Technical Analysis for Futures Traders
Part III, Institutional Chapter 12

Liquidity: Pools, Sweeps, and Imbalances

Stops do not just sit there. Someone is hunting them. The map of where they sit is the map of where price will travel.

12.0Why this chapter exists

Chapter 3 introduced liquidity as a primer. This chapter treats it with full institutional rigor: a complete taxonomy of pools, the canonical sweep mechanics by pool type, the FVG (fair-value gap) retest framework, and the regime-conditioning rules that distinguish reversal sweeps from continuation sweeps.

Liquidity-based reading is the most internally consistent framework in this book for explaining why price travels where it does. The price tape is, mechanically, a sequence of orders being matched at the tape. The orders that drive the tape are split between aggressive flow (who took out the resting liquidity?) and resting flow (where was the liquidity sitting?). The aggressive side is what Chapter 11 (Order Flow) addresses. The resting side, the pools where liquidity accumulated and is harvested, is this chapter.

We treat:

  1. The full pool taxonomy: equal H/L, PDH/PDL, prior-week extremes, round numbers, IB extremes, VAH/VAL, naked POCs.
  2. The sweep mechanics for each pool type: pierce magnitude, close-back-inside requirement, velocity gate.
  3. The FVG framework: retest probability, conditioning by impulse strength, regime sensitivity.
  4. The regime-conditioning rule: sweeps in Range = reversal; sweeps in Trend = continuation.
  5. Multi-pool harvest sequences: when one sweep cascades into the next.
  6. Failure modes specific to liquidity-based reading.

This chapter is one of the more practitioner-focused. The frameworks become operational immediately.

12.1The pool taxonomy

A liquidity pool is a price region where stop-loss orders, breakout orders, or other "if-touched" orders accumulate. Pools are invisible in the order book (most are stops, which appear only when triggered) but predictable in location.

Pool Type 1: Equal Highs / Equal Lows

When the chart prints two or more swing highs (or lows) at approximately the same price within tolerance, an equal high pool (or equal low pool) forms above (or below) the level.

  • Tolerance: 0.10 × ATR (volatility-aware proximity).
  • Stop density: highest among pool types; every short who entered at the first high parked a stop just above the second.
  • Reaction profile: very high probability of being swept. The question is not whether but when.
  • Defense profile: moderate. The participants who placed the stops are also the ones who would defend the level if the sweep failed.

Equal highs are flagged liquidity. When price approaches them, the trade decision is set up by the question: will the sweep occur, and what happens after?

Pool Type 2: Prior Day Extremes (PDH/PDL)

The high and low of the prior trading session, computed RTH-only on the institutional convention.

  • Stop density: high. PDH and PDL are tracked by every desk and almost every retail platform; trail stops, breakout stops, and protective stops cluster.
  • Reaction profile: ~60 to 70% sweep-then-react probability on liquid futures.
  • Decay profile: slow. PDH retains relevance for several sessions; once breached cleanly, becomes a former-resistance-now-support reference.

PDH and PDL are the most-marked horizontal levels on any desk's chart. The author's Trend & Levels indicator weights them at base 2.0 (per Chapter 5).

Pool Type 3: Prior Week / Month Extremes

PWH, PWL, PMH, PML, the high and low of the prior week and prior month respectively.

  • Stop density: very high but spread out. Multi-session traders, swing traders, and longer-horizon participants all reference these.
  • Reaction profile: strong; weekly extremes in particular act as durable magnets and durable resistance.
  • Time-frame relevance: dominant for swing decisions, contributory for intraday.

Pool Type 4: Round Numbers

Prices ending in 00 (or 50 on lower-priced contracts).

  • Stop density: moderate. Discretionary traders mentally round their stops to round numbers; algorithmic execution stratifies child orders at round-number increments.
  • Reaction profile: weak standalone (round numbers alone do not produce reliable reactions on liquid futures). Strong in confluence with other pool types.

Round numbers earn 0.5 base weight per Chapter 5; their role is in confluence boost, not standalone defense.

Pool Type 5: IB Extremes (Initial Balance High / Low)

The 09:30 to 10:30 ET range edges on ES (and product-specific equivalents).

  • Stop density: high during the session it forms. Day traders use IB extremes as primary intraday reference; breakout-anticipation orders and trail stops cluster.
  • Reaction profile: the IB-extension-or-fade decision drives the day-type classification (Chapter 9). Sweep-and-fail at an IB extreme is a textbook range setup.
  • Decay profile: decays after the session it forms. By the next RTH open, the IB has lost most of its reference value.

Pool Type 6: Value Area Extremes (VAH / VAL)

The upper and lower bounds of the prior session's 70% value area.

  • Stop density: moderate. Tracked by profile-aware traders.
  • Reaction profile: strong, especially when current session opens out-of-balance relative to prior VA. Mean-reversion to VA is one of the more documented intraday tendencies on ES.
  • Confluence with other pools: common; VAH and VAL frequently coincide with PDH/PDL or with key swing pivots.

Pool Type 7: Naked POCs

POCs from prior sessions that have not been retested since formation. The most institutionally-respected single pool type, with retest rates around 75% within 30 sessions on ES (Chapter 9).

  • Stop density: moderate; the naked POC is more of a magnet than a stop pool, attracting price toward it rather than through it.
  • Reaction profile: very high probability of retest; reaction at retest depends on regime.

Pool Type 8: AVWAP touches

AVWAP from event anchors (FOMC, prior settlement, weekly open) acts as a moving liquidity pool because real participants reference it for execution.

  • Stop density: lower than horizontal pools because AVWAP is a moving reference.
  • Reaction profile: strong; institutional defense of AVWAP is real and persistent.

12.2The sweep mechanic, formalized

A sweep is the canonical pattern by which a liquidity pool gets harvested.

The five-condition sweep template

The author's Futures Institutional Edge indicator and the practitioner consensus converge on the following:

  1. Approach: price approaches the pool from one side.
  2. Pierce: price exceeds the pool's level by ≥ 0.10 × ATR (a non-trivial pierce; a 1-tick wick does not qualify).
  3. Velocity: the pierce occurs in a single bar (or two consecutive bars) with elevated volume. The pierce is forceful, not a slow drift.
  4. Rejection: the bar (or the next bar) closes back inside the pool's level. The sweep is rejected.
  5. Order-flow confirmation (optional but recommended): CVD divergence at the wick, or absorption signature on the rejection bar (stacked bid imbalances on a sweep-of-highs rejection bar).

A pattern satisfying all five is a high-quality sweep. Three of five is a flag. One or two is noise.

Pierce magnitude and pool type

The 0.10 × ATR pierce requirement scales appropriately for ATR. But some pool types deserve different thresholds:

Pool type Pierce threshold
Equal H/L 0.05 to 0.10 × ATR (smaller because stop density is high; even small pierces trigger cascade)
PDH/PDL 0.10 × ATR (standard)
Round number 0.10 to 0.15 × ATR (round numbers attract probes that don't always cascade)
IB extremes 0.10 × ATR (standard)
VAH/VAL 0.10 × ATR
Naked POCs the pierce concept does not directly apply; naked POC is a magnet, not a stop pool
Prior week/month 0.15 to 0.25 × ATR (multi-session traders' stops are spread further)

These are tunable defaults; calibrate to your contract and timeframe.

Velocity gate detail

The velocity gate distinguishes a real cascade from a slow drift through a level. A real cascade has:

  • Volume on the pierce bar ≥ 1.5× to 2× trailing 20-bar average.
  • The pierce bar's range ≥ 1.0 × ATR(14).
  • Bid-ask aggressive flow showing the cascade direction.

A slow drift has: low volume, small range, balanced flow. Slow drifts that "just happen to" pierce a level often do not harvest stops because the flow was insufficient to trigger the cascade; the orders behind the level remain.

Rejection vs continuation

After the pierce:

  • Rejection: price closes back inside the level. The sweep is complete; the next move is typically away from the swept side.
  • Continuation: price holds beyond the level. The "sweep" has become a breakout. The pool was harvested and the level was broken; the next move is in the breakout direction.

The same pierce mechanic produces opposite outcomes depending on the next bar's behavior. The rejection vs continuation distinction is what regime conditioning resolves.

The sweep mechanic, opposite outcomes by regime Range regime: sweep reverses Equal high pool Pierce Reverses Trend regime: sweep continues Equal high pool Pierce Continues Same pattern. Opposite outcome. Regime determines which. Regime classification is mandatory before sizing a sweep trade.
Figure D12.4.Sweep regime conditioning. The same pattern reverses in Range regimes and continues in Trend regimes. Regime is the binding variable.

12.3The regime-conditioning rule

Stated formally:

In Range regimes (Range-Calm, Range-Vol), sweeps tend to reject and reverse. In Trend regimes (Trend-Vol, Trend-Calm), sweeps tend to continue and breakout.

The same pattern, opposite outcomes, depending on regime classification (Chapter 2).

Why this is true

In a Range regime, the level has been defended in the past sessions; institutional participants have established their inventory at value-area edges; sweeps are liquidity-harvest events by which institutions accumulate (on sweep-of-lows) or distribute (on sweep-of-highs) at the edges of the range. The sweep is a feature of the range structure.

In a Trend regime, the level is resistance to the trend. Aggressive flow is consistently in one direction; a sweep is the moment the level breaks and the trend extends. The sweep is the resolution of a level that was never going to hold.

How to use the rule

Step 1: classify regime by 10:30 ET (composite from Chapter 2). Step 2: when price approaches a pool, your prior on the outcome is set by the regime: - Range: bias toward sweep-and-reverse. Watch for confirmation; setup is Framework 3 (Range Fade with Sweep). - Trend: bias toward sweep-and-continue. Watch for confirmation; setup is Framework 2 (Trend Continuation), with the breakout extension as the entry.

Step 3: regardless of bias, require order-flow confirmation before sizing. The bias from regime is conditional probability; the order flow is the trigger.

Worked example: same level, two regimes

Imagine PDH on ES at 6,212.

Day 1, regime classified as Range-Calm: price approaches 6,212 from below, pierces by 1.5 points to 6,213.50, closes back at 6,212.25. CVD shows divergence; footprint shows stacked bid imbalances on the rejection bar. Sweep-and-reverse confirmed. Trade: short the rejection close, stop above 6,214, target IB midpoint at 6,206.

Day 2, regime classified as Trend-Vol up-trend: price approaches 6,212 from below, pierces to 6,213.50, holds, closes at 6,213. CVD continues higher (no divergence); footprint shows continued stacked ask imbalances. Sweep-and-continue confirmed. Trade: long on the breakout close, stop below 6,210, target the next structural level above (6,225 or wherever the next high-score level sits).

Same level, same approach, opposite trades, dictated by regime.

12.4The FVG (Fair-Value Gap) framework

Definition recap (from Chapter 3)

A three-bar FVG is formed when:

  • Up-impulse: bar 1's high is below bar 3's low. The price band between bar 1's high and bar 3's low did not transact during the impulse.
  • Down-impulse: bar 1's low is above bar 3's high. Symmetric.

The retest framework

FVGs in liquid futures retest at high empirical frequency under conditions:

  • Impulse strength: the impulsive bar (bar 2) must be ≥ 1.0 × ATR. Smaller "FVGs" formed by minor bars are noise.
  • Session context: RTH FVGs are more reliable than ETH FVGs; RTH on a non-news day is the strongest case.
  • Regime: in Trend regimes, FVGs in the trend's direction retest with very high frequency (the pullback to the FVG is the continuation entry). In Range regimes, FVGs are less reliable as standalone retest targets.

The retest patterns

When price approaches an FVG:

  • Proximal-edge retest: price touches the closer edge of the FVG (the edge first encountered on the retracement) and reverses. This is the most common retest pattern; the FVG is partially filled.
  • Mid-FVG retest: price moves into the middle of the FVG and reverses. Less common; suggests continued impulse strength.
  • Full-fill retest: price moves through the entire FVG to the far edge. The FVG is fully filled; the impulsive move's structural significance has been consumed.

Empirical estimate: on ES 5-min during Trend regimes, ~75% of qualifying FVGs see at least proximal-edge retest within 20 bars; ~40% see mid-FVG retest; ~15% see full-fill.

Trade structure

The FVG retest is a continuation entry in Trend regimes:

  • In an up-trend: wait for price to retrace into an FVG that formed during a recent up-impulse. Enter long at the proximal edge with order-flow confirmation. Stop below the FVG. Target the prior swing high.
  • In a down-trend: symmetric.

Failure modes

  • FVGs in Range regimes: less reliable; the chop within a range can fill an FVG as part of normal noise, not as a continuation entry.
  • FVGs at the very start of a trend: the trend is too young to have established the continuation behavior; size smaller until the trend confirms.
  • FVGs that have already been partially filled: the proximal-edge retest has already happened; the FVG is now a "tested" structure with reduced reaction probability.

12.5Multi-pool harvest sequences

Liquidity-based reading scales beyond single sweeps. Many sessions feature cascades where one sweep triggers the next.

The classic multi-pool sequence

  1. Session opens. Equal highs from yesterday's afternoon at 6,212.
  2. Price pushes up, sweeps the equal highs at 6,213.50, rejects, reverses to 6,205.
  3. The reversal pushes through equal lows at 6,206 (formed earlier in the morning), sweeps to 6,204, rejects, reverses to 6,210.
  4. The chop between 6,206 and 6,212 has now harvested liquidity on both sides; the next directional move is on cleaner liquidity.

Why this matters

A trader who tries to fade both sweeps gets two trades. A trader who sees the multi-pool cascade as a unit recognizes that after both pools are harvested, the next directional move has elevated probability of holding (no fresh liquidity to harvest). The third trade in this sequence is the high-conviction one.

The "third sweep" pattern

A specific high-quality variant: after two sweeps in opposite directions (a high-pool sweep then a low-pool sweep, or vice versa), the third attempt at one side often runs because both pools are exhausted. Recognize the cascade and do not fade the third move; it is structurally the breakout.

12.6Liquidity in news and event windows

News blackout protocol applies, but with nuance: news creates liquidity events that are different from organic sweeps.

News-driven moves

When a tier-one event (FOMC, NFP, CPI) drops:

  • The initial reaction (first 1 to 5 minutes) is dominated by spread-takers and HFT response. Order flow during this window is mechanical, not informative.
  • Price often spikes through nearby liquidity pools (PDH, PDL, equal highs/lows) on the news bar.
  • The "sweep" mechanic looks like the canonical pattern but is not an institutional liquidity-harvest in the normal sense; it is a forced repricing event.

How to handle

Skip the news bar for sweep-trade analysis. Wait at least 15 to 30 minutes after the news. The pools that were swept during the news event were harvested for non-institutional reasons (mechanical stop cascades from forced rebalancing); the post-news structure is what to read.

The new pools that form post-news (new equal highs, new IB-equivalent ranges) are the relevant frames for the next several hours.

12.7Operational considerations

Pool tracking

A trader needs a system for tracking pools as they form and as they are swept. Options:

  • Manual: mark equal H/L and PDH/PDL on the chart at session start; update during the session.
  • Indicator-assisted: use a tool that tracks equal H/L automatically (the author's Trend & Levels and Futures Institutional Edge both do).
  • Mental: for pools that aren't on the platform's automated list (rare events, unique structural levels).

The discipline is consistency: the same pool tracking method every session.

Sweep alerts

Some platforms support "sweep alerts" that fire when the canonical pattern completes. These can be useful but require careful tuning:

  • Too sensitive (every wick fires): false positives dominate.
  • Too strict (only big sweeps): misses 80% of valid sweeps.

The right calibration is on a per-contract basis; tune over a 30-session sample.

Visualization

Pool visualization on the chart should be subtle, not overwhelming. The author's preference: thin lines or small ticks at pool levels, with sweep events marked retrospectively (an X marker after the fact, not a flashing alert).

12.8Failure modes

  1. Trading every sweep. Most sweeps in choppy market structure are noise. The conditioning rules (regime, structural confluence, order-flow confirmation) are what distinguish tradeable from noise.

  2. Fading sweeps in Trend regimes. The most expensive single mistake. Same pattern, opposite outcome; the regime-conditioning rule is non-negotiable.

  3. Treating FVGs in Range regimes as retest targets. The retest behavior is a Trend-regime property; in Range, FVGs fill as part of normal chop without producing tradeable continuations.

  4. Pierce magnitude blindness. A 1-tick wick is not a sweep; the pierce must be ≥ 0.10 × ATR for the cascade to have triggered.

  5. Velocity blindness. A slow drift through a level is not a sweep cascade. The velocity gate (volume + range) is part of the pattern, not optional.

  6. Pool age blindness. An equal-high cluster from 3 weeks ago has decayed in stop density. New stops have replaced or surpassed old ones; the pool's location may be stale.

  7. Multi-pool cascade misreading. Trying to fade every sweep in a cascade burns capital and ignores the structural meaning of the cascade. After two sweeps in opposite directions, the third move is often the runner.

  8. News-driven sweep over-trust. News-driven sweeps are forced repricings, not institutional harvests; they do not produce the normal post-sweep behavior.

12.9The integrated stack treatment

Liquidity is Layer 4 of the institutional stack: structural levels (Layer 1) → volume profile (Layer 2) → AVWAP (Layer 3) → liquidity flags (Layer 4) → order-flow confirmation (Layer 5).

Liquidity flags identify where the action will likely happen. Order flow confirms whether the action is the high-quality version. Together they form the most actionable two-layer combination in the book.

The sequence:

  1. Identify pools at session start (mark on chart).
  2. Note pools that have been swept; remove from active list (they are harvested).
  3. Watch active pools for approach.
  4. On approach, await sweep mechanic (pierce + close-back-inside + velocity).
  5. Check order flow on the rejection bar.
  6. Apply regime-conditioning rule.
  7. Trade per the appropriate framework (2 or 3).

12.10Diagram concepts referenced in this chapter

  • D12.1: Pool taxonomy gallery. A grid of small annotated charts, one per pool type (equal H/L, PDH/PDL, round number, IB, VAH/VAL, naked POC), each showing the typical sweep mechanic.
  • D12.2: Five-condition sweep template walkthrough. A single sweep example with each of the five conditions called out at the bar where it is satisfied.
  • D12.3: FVG retest by impulse strength. Two panels: a strong-impulse FVG that retested, a weak-impulse FVG that did not. The conditioning is visible.
  • D12.4: Sweep regime conditioning. The same equal-high pool shown in a Range-Calm session (sweep-and-reverse) and a Trend-Vol session (sweep-and-continue).
  • D12.5: Multi-pool cascade. A 90-minute chart segment showing three sweeps in succession: high pool, low pool, high pool again (the third runs).
  • D12.6: News-driven false sweep. A pre-FOMC chart showing a "sweep" pattern that was a news-driven repricing, not an institutional harvest. Annotation explains.

12.12Exercises

Exercise 12.1: Pool tracking session. For one ES RTH session, start at 09:30 ET and mark every pool type as they form (equal H/L, IB, VAH/VAL from prior session, etc.). Maintain the active list through the session. By close, count: how many were swept, how many held, how many were broken.

Exercise 12.2: Sweep regime audit. For 20 sessions split between Trend and Range regimes, identify all sweeps that satisfied the five-condition template. Tabulate outcome (reverse vs continue) by regime. The hypothesis: Range sessions show reversal-dominant outcomes; Trend sessions show continuation-dominant.

Exercise 12.3: FVG retest tracking. Identify all FVGs ≥ 1.0 × ATR formed during RTH on 30 sessions of NQ 5-min. Track which retest within 20 bars. Stratify by regime. The retest rate in Trend regimes should be visibly higher than in Range.

Exercise 12.4: Multi-pool cascade identification. Find a recent session where two or more sweeps in opposite directions occurred within 60 minutes. Annotate. Note whether the third move (after both pools harvested) ran or chopped. Build the pattern recognition.

Exercise 12.5: News blackout pool audit. For a recent FOMC session, identify all pools that were swept during the news bar. Compare to sweeps after the 30-minute post-news window. Note the difference in subsequent behavior. The point: news-driven sweeps do not behave like organic sweeps.

Next chapter: the Open, Open Type classification, IB extension probabilities, and the day-structure framework that anchors every intraday decision.