Skip to main content

Bid/Ask Stacking

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,211 words

Bid/ask stacking is the build-up of resting limit-order liquidity across multiple price levels on one side of the order book (the Depth of Market, or DOM). When many large bids "stack" below the current price, more aggressive sell volume is required to push price down through them — so stacked liquidity is read as a wall of support; stacked asks above price are read as resistance. The concept is a staple of order-flow and DOM scalping, but it carries a permanent, unavoidable tension: a resting limit order is a promise of intent, not a contract, and in modern electronic markets the overwhelming majority of those promises are cancelled before they ever trade. Stacking is therefore a real, observable phenomenon whose informational content is genuinely contested.

How it's formed

The DOM (or Level 2 / "ladder") shows the quantity of resting limit orders queued at each price level — typically the best bid/ask plus the next several levels of depth (the twofox Sierra Chart study, for example, monitors up to the first 15 levels per side). "Stacking" is simply liquidity being added at successive levels: the displayed size at one or more price levels grows over time. Its opposite is pulling — liquidity being withdrawn (cancelled) ahead of price. The standard order-flow framing:

  • Stacking ahead of price = more market orders are needed to move price a given distance → the move may be slowing or reaching exhaustion.
  • Pulling ahead of price = fewer aggressive orders needed to move price → liquidity is getting out of the way, often preceding continuation.

A related, distinct measure is order-book imbalance: the ratio of total bid depth to total ask depth (e.g. bid_size / (bid_size + ask_size) over the top N levels). Stacking is the dynamic (depth growing over time) view; imbalance is the static snapshot. See the sibling node on order-book imbalance.

How it's used in practice

DOM and order-flow traders — almost exclusively very-short-horizon scalpers in liquid futures (ES, NQ, CL, Treasuries) and some crypto — use stacking three ways:

1. Locating intraday support/resistance. A large, persistent stack at a round number or prior level marks where price is likely to stall. Traders may lean on it (place entries just in front) or fade moves into it. 2. Reading stack behaviour as the move develops. The diagnostic value is in what the stack does when price reaches it. If a stacked bid wall absorbs incoming sell market orders without breaking (size refreshes as it's hit), that is genuine demand — bullish. If it pulls (cancels) as price approaches, the support was illusory and price often accelerates through the vacuum. 3. Stack exhaustion. Watching a defended level's size tick down as aggressive orders eat it; when the queue empties, the level breaks. This is the highest-quality use because it requires the orders to actually trade, not merely be displayed.

The reliable signal is always resting orders being consumed (absorption/exhaustion, visible on a time-and-sales or footprint chart), not resting orders merely being displayed.

Adoption, debate & evidence

Stacking is widely used in the discretionary DOM/order-flow community (Bookmap, Sierra Chart, Jigsaw, ATAS) and is a building block in quant order-book-imbalance signals. But its core premise — that displayed depth is informative — is empirically fragile:

  • Cancellation rates are near-total. Hasbrouck and Saar document that a large fraction of orders in the millisecond environment are cancelled almost immediately ("fleeting orders"), closely tied to HFT activity (their October-2004 INET data showed roughly a third of limit orders cancelled within two seconds). More recent studies commonly cite eventual cancellation of around 95–96% of limit orders in US equity markets — i.e. a displayed stack is, statistically, far more likely to be cancelled than filled. (As a benchmark, U.S. v. Coscia cited the defendant's firm cancelling over 95% of the orders it placed.)
  • Displayed depth is not total depth. Iceberg / reserve orders show only a sliver of true size and are explicitly permitted by the CME, CFTC and DOJ. A thin-looking level may hide enormous real liquidity; a thick-looking one may be hollow.
  • Spoofing and layering are the deliberate weaponisation of stacking. Spoofing — bidding/offering with intent to cancel before execution — was made illegal by the 2010 Dodd-Frank Act (and CME Rule 575). Layering places multiple non-bona-fide orders across several levels precisely to manufacture a fake stack and induce other participants to react. Enforcement is real: U.S. v. Coscia (2015) was the first criminal spoofing conviction; Navinder Sarao used "dynamic layering" of S&P 500 e-mini orders implicated in the 2010 Flash Crash; the DOJ announced a ~$5.56M BofA Securities settlement in September 2025 over Treasury-market manipulation in which traders used "iceberg" cover while fully displaying spoof orders on the opposite side (one trader allegedly inflated perceived liquidity by ~2,600%, per DOJ/press reporting). The existence of these cases is direct evidence that stacked displayed depth routinely lies.

Net: stacking is a genuine market microstructure phenomenon, but displayed stacking has weak and easily-faked standalone predictive value. Its informativeness rises sharply when corroborated by execution (absorption/exhaustion on the tape) and falls toward noise when read as a static picture of "walls."

Strengths & limitations

Works when: the timeframe is seconds-to-minutes in a deep, centralised limit-order-book market; the trader confirms stacks by watching whether they absorb or pull under live trading; and stacking is one input among tape and price action.

Fails when: treated as a reliable static map of support/resistance ("there's a big wall, it'll hold"). Spoofers exploit exactly that belief. It also degrades in fragmented or off-exchange markets (US equities route much volume to dark pools and internalisers, so the lit book understates true depth) and is largely meaningless on delayed or aggregated retail feeds.

The single most common misuse: trading off the displayed size alone — entering because a large bid "appears," only to watch it vanish (pull) the instant price arrives, leaving the trader long into a vacuum. Displayed depth is intent, not commitment; the only depth that counts is the depth that fills.

Sources

Disputes flagged: the predictive value of displayed (vs executed) stacking is genuinely contested; academic microstructure work treats unfiltered displayed depth as noisy and manipulable.