Fill-or-Kill Order

A fill-or-kill order is an order that must be executed in full the moment it arrives or not at all. The definition comes from exchange order-handling rules rather than from a statute, and it is built on a limit order rather than on a market order. The formulation an SEC filing records for one exchange's options market is precise: a Fill-or-Kill Order is "a Limit Order that is to be executed in its entirety as soon as it is received and, if not so executed, treated as cancelled." Read the two halves separately, because each is doing different work. "In its entirety" is a quantity condition. "As soon as it is received" is a duration condition. A fill-or-kill order is the compound of the two.

That compound is what distinguishes it from its two neighbors in the same rulebooks. An All-or-None Order is "a Market or Limit Order which is to be executed in its entirety or not at all", and is treated as carrying an immediate-or-cancel time-in-force. An immediate-or-cancel order is "to be executed in whole or in part upon receipt", with any unexecuted portion "treated as cancelled". So all-or-none is the entirety condition, immediate or cancel is the immediacy condition with partials allowed, and fill or kill insists on both. One detail is easy to miss and worth keeping: the fill-or-kill definition is written for a limit order, while the all-or-none definition reaches a market order as well.

The condition that makes it useful is the condition that makes it fail. An ordinary order can be met by several counterparts at several prices, filling in pieces until it is satisfied or until its price condition stops it. A fill-or-kill order forbids that. Everything it needs has to be available at one instant, at prices its limit allows. On a security where the resting size at any single price is modest, the ordinary result is that the order is cancelled untouched. That is not a malfunction; it is the instruction working. But it means an investor who reaches for fill-or-kill in order to avoid a partial fill has usually traded a small, manageable problem for a larger one, which is holding no position at all.

The entirety condition also changes how the order interacts with the venue. In the rules read for this page, a fill-or-kill order sits in a small group of what the exchange calls contingency orders, alongside all-or-none and stop orders, and those "will not route". An order that cannot be routed away has to be satisfied out of that one venue's own book or cancelled. So the pool of available size is not the whole market's displayed size but one venue's, which narrows the odds further than the entirety condition alone would suggest.

All-or-none carries a simultaneity requirement that is easy to overlook. The same filing adds that all-or-none orders "will only execute against multiple, aggregated orders if the executions would occur simultaneously", and explains why: because of the size condition, the pieces have to be satisfied at once to avoid a priority conflict on the book. It also records that all-or-none orders received before the opening process or after the close are rejected. Both facts point the same way. An entirety condition is not merely a preference the venue tries to honor; it constrains the matching itself, and the venue would rather refuse the order than partly fill it.

There is a middle instruction, and it is being withdrawn on at least one venue. A minimum quantity order sets a floor rather than demanding everything: it "requires that a specified minimum quantity of contracts be obtained, or the order is cancelled", and like all-or-none it is treated as immediate-or-cancel. That is the instruction most investors actually want when they say they do not want a token fill, because it accepts a large partial while refusing a trivial one. The exchange whose rules define it proposed to stop offering it, saying the order type "is not utilized frequently on the Exchange" and that it was dropping it for lack of demand. So the practical menu at any venue is narrower than the concepts, and which of these instructions a particular broker exposes is a question about that broker rather than about the rules.

An all-or-none instruction takes the order out of the execution-quality statistics. The monthly order-execution reports brokers and market centers publish under Rule 605 cover "covered orders", and 17 CFR 242.600(b)(27) excludes from that term "any order for which the customer requests special handling for execution, including, but not limited to, orders to be executed at a market opening price or a market closing price, orders to be executed only at their full size, orders to be executed on a particular type of tick or bid, orders submitted on a 'not held' basis, orders for other than regular settlement, and orders to be executed at prices unrelated to the market price of the security at the time of execution." An order that must execute only at its full size is squarely in that list. The consequence is that the class of orders carrying an entirety condition is invisible in the published statistics, so there is no aggregate evidence about how well they are handled.

The order arrives at the venue carrying both conditions. The matching engine looks at what is available at that instant at prices the limit permits. If the full quantity can be satisfied there and then, it executes in one go. If it cannot, the order is cancelled in its entirety and nothing trades. There is no resting, no queue position and no second look.

A hypothetical illustration of the cost of the entirety condition. Suppose the resting interest on a venue shows 300 shares offered at $12.40 and 900 more at $12.45, and an investor wants 1,000 shares. An ordinary limit order at $12.45 takes both levels: 300 × $12.40 = $3,720 plus 700 × $12.45 = $8,715, for $12,435 and an average of $12.435 a share. A fill-or-kill order for the same 1,000 shares at a limit of $12.40 finds only 300 shares available at that price at that instant, cannot satisfy the whole quantity, and is cancelled: zero shares, zero cost, and no position. Raising the fill-or-kill limit to $12.45 would let it take both levels and execute, at the same $12,435 the ordinary limit order paid, so the condition bought nothing in that case and would have cost the whole trade in the first.

The general shape of the trade-off follows from that. A fill-or-kill order is worth using where a partial position is genuinely worse than no position, which is a real situation in a multi-leg strategy or where a minimum size is needed for a hedge to work. It is worth avoiding where the investor simply dislikes the untidiness of a partial fill, because the alternative outcome is not a tidy full fill but no trade.

Pros

• It removes the possibility of an unwanted partial position, which matters where a half-size holding would not serve the purpose.
• It resolves instantly, so there is no resting order to forget about and no exposure to a price move while the order waits.
• Being a limit order by definition, it never trades at an unbounded price: the entirety condition sits on top of a price condition, not instead of one.

Cons

• It is the order type most likely to do nothing, because everything it needs must be available at one instant.
• Contingency orders of this kind are typically not routed away, so the order is measured against one venue's resting size rather than the whole market's.
• An investor who wanted to avoid a token fill has probably asked for the wrong instruction: a minimum-quantity floor does that job, and it is not offered everywhere.
• Orders that must execute only at their full size are excluded from the covered-order definition, so they appear in none of the published execution-quality reports.
• Availability varies by venue and by broker, and the concept existing in a rulebook does not mean the instruction is on your platform.