certainty. Others regard that as an argument about the trader and not about the order type, and note that the same objection would rule out every risk control ever devised. There is no data here to settle it, so both positions are left standing.