What a take profit order is

What it is attached to 

A take profit opens nothing. It is written against a position that already exists, or against a pending order that has not dealt yet, in which case it is not an order at all until that pending order deals and the position appears. That dependency is the defining property of the instruction, and it is why the word attached is used rather than placed. It closes the position it belongs to, it is written for that position's size, and it stops existing the moment the position does, whether the position was closed by the level, closed by hand or closed by the firm.

Two consequences follow from being attached rather than standing alone. The first is that a take profit can never be the instruction that establishes exposure, so it can never be the reason a position exists. The second is that a position closed in part takes its attached order down in size with it on most platforms, so what remains after a partial close is a smaller position with a smaller order against it, rather than an order still written for a size that is no longer there. Where a platform behaves otherwise it is stated in that platform's own reference, and the difference shows up in the order list rather than on the ticket.

The order type it resolves to 

Underneath the label, a take profit is a limit order, and everything the earlier lesson established about limit orders applies to it without amendment. It carries the same clause: deal at the stated level or at a price better than it, and at no other price. For a long position the closing deal is a sale, so the attached order is a sell limit and it rests above the current bid. For a short position the closing deal is a purchase, so it is a buy limit and it rests below the current ask. Which side of the quote it is measured against follows from the direction of the closing deal, and from nothing else.

Set beside a stop attached to the same position, the symmetry is exact and the behaviour in a jump is opposite. Both orders close the same position, both rest on the price axis on opposite sides of it, and both wait. A stop deals at whatever is available once it is triggered, so a market that reopens some distance away carries it to a price worse than the level. A take profit is constrained to its level or better, so the same reopening deals at the price it finds on the permitted side. One event, two attached orders, and the distance the market jumped counts against one of them and not against the other.

How the level is attached 

The field exists on both of the platforms YAL runs, MetaTrader 5. On both it sits on the order ticket next to the stop loss field, and on both it is empty by default, so a position opened without touching it has no attached order of any kind. On MetaTrader 5the value is entered as a price. A position that is already open is amended through its own dialog, or by dragging the line the platform draws on the chart at that level, and a drag is an amendment to the existing instruction rather than a second order.

Both platforms hold the instruction on the server rather than in the terminal, which is the practical reason attached orders are used at all: a terminal that is closed, a machine that is switched off and a connection that drops leave the order where it was written. It follows that removing the level is a deliberate act, exactly as entering it was. An attached order that is forgotten is still working.

One detail is invisible on the ticket and worth stating. A deal that closes a position carries the same costs as any other deal on that instrument, so a level is a price rather than a net result, and the charges that apply to a closing deal entered by hand apply in the same way to one made by an attached order.

Reached without dealing 

Because a take profit is a limit order, it inherits the limit order's characteristic failure: the level can be reached, or appear to have been reached, with nothing dealt. Three mechanisms account for most of it, and all three were introduced with the limit order rather than being peculiar to attached ones. The level was touched on the other side of the quote, so the price the order is measured against never arrived. The touch happened inside a single update that reversed immediately, so there was a moment at the level during which no counterparty dealt. Or the size available at the level was consumed by orders that were already resting there, on a venue that fills in the sequence orders arrived.

The first of those is where a chart and an order most often disagree, and the disagreement is sharpest on a short position. Most charts are drawn on the bid, while a short position closes by buying and its attached order is therefore measured against the ask. In a fast market the two sides of a quote widen, and they do not widen symmetrically, so the bid can print through a level while the ask stayed above it for the whole of the move. Nothing has malfunctioned. The line on the screen and the price the order is watching are two different series, separated by the distance between the two sides of the quote, which is itself variable.

The level is a plan input, not a forecast 

A level is a number chosen before the fact, and the market has no knowledge of it. Entering one is not a statement that price will reach it, and no attached order carries any information about whether it will. It is a conditional instruction and nothing more: if this price arrives, close at it or better; if it does not, do nothing. Read that way, the level is an input to a plan, in the same category as the size in the volume field, and it is decided by whoever writes it rather than discovered in the market.

Traditions for deriving the number differ, and each carries a limit that its own adherents state. One convention sets the distance as a multiple of the distance to the stop level on the same position, which makes the two numbers consistent with each other but takes no account of whether that distance is one the instrument typically covers. Another places the level at a structural feature on a chart, a prior high or low, which is legible to anyone reading the same chart but is drawn by the person drawing it, and two chartists marking the same instrument rarely mark the same price. A third scales the distance by a volatility measure, which describes the window the measure was computed over and not the session ahead. A fourth uses no level at all and closes by discretion or by a trailing rule. None of these is put forward here, and this page knows nothing about any reader's circumstances.

Where practitioners disagree 

The standing argument is between a fixed exit and a discretionary one, and it is not decidable from the mechanics. The case for the fixed level is that it is written before the position exists, when nothing is yet at stake, and that it is recorded, so what was intended can be compared afterwards with what happened. The case against it is that a number fixed in advance is fixed against conditions that have since changed, and that closing at it is a decision made by someone who could not see the current market, namely the person who wrote it. Both descriptions are accurate. The argument persists because only one side of it is measurable: whether the level was reached is a matter of record, and what a position would have done afterwards is not, so the two traditions are comparing a fact against a counterfactual.

A narrower disagreement concerns leaving an attached order working across sessions. One view holds that a level held on the server is the only instruction that survives a closed terminal, an outage or an unattended market, and treats that survival as the whole point of writing one. Another observes that a market can reach a level for a reason that would have changed the intention had it been visible, and that an instruction written days earlier has no way to know that. Both are describing the same property of an attached order, which is that it is indifferent to context, and they disagree about whether indifference is the feature or the flaw.