Market vs. Limit Orders

Every order you send is one of two fundamentally different objects. A market order demands immediate execution at whatever price the book offers; a limit order offers to trade at a price you name, and waits. Choosing between them is the atom of execution, the tradeoff of immediacy versus price that every larger algorithm is built out of.

The tradeoff, stated

• A market order guarantees the fill but not the price: you cross the spread and pay (at least) the half-spread $s/2$ relative to the mid, plus walk-up if your size exceeds the touch. You are a liquidity taker.
• A limit order guarantees the price (or better) but not the fill: you rest in the Order Book Mechanics queue and may earn the spread, but you might never trade, and when you do trade it is disproportionately when the market is moving against you. You are a liquidity maker.

The right choice depends on how much you value certainty of execution versus price improvement, and on how informed you fear the flow on the other side is.

A limit order is an option you write for free

The deep insight (Copeland & Galai, 1983) is that a resting limit order is a short option position. Post a limit sell at price $a$ above the mid. You have written the market a call: anyone may "exercise" by lifting your offer. They will do so precisely when it is advantageous to them, i.e., when the true value has risen above $a$. So you get exercised (filled) mostly in the states where you'd rather not have sold. This is adverse selection, and it is the cost that offsets the spread you hoped to earn. The limit order's payoff is the classic written-option profile: a small premium (the spread) most of the time, and a loss in the tail when you're picked off. See Adverse Selection.

A decision model

Suppose you want to buy one unit; the mid is $m$ and the half-spread is $s/2$. Compare two policies.

Market order. Immediate, certain cost relative to the mid:

$C_{\text{mkt}} = \frac{s}{2}.$

Limit order at the bid. With probability $\pi$ you fill and capture the half-spread; with probability $1 - \pi$ you do not fill and must eventually cross anyway (or miss the trade), and conditional on filling you suffer expected adverse selection $\alpha$ (the mid tends to have dropped when your buy fills). A stylized expected cost:

$C_{\text{lim}} = \pi\Big(\underbrace{-\tfrac{s}{2}}_{\text{spread earned}} + \underbrace{\alpha}_{\text{adverse selection}}\Big) + (1-\pi)\,\underbrace{\big(\tfrac{s}{2} + \kappa\big)}_{\text{chase / opportunity cost}}.$

Here $\kappa$ is the penalty for non-execution, the adverse price drift while you waited plus any missed alpha. The limit order wins when the fill probability $\pi$ is high, the adverse selection $\alpha$ is small, and your urgency $\kappa$ is low. It loses when you must trade (high $\kappa$) or when you are trading against sharper counterparties (high $\alpha$). This is exactly why urgent, alpha-driven orders lean toward market orders while patient, liquidity-providing flow leans toward limits.

The signing of the tradeoff

Rearranging, post a limit order rather than crossing when the expected saving beats the expected penalty:

$\pi\,\Big(s - \alpha\Big) \;>\; (1-\pi)\,\kappa.$

Read off the comparative statics: wider spreads $s$ favor posting (more to earn), higher adverse selection $\alpha$ or urgency $\kappa$ favor crossing, and a fill probability $\pi$ near zero makes limit orders pointless. A market maker with no directional view and low urgency is on the "post" side of this inequality essentially always, that is the The Avellaneda-Stoikov Model problem. A momentum trader who must get in before the move completes is on the "cross" side.

Worked example

A stock trades $50.00 / $50.02, so s = \0.02$and$m = $50.01$. You want to buy 1,000 shares.

• Market order: lift the offer at $50.02, cost s/2 = \0.01 per share versus mid, i.e. \10 total, filled now.
• Limit at $50.00: if you fill you earn $0.01 versus the mid. Suppose historically $\pi = 60\%$ of such orders fill within your horizon, conditional adverse selection is \alpha = \0.006 (the mid drifts down \0.006 by the time you're hit), and the non-fill chase penalty is \kappa = \0.015$. Then$C_{\text{lim}} = 0.6(-0.01 + 0.006) + 0.4(0.01 + 0.015) = 0.6(-0.004) + 0.4(0.025) = -0.0024 + 0.010 = $0.0076 per share, slightly cheaper than the \0.01 market order, but with variance and execution risk. If your urgency $\kappa$ were higher (say $0.05 because the stock is running), the limit order's expected cost jumps and crossing wins.

Order-type nuances

Real venues expose more than two order types, each a point on the immediacy–price spectrum: marketable limit (a limit priced through the book, immediacy with a worst-case price cap), immediate-or-cancel (IOC) and fill-or-kill (FOK) (take only what's available now), post-only (reject if it would cross, guaranteeing maker status and the rebate), iceberg (hide size), and pegged (track the mid or touch). Each is a way to tune the same tradeoff.

Failure modes and caveats

• Non-execution is a real cost. Backtests that assume limit orders always fill at the posted price are fantasy; unfilled limits leave you exposed to exactly the moves you wanted to trade. Model $\pi < 1$.
• Adverse selection dominates in fast markets. Passive fills cluster at the worst moments; measuring realized spread net of the post-trade drift (the Bid-Ask Spread Decomposition) reveals whether your "spread capture" is real.
• Queue jumping and latency. In a one-tick market, whether your limit fills is mostly about queue position, which is a Latency & High-Frequency Trading game.
• Signaling. A large limit order displays your intention and can be front-run or faded; this is why big orders are sliced by execution algorithms rather than posted whole.

In interviews

The canonical question is "when would you use a limit order versus a market order?", and the answer that separates candidates is framing the limit order as a written option whose premium is the spread and whose cost is adverse selection plus non-execution. Be ready to write the decision inequality $\pi(s - \alpha) > (1-\pi)\kappa$ and reason about its comparative statics: urgency and informed counterparties push you to cross, wide spreads and patience push you to post. A good follow-up connects to execution algorithms, slicing a parent order is precisely managing this tradeoff repeatedly over time, which is Optimal Execution.