What a limit order is

The instruction, and what it fixes 

The market order described in the previous lesson names a size and accepts whatever price the venue can do at the instant the message arrives. A limit order reverses both halves of that arrangement. It names a price and accepts uncertainty about whether a deal happens at all. One instruction fixes the timing and leaves the price open, the other fixes a boundary on the price and leaves the timing, and the existence of the deal, to the market. Neither is a version of the other with a setting changed.

The word limit is narrower than it sounds. It limits the price, and nothing else: not the size of a position, not how long the order lives, and not what the position can lose once it exists. An order that has dealt is a position like any other, and the level that governed how it opened has no further bearing on it.

Or better, and what better means 

Two words carry most of the definition, and they mean opposite things on the two sides of a deal. For a purchase, better means lower, so a buy limit deals at its level or at any price beneath it. For a sale, better means higher, so a sell limit deals at its level or at any price above it. The constraint is one sided in both cases. A limit order sets a ceiling for a purchase or a floor for a sale, never a band with an edge on each side, and there is no price on the permitted side of the level that the order refuses.

A deal that happens on the permitted side rather than exactly at the level is conventionally called price improvement. It describes where a fill landed relative to the instruction, and it arises for an ordinary structural reason: a market can arrive at a level in a jump rather than a step, and an order constrained to that level or better takes the price it finds rather than the price it asked for. The same structure means such an order can never deal on the other side of its level, which is the property that distinguishes it from every other instruction in this module.

Which side of the quote the level is measured against 

A quote is two prices, and a limit order is measured against exactly one of them: the price on which its own deal would happen. A purchase deals at the ask, so a buy limit is monitored against the ask. A sale deals at the bid, so a sell limit is monitored against the bid. Nothing about this is a platform setting, and it follows from the direction of the deal alone.

This is the source of the most common complaint about limit orders, and it is not a fault. Most charts are drawn on the bid, so a chart can show price descending through the level of a buy limit while the ask, the only price that order is measured against, remained above it by the width of the spread for the whole of the move. The line on the screen and the price the order is watching are two different series, separated by exactly the gap the first lesson in this module described.

Above the market, below the market 

Because better means lower for a purchase and higher for a sale, the two kinds of limit order sit on opposite sides of the current price, and that position is not a choice. A buy limit sits below the current ask, a sell limit above the current bid. Any other arrangement asks the venue to deal at a price it is already offering, which is a different instruction entirely.

What happens to a limit order entered on the wrong side of the market is a genuine divergence in convention rather than a rule. Some venues reject the order outright and return an error naming the level as invalid. Others treat it as immediately marketable and deal it at once at the best available price, on the reasoning that a purchase capped above the current ask is satisfied by the current ask. The two behaviours produce very different results from the same keystrokes, and which one applies is a property of the venue and the platform, stated in their documentation and not inferable from the order type.

The fill that never happens 

A limit order that is never reached simply continues to exist, which is the ordinary case rather than a malfunction. The harder case, and the one that produces most of the confusion, is an order that is reached and still does not deal. Three mechanisms account for nearly all of it. The level was touched on the other side of the quote, as the section above describes, so the price the order watches never arrived. The level was touched by a single update that reversed immediately, so the market was there for an instant during which no counterparty dealt. Or the level was reached and the size available at it was consumed by orders already resting there, on a venue that fills resting orders in the sequence they arrived.

That third mechanism works differently depending on where the order lives. On an exchange traded book the resting orders at a price form a queue with time priority, so an order entered later deals only once the size in front of it has traded, and a touch that consumes less than the queue leaves the later orders untouched. In a bilateral over the counter arrangement there is no public queue: the firm's own quote reaching the level is what triggers the deal, and the size available there is whatever the firm and its liquidity sources can do at that moment. Traders arriving from exchange markets frequently carry queue reasoning into a context where no queue exists, and traders going the other way assume any touch deals when priority says otherwise.

Partial fills 

The fourth row above is a partial fill, the outcome most often left out of an explanation of limit orders. It happens when the size available at the level is smaller than the size the order asked for. Part of the order deals, the remainder stays where it was written and keeps working at the same level, and the two parts are separate fills at separate moments. The resulting position is smaller than the one specified until, and unless, the remainder deals as well.

Where more than one fill occurs, each carries its own price, and the parts can differ because everything on the permitted side of the level is permitted. Reporting splits two ways. Some platforms aggregate the fills into a single position and report a size weighted average opening price, which is the arithmetic below. Others keep each fill as its own position with its own price, leaving several small positions on the same instrument where one was intended. Cost treatment follows the same split, because a charge levied per deal is levied on each fill.

How long a working order lives 

Until it deals, a limit order is a pending instruction rather than a position. It appears in a separate list from open positions, it can be amended or withdrawn while it waits, and it produces no profit or loss of its own because nothing has been opened. Its lifetime is set when it is written: valid for the trading day and cancelled at the close, valid until a stated date and time, or valid until cancelled, which means it persists across sessions until it deals or is removed.

Two consequences of that pending status are worth stating plainly. Whether a working order reserves any margin before it deals is a firm level convention rather than a universal rule, published in a firm's own terms. And an order that persists across sessions persists across the news, the closes and the openings in between, so an instruction written against one set of conditions can deal weeks later against another. The order does not expire because circumstances changed. It expires when its stated validity ends.

Where practitioners disagree 

What is contested about limit orders is how their cost is counted. One tradition holds that specifying a price removes an uncontrolled variable, since the dealt price can never be worse than the level, and treats the certainty of that boundary as the point of the instruction. Another answers that the missed deal is itself a cost, and an invisible one, because an order that never dealt appears on no statement and in no record, so the tally of what limit orders cost is systematically incomplete. Both descriptions are accurate, and the argument persists because one of the two costs is measurable and the other is not.

A narrower dispute concerns price improvement. One view treats it as a measurable property of an execution arrangement and reports it. Another argues that improvement on a single fill is indistinguishable from ordinary variation in how a market arrives at a level, and that a figure aggregated from it says more about how prices moved during the period measured than about the arrangement. Neither position resolves the other, and both are worth holding whenever a published improvement statistic is read, on any firm's materials.