What price a stop order watches

Your stop triggered, and the chart on the same venue shows a low one increment above the level you set. Or the reverse: the chart clearly traded through the level and nothing happened.

Neither is a malfunction. A trigger is a comparison between a reference price and a level you chose, and the reference price is a venue setting that is easy to leave at a default and never look at again. The chart is a rendering of one of the candidate references, and often not the one doing the comparing.

A trigger has two halves

Every conditional order — a stop, a stop-limit, a take-profit — is two objects glued together. There is a trigger, which is a rule the venue evaluates, and there is a child order, which is what the venue submits once the rule is satisfied. Neither half guarantees a price, and they fail in different ways, so it is worth holding them apart.

The trigger’s rule needs three inputs: a level, a direction, and a price to compare against the level. The level and the direction come from you. The third input frequently does not.

The three candidate references

Derivatives venues generally maintain three prices for an instrument, and any of them can be wired to a trigger. They are described in full elsewhere; what matters here is how each one behaves as a trigger input.

Last traded price. The most recent print on this venue’s book. It is the price the chart usually draws, so a last-referenced trigger is the one that behaves as the chart implies. It is also the most local and the twitchiest: a single small aggressive order in a thin book produces a print, and a print is all the comparison needs.

Mark price. The venue’s fair-value estimate, typically built from an external index. It moves smoothly and resists a local wick, which is why liquidation generally references it. As a trigger input this cuts both ways: it will ignore a local spike your chart shows plainly, and it will fire on an index move that never printed on this venue at all.

Index price. The external composite itself, with no contract-specific adjustment. Even less local than the mark, and further still from what the chart draws.

Spot venues typically offer only the last price, so the question does not arise there. On a derivatives venue it arises on every conditional order, and the default is chosen by the venue.

The condition, and how often it is checked

The comparison is not always the naive one.

Direction and inclusivity. A sell stop at a level fires when the reference falls to or below it on most venues, and strictly below it on some. One increment of difference decides whether an exact touch counts.

Continuous versus sampled evaluation. A matching engine that evaluates triggers on every price update will catch a level touched for a microsecond. A venue that evaluates on a periodic cycle sees only the reference’s value at each check, so a level crossed and recovered between two checks was never crossed as far as the trigger is concerned. This is the same distinction that makes a snapshot feed differ from the book, applied to a rule instead of a display.

Sustained-breach conditions. A few venues require the reference to hold beyond the level rather than merely touch it. Where that exists it is a documented parameter, not a hidden smoothing.

Triggering is not filling

Once the rule is satisfied, the trigger is finished and its job is done. What exists from that moment is an ordinary order, entering the matching engine with no special standing, at whatever the book looks like in that instant.

Two consequences follow. First, a stop-market child will execute and its price is whatever the levels give it, which is the entire reason a stop is a trigger and not a price. Second, a stop-limit child can fail to execute at all, because its limit constrains price and therefore cannot constrain execution.

There is also a crowding effect that is purely mechanical. Conditional orders are not resting in the book — they are held as instructions and are invisible in depth — so a level with many triggers behind it looks exactly like a level with none. When the reference reaches it, all of them convert to aggressive orders within the same interval and consume the same depth. The impact is caused by the triggered flow, not observable in advance from the ladder.

The mechanism

THE MECHANISM — a stop's trigger

  · You set a level and a direction
                    → the venue compares it to a
                      REFERENCE PRICE you may not have
                      selected.

  · Reference is the last traded price
                    → behaves as the chart implies. A
                      small print in a thin book
                      satisfies it.

  · Reference is the mark or index price
                    → can fire with NO local print, or
                      ignore a wick the chart shows.

  · The rule is satisfied
                    → trigger spent. An ordinary order,
                      with no standing in the queue.

  · Child is a market order
                    → executes. Price is whatever the
                      levels give it.

  · Child is a limit order
                    → NO GUARANTEE OF EXECUTION. The
                      limit binds price, so it cannot
                      also bind the fill.

  · Many triggers behind one level
                    → invisible in depth, then all
                      aggressive at once.

  · Default reference, touch versus
    sustained breach, evaluation cadence
                    → VENUE-SPECIFIC. Each of them is
                      a per-venue choice.

Worked example

Illustrative figures throughout, synthetic and round, describing no real venue.

Suppose a long position with a sell stop set at a level of 39,500, and suppose that at the moment of interest the venue’s last traded price is 39,540 while its mark price is 39,530.

Case one: last-referenced, thin book. A small aggressive sell of 0.1 units clears the nearest bids and prints at 39,498. The comparison is satisfied on that print. The child market order is submitted and walks the remaining bids, filling an average of 39,430 because the depth that would have absorbed it was the depth the 0.1-unit order just consumed. The chart shows a thin spike to 39,498 and a fill 68 below it.

Case two: mark-referenced, index move. The external index falls; the mark follows to 39,499 while this venue’s last print is still 39,540, because nobody has crossed the book here yet. The trigger fires. The chart shows no candle anywhere near 39,500, and the position is closed.

Case three: mark-referenced, local wick. A large aggressive order prints 39,480 on this venue while the mark, anchored to the index, only reaches 39,528. Nothing triggers. The chart shows a clean break of the level and the stop is still open.

All three are the same order with the same level, and the only difference is which price the venue was comparing.

The failure mode

The characteristic error is treating the level as a property of the chart. It is a property of a comparison, and both inputs to that comparison are configurable — one by you, one often by default.

Three ways that bites, all documented behaviour. A last-referenced trigger can be satisfied by a print so small it would be invisible on the tape as anything but a single record, and the aggressive flow that produced the print is also the flow that removed the depth the child order then needs. A mark-referenced trigger can fire without a single trade at the level on the venue holding the order, which makes the fill impossible to reconstruct from that venue’s chart alone. And if the reference reaches the level while the instrument is suspended or in an auction, the trigger may fire into a book that is not matching, so the child order rests or is rejected instead of executing.

None of this makes a stop useless and none of it makes one protective. It is a trigger. The reference it watches is discoverable in advance from the venue holding the order, and that is the only place the answer lives.