What an Order Type Gives Away

An order type is a choice about which of two things you leave uncertain, and there is no third option. A market order fixes the fill and leaves the price open. A limit order fixes the price and leaves the fill open. The limit looks like the free one, because it collects the spread instead of paying it, and lesson 3 already said why it is not: the fills you get are the ones somebody else chose to give you. Measured on this course’s sixty closes, a buy limit posted eight tenths of a point under the close fills 16 times in 56 chances and banks that eight tenths every time. Valued three bars on it still averages a loss of 0.375 a share, while simply buying at the close averages a gain of 0.373. Take the series’ drift out and the split is starker: buying at the close comes to −0.020 and posting to −0.768. The discount is real, it is eight tenths, and what it selects for costs more than it pays.

Prerequisites: Lesson 3, which established that a fill is one counterparty choosing that moment to trade against you, and that this makes your fills a biased sample rather than a random one, lesson 11, for what a market order costs once your size is larger than what rests in front of it, and lesson 55, for the resting quote as an option you have written, which is the same object a limit order is.

And the third thing an order gives away

Everything above is about price against fill. There is a third quantity, and it is the one the lesson is named for: a displayed order tells everyone your price and your size. Lesson 57 explained why that matters to a desk working a schedule — an order that is 205 shares a minute for twelve sessions cannot afford to announce itself — and the platform’s answer is the display rule. An iceberg shows a round lot and keeps the rest in reserve, refilling as the visible part fills.

What it gives away in exchange is queue position. On most venues the hidden part loses its place when the displayed part refreshes, so the same total size sits further back and fills less often. That is the same currency as the first table, paid in a different denomination: the iceberg buys concealment with fill probability, and the fills it does get are selected the same way every passive fill is. It does not escape the table above; it moves down it.

The honest summary of the whole page is one sentence with three terms in it. Price, fill, and disclosure: you can fix any one of them, you can usually fix a second by paying for it in the third, and you cannot fix all three. A market order takes none of the price certainty and all of the fill certainty, and discloses nothing until it prints. A displayed limit fixes the price, leaves the fill open and tells everyone. An iceberg fixes the price, leaves the fill more open than that, and tells them less. Read the platform’s menu as positions in that space rather than as a list of features, and it stops being a menu of tricks.

For a retail order none of this is close to the largest thing in the account. On a series like this one the passive penalty at three bars is 0.75 of a point on a series that trades near a hundred, which is three quarters of one per cent, and lesson 12’s ratio would already have told you whether that is large next to what your instrument moves in a day. The reason to know it is narrower and more useful: it explains why a limit order that fills feels like a good entry and measures out as a bad one, and it is the same reason lesson 3 gave, now with a number on it.

What this does not settle

That closes are enough to decide whether a limit filled. They are not. A real order fills on any trade through its price, so a bar that dipped below the limit intrabar and closed above it counts as a miss here and would have been a fill in the market. The true fill rate is higher than 16 in 56, and the fills the test misses are the ones where the price came back, which are the good ones. The correction runs against the finding, and its size is not knowable from a close-only series. Every figure above should be read as the effect measured on daily closes, not as the effect on a real book.

That 56 observations settle anything. Lesson 19’s arithmetic says they do not, and the five distances in the table are five views of the same sixty numbers rather than five independent tests. What makes the result worth printing is not the sample but the sign: it is negative in all five rows, in all four horizons past one bar, and it gets worse rather than better when the drift is removed. That is a consistent pattern in one series, which is a reason to go and measure your own, not a settled fact.

That three bars is the right horizon. It is a choice, and it is the choice that makes the effect visible; at one bar the comparison reverses. If you hold for minutes rather than days the one-bar row is the one that applies to you, and on this series it says the opposite of the headline. The lesson’s finding is about holding periods long enough for the selection to show, and the crossover on this series is between the first and second bar.

That the test contains a spread. It does not, and this is the largest omission on the page. A real passive order also earns the half-spread the taker pays, and on a venue with a maker rebate it earns lesson 54’s 0.0030 a share as well. The 0.80 discount here stands in for that and is far larger than any real spread on a hundred-point instrument, so the test is generous to posting on price and still finds against it on selection. Charge the taker a realistic half-spread and the comparison narrows; whether it reverses depends on the ratio lesson 12 measured, the spread against what the instrument moves in a bar.

That the sell side agrees. It does not, on this series, and the reason is that a series rising 5.76 per cent makes every short lose and drowns the effect. A test that only works in one direction is a test that has not separated the effect from the trend, which is why the detrended column exists and why it is the column doing the work. A reader who repeats this should run both sides on a series with no drift in it before believing any of it.

And the concession that costs this lesson most: it has measured the cost of a passive fill and has not measured the cost of not trading. The 40 bars where the limit missed are recorded here as opportunities the taker got, and in a real account they are also 40 occasions on which no capital was committed, no risk was carried and no commission was paid. A policy that trades a third as often is not simply a worse version of one that trades always; it is a different exposure, and comparing their per-trade averages hides that. The right comparison is per unit of time and per unit of risk carried, which needs a position-sizing rule this page has not specified. Everything above is true of the fills. Whether it is true of the strategy is a question this lesson has not asked.

Problems

1. Run the two policies on your own instrument. Take two hundred bars at whatever timeframe you trade. For each, record the close, whether the next bar traded through a limit a fixed distance below it, and the close three bars later. Average the outcome for the fills and for the misses. Pick the distance to be about the size of your instrument’s median bar move, which is the second problem below. An hour with a spreadsheet, and it tells you whether the sign of this lesson holds where you trade.
2. Measure your instrument’s median bar move and its worst one. Take the absolute change from bar to bar over two hundred bars, sort them, and write down the median and the maximum. On this course’s series those are 0.90 and 4.10, a ratio of four and a half. That ratio is the whole stop-limit problem: it is how wide the band has to be to survive the worst bar, expressed in units of an ordinary one. Ten minutes, and it is the number to set any band against.
3. Find your own crossover. Repeat the first problem at one, two, five and ten bars and find the horizon where posting stops winning. On this series it is between one and two. If yours is at ten, passive orders are close to free for you; if it is at one, they are not, and the guide that told you to always use limit orders was written for somebody else’s holding period. An afternoon, and it settles the order-type question for your own trading rather than in general.

Sources. Thomas Copeland and Dan Galai, “Information Effects on the Bid-Ask Spread” (The Journal of Finance, 1983), for the quote as a written option, which is where lesson 55’s framing and this lesson’s come from. Puneet Handa and Robert Schwartz, “Limit Order Trading” (The Journal of Finance, 1996), for the two sides of the passive trade set against each other: the spread earned and the adverse selection borne. Lawrence Harris and Joel Hasbrouck, “Market vs. Limit Orders” (Journal of Financial and Quantitative Analysis, 1996), for the same comparison run on real order data rather than on sixty closes, which is the study this page is a small imitation of. U.S. Securities and Exchange Commission, investor bulletin on stop and stop-limit orders, for the regulator’s own account of what happens to stop orders in a fast market, written after a morning when a great many of them filled a long way from their triggers.

There are two orders and everything else is a rule attached to one of them. The market order leaves the price open; the limit order leaves the fill open; nothing leaves neither open. The limit’s discount is real and the selection behind it is larger: eight tenths collected on this course’s series, against a difference of 2.17 between the bars it caught and the bars it missed, and a detrended average of −0.768 against −0.020 for simply trading. It reverses at a one-bar horizon and nowhere past it. The stop is the same trade with a trigger on it, and no band both protects the price and reliably fills, because 22 of 59 bars here moved further in one bar than a one-point band. And a displayed order gives away a third thing, which is what you are doing, at the price of a queue position. Lesson 59 takes the one cost this page kept deferring — what your own trading does to the price while you do it — and puts a number on it.

Educational only. Trading involves substantial risk of loss. Not financial advice. Past performance does not guarantee future results.