Part II · Markets · Chapter 2C

The order book

Making and taking liquidity

Strategies fall into two execution families with fundamentally different profit-and-loss characteristics. Making adds liquidity to the order book: it earns the spread and pays low or even negative fees. Taking removes liquidity: it pays the spread and higher fees. The same predictive model can be wrapped in either execution layer, with very different results. Understanding the mechanics of the order book is what separates retail traders who lose 50 basis points per round trip to slippage from professionals who lose 5.

1 · The order book

The order book is the live record of supply (asks) and demand (bids) for a security, organized in price levels. The best bid is the highest price any buyer is willing to pay at that moment. The best ask, or offer, is the lowest price any seller is willing to accept at that moment. The gap between them is the bid-ask spread, and its midpoint is the mid price. Writing $b$ for the best bid and $a$ for the best ask:

Here is a small book, with each level's price and the number of contracts resting there:

ASKS (sellers)
$118 @ 12   ← deeper ask levels (worse prices, more quantity)
$117 @ 8
$116 @ 5
$115 @ 3    ← best ask (top of book)
        ____________________  spread = $2  mid = $114.0
$113 @ 5    ← best bid (top of book)
$112 @ 8
$111 @ 10
$110 @ 15   ← deeper bid levels (worse prices, more quantity)
BIDS (buyers)

The best ask is $115 and the best bid $113, so by (2C.1) the spread is $2 and the mid $114. The levels nearest the mid, the top of book, carry the best prices and the least quantity. When a new order arrives, one of three things happens:

1. A limit order priced away from the book goes onto the book as a resting order. It becomes part of the bid or the ask at its price level, and the trader is adding liquidity.
2. A limit order priced to cross the book takes liquidity. A buy limit at or above the best ask matches against the resting ask, and a sell limit at or below the best bid matches against the resting bid.
3. A market order fills immediately at whatever prices are available on the opposite side. It walks through the book as deep as necessary to fill the requested quantity.

How deep the book is

Every liquid security has two queues of resting orders: bids, from buyers willing to take the security at one price, and asks, from sellers willing to give it up at another. The highest bid and the lowest ask across all venues form the National Best Bid and Offer (NBBO), the regulated reference price for American equities under Regulation NMS (Chapter 2B). Behind them sit deeper levels with worse prices but more quantity available.

For SPY on a typical day, the top of the book is millions of shares deep. The bids and the asks at the NBBO might each carry 500,000 to 2 million share-equivalents of resting orders across all venues. Behind them, levels 1 cent away in each direction carry similar size. The book is unusually thick relative to most stocks because SPY is the most-traded ETF in the world.

For SPX index options, the book is much thinner per strike. Each strike has its own order book, and there are thousands of strikes across hundreds of expiries. At-the-money 0DTE strikes on a typical day might have 500 to 2,000 contracts at the NBBO, and deep out-of-the-money strikes 5 to 50. The spreads are wider in absolute dollar terms, often $0.10 to $0.50 on liquid SPX strikes against $0.01 to $0.02 on SPY. The per-contract notional is also 10 times larger, so the percentage spread is comparable.

The percentage spread is the spread divided by the mid, $s/m$. A hypothetical SPY option quoted with a $0.02 spread around a $5.00 mid has $0.02/5.00 = 0.4\%$. The matching SPX option, ten times the size, might be quoted with a $0.20 spread around a $50.00 mid, which is also 0.4%.

Which book your order meets

For American equity options and shares, retail orders predominantly do not touch a public exchange book directly. They are routed to a wholesaler, such as Citadel Securities, Susquehanna, Virtu, Wolverine, Jane Street or Hudson River Trading, which fills the order from its own principal book. The wholesaler returns a small fraction of the captured spread to the broker as payment for order flow (PFOF). For SPX and XSP there is no wholesaler bypass: orders route directly to the Cboe order book (Chapter 2B).

Two consequences flow from this routing. First, SPY and QQQ retail flow is largely internalized, so the gamma exposure (GEX) computed from public open interest (OI) on these names double-counts what wholesalers cross internally. Second, SPX and XSP flow hits a single transparent book. That is one of several reasons the dealer-mechanics chapters that follow use SPX as their laboratory rather than SPY.

2 · The bid-ask bounce

The "last trade" price alternates between the bid and the ask, as trades alternate between buy-driven and sell-driven flow. This produces a sawtooth pattern called the bid-ask bounce, which adds noise to any series of last prices. Professional quantitative work uses the mid price instead, because the mid is a less biased estimator of the security's true value.

For SPY, where the bid-ask spread is typically $0.01 to $0.02, the bounce is small. SPX, the cash index, has no bounce at all, because it is not directly tradable: its level is computed from the constituent stocks. For SPX options, where spreads can be $0.10 to $0.30 wide, the bounce dominates the high-frequency dynamics. Any backtest that uses "last" instead of "mid" prices on them is systematically biased.

3 · Walking the book

A large market order that exceeds the quantity available at the top level walks the book. Its fills are spread across several levels, and each successive contract fills at a worse price. For a buy of $Q$ contracts that takes $q_j$ contracts at price $p_j$ on each level $j$, the average fill and the slippage against the best ask $a$ are

The figure uses the book above. The slider sets the size of a market buy, from 1 to 20 contracts, starting at 3. The order takes the best ask first and then each next level up until it is filled, and the amber part of each bar shows what it took. Readouts give the number of levels eaten, the average fill price, the slippage against the $115 best ask and any quantity left unfilled. Start at 3, then drag the slider up through 8 and on to 20.

market buy

3 contracts

levels eaten = 1

avg fill price = $115.00

slippage = +$0.00

unfilled = —

At 3 contracts the whole order fills at $115: one level eaten, no slippage. At 8 it has emptied the second level too, for an average of $115.63. At 20 it eats all four levels and averages $116.65, $1.65 of slippage on each contract. The average climbs in steps, because each new level is a dollar worse than the last, and the book never leaves any of the 20 unfilled.

For a small retail SPY market buy of 100 shares against a top-of-book ask of 1,000 shares, the entire order fills at the best ask with no slippage. For a 100,000-share market order, the order walks several levels, and the average fill price might be 5 cents worse than the top-of-book ask. That is about 0.9 basis points of slippage on a $580 stock, since $0.05/580 \times 10{,}000 = 0.86$. For a 10-million-share market order, the order might move the price by several dollars before completion. These examples assume a thin top of book, 1,000 shares at the best ask. At the depth quoted above for SPY, 500,000 to 2 million shares at the NBBO, a 100,000-share order would fill at the top.

This is why institutional-scale execution is its own specialty. Large orders are broken into many small pieces and executed over time, using algorithms that trade off market impact against opportunity cost:

• VWAP (volume-weighted average price) algorithms spread an order in proportion to the market's own volume through the day. Their benchmark is the day's average trade price weighted by volume, $\sum_k p_k v_k / \sum_k v_k$ over trades at price $p_k$ and volume $v_k$.
• TWAP (time-weighted average price) algorithms spread an order evenly over time.
• Implementation-shortfall optimizers minimize the gap between the price when the decision was made and the average price actually achieved, counting the cost of any part left unfilled.

The optimal execution problem was formalized by Almgren and Chriss (2000) and remains an active research area. Chapter 2D gives the rule of thumb for how impact grows with order size, the square-root law.

4 · Maker-taker fees

American exchanges operate a maker-taker fee model for most equity and options trading:

• Makers, resting orders that get filled when someone crosses them, typically receive a small rebate: they are paid to provide liquidity.
• Takers, incoming orders that cross the spread to fill against resting orders, pay a fee: they are paying for immediate execution.

For American equity exchanges, the maker rebate is typically 0.2 to 0.3 cents per share and the taker fee typically 0.3 to 0.4 cents per share. The exchange keeps the small difference. On 100 shares of SPY traded round trip as a taker on both legs, that is roughly $0.60 of exchange fees: small but real.

For American options, the maker-taker structure is more variable across exchanges and contract types. SPX options on Cboe have a different fee schedule from SPY options, which trade across multiple competing venues. Institutional traders care deeply about these differences. Retail traders rarely see them, because their broker (Robinhood, Fidelity, Schwab) routes orders through wholesalers that internalize the fee structure (Chapter 2B, on payment for order flow).

Inverted markets

A small number of American equity exchanges run inverted markets, also called subsidy markets: they pay takers and charge makers. The economic rationale is that traders who want immediate execution badly enough are willing to pay for it explicitly, while liquidity is harder to attract. Inverted markets exist at small share, but they provide a price-improvement venue for orders that absolutely must fill now.

5 · Queue position

When you place a limit order at a price level that already has resting orders, you go to the back of the queue at that price. When a market order arrives and consumes some of the resting quantity, it fills against the front of the queue first. The order at position 50 in a queue of 100 contracts has to wait for 49 fills ahead of it before it fills.

Queue position matters enormously for market makers, who rely on being at the front of the queue to capture the most order flow. Latency arbitrage, the ability to update quotes faster than competitors, is largely about queue-position management. A market maker that can see incoming flow microseconds before its competitors can adjust its resting orders accordingly, capturing or avoiding the resulting fills.

For retail traders, queue position is invisible. Retail orders are routed to wholesalers rather than directly to exchanges, and wholesalers internalize execution against their own principal books. But understanding that queue position exists matters. A limit order at the NBBO might not fill even if the price touches your level, because 10,000 other contracts were in the queue ahead of you.

6 · Adverse selection

Making has a subtle and important risk. When your resting order fills, the trade was almost certainly initiated by someone with marginally more information than the market consensus. If the price was about to move up and you had a resting ask, your ask gets filled and the price immediately moves above your fill: you sold too cheaply. If the price was about to move down and you had a resting bid, your bid gets filled and the price immediately moves below your fill: you bought too high. This is adverse selection, the systematic tendency of your fills to come from informed counterparties at exactly the worst moments.

Adverse selection is the central risk that market makers manage. The bid-ask spread exists, fundamentally, to compensate the market maker for it. If spreads were too narrow, market makers would systematically lose to informed flow. If spreads were too wide, no one would trade and the market maker would earn nothing. The equilibrium spread is the one where the spread captured exactly compensates for adverse-selection losses, plus a margin.

In general, if the informed traders know of a move of size $\delta$, the same balance, $(1 - \pi)\,h = \pi\,(\delta - h)$, gives the break-even half-spread:

The spread widens with both the share of informed traders and the size of what they know. A market maker's costs and profit margin sit on top, so a quoted half-spread runs a little above $\pi\,\delta$. A retail limit order resting at the bid is a small market maker without the spread's protection: it fills most readily when the price is about to fall through it.

7 · Order types

American brokerages typically expose five order types to retail traders, and those five cover about 95% of retail decisions. Three of them cover about 90%. Each has its own contract: what it promises, what it does not, and when it is the right tool. Knowing them is one of the cheapest sources of slippage reduction available.

• Market order: buy at the best available ask, or sell at the best available bid. It crosses the spread immediately and promises a fill but no price: you fill at whatever price is available. In thin markets, at the open or in stressed conditions, it can fill far from the NBBO. Use it cautiously, and only for tiny orders in highly liquid names.
• Limit order: buy at or below a stated price, or sell at or above it. It promises a price but no fill: the order sits passive on the book until it is matched or canceled. Use it when price matters more than execution.
• Marketable limit order: a limit order set at or through the current NBBO. It behaves like a market order in normal conditions, filling at the NBBO, but caps slippage in stress, because it never fills beyond its limit. It is the recommended default for any retail order of meaningful size.
• Immediate-or-cancel order (IOC): fill whatever portion you can at the limit price right now, and cancel the rest. It is useful when you want to take some liquidity without leaving a resting order behind.
• Market-on-close order (MOC): execute at the official closing print, which the closing auction determines. An MOC order is certain to fill at the close, but the auction's imbalance sets the close price itself.

The recommended discipline is simple: use marketable limits as the default. They get you filled in normal conditions and protect you from blow-up fills in stress. A marketable limit set at the offer is essentially identical to a market order on normal days. It protects you from a fat-finger blowup, the fill far from any sensible price that occurs when an order arrives during a quote dropout, a fast market or a circuit-breaker scenario. The cost of using marketable limits as the default is approximately zero. The benefit is that on the day something genuinely goes wrong with the quote infrastructure, you do not fill 200 points away from the NBBO. Always use marketable limits. Save market orders for tiny orders in highly liquid names, where slippage is bound to be negligible.

8 · The closing auction

Between 3:50 PM and 4:00 PM ET, American equity exchanges run their closing auction. Instead of matching orders one at a time, the auction gathers them and fills them all at a single price. Starting around 3:50, the exchanges publish imbalances every few seconds, such as "12 million shares of SPY to buy at the close, market-on-close". Liquidity providers then know to step in if they want to take the other side. The final 10 seconds of trading set the official close. That print is used by MOC orders and by the calculation of an ETF's net asset value (NAV), its holdings' worth per share. Index benchmark closing levels use it, including S&P's official daily index level, and so does most institutional performance measurement.

Estimates of the auction's size vary with the year and the measure. One puts it at approximately 10% to 15% of daily NYSE volume. A time series shows it growing with passive index funds:

• 2010: about 5% to 7% of NYSE volume.
• 2020: about 10% to 12%.
• 2025: estimates ranging from 12% to 18% of total NYSE consolidated-tape volume on typical days, higher on rebalance dates.

For index ETFs such as SPY, QQQ and IWM, the share is higher, closer to 20% to 25%, because passive index funds rebalance to the close. NYSE's own reporting through 2024 documented that closing auctions had reached new highs, both in absolute daily traded shares and notional and in their share of consolidated American equity volume.

The microstructure consequence is that the end of the day has a different character from the rest of it. The last 15 to 30 minutes are dominated by mechanical, non-fundamental order flow. This is where dealer hedging concentrates, where index funds rebalance, where pension funds size positions to specific weights and where volatility-targeting strategies execute their daily risk adjustments.

This mechanical flow is also where dealers' end-of-day hedging happens. Two papers in the Journal of Financial Economics measure intraday momentum, the tendency of the last half-hour to continue the day's move. Gao, Han, Li and Zhou (2018) found in SPY, from 1993 to 2013, that the first half-hour's return predicts the last half-hour's return. Baltussen, Da, Lammers and Martens (2021) used a different predictor, the return over the rest of the day, from the previous close to the last half-hour. They found the pattern, with statistical significance, across more than 60 futures contracts around the world from 1974 to 2020. Chapter 6C examines their paper in detail. Only the second paper ties the pattern to hedging, by option dealers and by leveraged funds that are short gamma. Gao and co-authors proposed other causes: late-informed trading and infrequent rebalancing.

In the Dealer Mechanics Lab's account, the mechanism is the same in every market. Dealers carry net option exposure through the day, accumulate hedging needs as it progresses and concentrate those hedging trades into the close. The composition of order flow in the last half-hour differs, so its statistics differ. Part VI picks the story up from here. Chapter 6C sets out the limits of the evidence.

For retail traders the implication is operational. Avoid placing market orders in the last 15 minutes of the trading day unless you have thought through what is happening in the auction. Spreads can widen, the price can swing on imbalance updates, and the closing print itself may be substantially different from the level you saw 30 seconds before 4:00.

9 · Trading outside regular hours

As of late 2025 and early 2026, several American brokers, among them Robinhood, Schwab and Interactive Brokers, have launched 24/5 equity trading. These extended-hours operations allow trades in major equities, including SPY and QQQ, outside the 9:30 to 4:00 session, through partnerships with Blue Ocean ATS and similar overnight venues.

Liquidity in 24/5 sessions is thin compared with regular trading hours. Spreads are wider, often by 5 to 20 times the regular-hours spread. Data disclosed by Citadel Securities suggest that post-market activity is rising structurally, to about 2% of notional and about 6% of shares traded for retail. It remains a small fraction of total volume. For most retail strategies, 24/5 trading is not yet operationally relevant. For traders who need to respond to overnight news, it provides a path to execute that did not exist in 2020.

The structural risk is that a market order placed during 24/5 hours can fill at a price far from where the underlying opens in the regular session. The bid-ask spread is wider and the depth thinner. The participants on the other side are different, too: often professional liquidity providers willing to quote only at prices unfavorable to retail. Use only limit orders during extended hours. Better still, as a rule, do not trade during extended hours unless you have a specific reason.

10 · Check yourself