Absorption, Imbalance & Initiative
Execution Precision
8 min read
Learn the language of real-time control by identifying absorption, volume imbalance, and initiative patterns in the order flow.
Loading
8 min read
Learn the language of real-time control by identifying absorption, volume imbalance, and initiative patterns in the order flow.
Candles are footprints. Order flow is the muscle. Learn to see who's pushing, who's holding, and who's about to get run over.
In one paragraph: absorption is large aggressive volume hitting a level without commensurate price progress; imbalance is one-sided volume dominance — typically a diagonal footprint ratio of 3× or more — between adjacent prices; initiative is taker-driven aggression that crosses the spread to push price away from value. All three describe who controls the auction right now. The rest of this lesson teaches you to spot, quantify, and avoid the false-positive trap each pattern hides.
You should already know:
These three concepts are how informed flow — desks, market makers, and liquidating size — leaves footprints in the book. They are not magic. Roughly half of textbook-clean reads fail; the edge is in the asymmetry of the ones that resolve, not in being right often.
Absorption = aggressive volume × time at a level without commensurate price progress. A working quantitative definition: at least 2× the prior-bar aggressive volume gets traded while price advances less than 0.25 × ATR(14). Treat absorption as a condition, not a signal.
absorption := V_agg(t) >= 2 * V_agg(t-1) AND |dP| < 0.25 * ATR(14)
BTCUSDT perp (Binance), NY session. Aggressor buys ~1,200 contracts (~$74M notional) cross the spread in 10s. Best ask refills 4x; two 3-tick wicks print; price reverses 0.4% over the next minute. Treat as condition, not signal — confirm with CVD divergence.
Caveat: what looks like a defended level can also be (a) iceberg liquidity that runs out one bar later, or (b) a spoofed bid that vanishes when probed. The next lesson — spoofing, stacking, and iceberg orders — covers detection. Until then, treat absorption alone as a low-confidence read, confirmed only by a follow-through delta divergence on CVD.
Imbalance = one-sided volume dominance at a footprint cell, measured diagonally. Convention: compare bid-traded volume at price P to ask-traded volume at price P + 1 tick. When that diagonal ratio exceeds 3× for three consecutive rows, the row stack is flagged as a stacked imbalance. The 3× threshold is heuristic — popularized by Bookmap, ATAS, and Sierra Chart defaults; 4× is stricter, 2× is noisy.
stacked_imbalance := for k in {0,1,2}: V_bid(P-k) / V_ask(P-k+1) >= 3
One-sided aggression eats resting liquidity. The next market order has to walk further down the book, triggering stops at the next levels, which fire more market orders — a positive-feedback cascade. This is the mechanical reason imbalance often precedes breakouts, stop runs, and continuation legs.
The same signature appears at session opens (everyone agrees on direction for 30 seconds, then mean-reverts) and around funding-rate flushes. Filter: imbalance into a known POI is signal; imbalance in chop is noise.
From Dalton's auction-market framework: initiative participants take liquidity to push price away from value; responsive participants provide liquidity into perceived edges of value. Use the term "responsive" — calling it "passive" conflates passive limit-order mechanics with the auction concept.
| Type | Description | Behavior |
|---|---|---|
| Initiative | Market orders crossing spread to drive price | Taking control (attack) |
| Responsive | Limit orders placed to absorb or slow price | Defending value (resist) |
Key Question:
Is the market moving because someone's chasing, or stalling because someone's absorbing?
When initiative + imbalance + structure align → a higher-probability read — but do not confuse this with certainty. In our own log of ~400 confluence setups on BTC perp, the resolved-as-expected rate hovers around 55–60% before fees. Edge comes from the asymmetry (1:2+ R), not hit rate. Size accordingly; the pattern that "looks textbook" fails often enough to bankrupt anyone who oversizes.
n = ~400 journaled setups, before fees. Edge is in 1:2+ R asymmetry, not hit rate.
| Pattern | Footprint signature | DOM signature | Common failure | When to trade |
|---|---|---|---|---|
| Absorption | High row volume, range < 1 tick | Limit order keeps refilling | Iceberg runs out → continuation | Only with delta-divergence confirm |
| Imbalance | ≥ 3× diagonal ratio, 3 stacked rows | Thin opposite side | Mean-reverts at session open | At a known POI, on retest |
| Initiative | Wide rows, fast tape, sweep | Spread widens, depth peeled | Exhaustion at swing extreme | Only aligned with HTF trend |
| Responsive | Stacked refills against the move | Pull-and-replace pattern | Defender capitulates | At edges of established value |
Open Trading Glass footprint on BTCUSDT 1-minute bars at the NY open. The drill (do this for 30 samples before forming an opinion):
Typical reader base rate after 30 NY-open samples
Your own log overrides this. The ~15% outright-failure tail is why sizing for hit rate (rather than asymmetry) bankrupts traders.
If you want a setup off this drill: entry on the retest of the absorption wick or after a 1-minute reclaim of the imbalance origin; stop a few ticks beyond the absorption wick (or the failure level of the stacked imbalance); first target ≥ 1× the stop distance, runner managed structurally. You read a plausible intention. Treat it as a hypothesis with an invalidation — not a verdict.
Absorption is when large aggressive market orders hit a price level without moving price — passive limit orders soak up the flow without yielding ground. A working quantitative threshold: at least 2× the prior-bar aggressive volume traded while price advances less than 0.25 × ATR(14).
Imbalance is one-sided volume dominance with price moving — typically a diagonal footprint ratio of 3× or more between adjacent prices. Absorption is large aggressive volume with price stuck. Imbalance is aggression finding no resistance; absorption is aggression meeting a defender.
A stacked imbalance is three or more consecutive footprint rows where the diagonal volume ratio (bid volume at price P vs. ask volume at price P + 1 tick) exceeds 3×. The 3× threshold is the Bookmap / ATAS / Sierra Chart default; 4× is stricter, 2× tends to be noisy.
No. Roughly half of textbook-clean absorptions fail — the apparent defender is an iceberg that runs out, a spoofed bid that vanishes, or a level that simply gets overrun. Treat absorption as a condition requiring confirmation (typically a CVD divergence and a structural trigger), not as a standalone signal.
From Dalton's auction-market framework: initiative participants cross the spread with market orders to push price away from value; responsive participants post limit orders into perceived edges of value to fade extension. Initiative is aggression; responsive is defense of fair value.
Structure shows you where. Order flow shows you when.
Imbalance = aggression. Absorption = defense. Initiative = commitment.
When they align with your setup, you don't have a signal — you have a hypothesis with an invalidation, backed by intentional flow. That is the most a probabilistic edge can ever be.
Next: spoofing, stacking, and iceberg orders — absorption is exactly where icebergs and spoofs hide. The next lesson teaches you to filter them out before they cost you a stop.