Absorption and Exhaustion

Two bars can share their open, high, low and close, their total volume, and even the volume traded at every individual price, and still carry deltas of +2,820 and −1,680. One of them closes thirty cents up on net aggressive selling, which one school calls exhaustion and fades and the other calls absorption and buys. Both readings are widely taught, both are internally consistent, and the footprint does not decide between them.

Prerequisites: Lesson 26, which established that an untested level is a number somebody typed, and lesson 17, because a reading of a bar is worth whatever it moves p by and nothing else.

Two bars that a candle cannot tell apart

Here are two bars. Both opened at $100.00, made a high of $100.50 and a low of $99.80, and closed at $100.30. Both traded ten thousand units. More than that: both traded the same volume at every individual price — 300 at the high, then 600, 1,200, 1,500, 1,600, 2,000, 2,000, and 800 at the low. Every quantity a candlestick records, and several it does not, are identical.

Price	A buys	A sells	A delta	B buys	B sells	B delta
$100.50	140	160	−20	60	240	−180
$100.40	320	280	+40	150	450	−300
$100.30	700	500	+200	350	850	−500
$100.20	950	550	+400	550	950	−400
$100.10	1,000	600	+400	700	900	−200
$100.00	1,400	600	+800	950	1,050	−100
$99.90	1,400	600	+800	1,000	1,000	0
$99.80	500	300	+200	400	400	0
Total	6,410	3,590	+2,820	4,160	5,840	−1,680

Check the constraints before reading anything into it, because that is the habit worth building. Every row adds to the volume at that price in both columns. Every delta is smaller than its row’s volume. The totals reconcile: 6,410 and 3,590 make ten thousand, and so do 4,160 and 5,840. Both bars are possible. Now they can be read.

In bar A the aggression was buying and it happened low in the range: +1,800 of the +2,820 came from the bottom three prices, and the top three contributed only +220 between them. In bar B the aggression was selling, most heavily at the top: −980 of the −1,680 came from the top three prices, while the bottom three came to −100. Two accounts of the same hour that share every number a chart displays.

Here is where the folklore turns on itself, and it is the reason this lesson exists. Bar B closed up thirty cents on net aggressive selling of 1,680 units. One school reads that as exhaustion: sellers were leaning on it, the close is a fake, fade it. The other reads it as textbook absorption: aggressive sellers hit bids all bar and price went up anyway, which means someone was quietly buying everything on the bid, and that is strength. Both readings are internally consistent, both are widely taught, and they are opposite trades on identical evidence.

The lesson is not that one school is right. It is that the footprint did not settle it, and no amount of staring at bar B will settle it either, because the disagreement is not about what the numbers are. What would settle it is the thing lesson 26 asked for and this lesson can now ask for again: how often, on your instrument, does a bar of shape B get followed by a higher close? That is a base rate, it takes tallies rather than trades, and it is the only thing on this page that can turn bar B into a reason to do anything.

What this does not settle

That the delta on your screen is a measurement. It is on some venues and an estimate on others, and you may not currently know which yours is. That is answerable in an afternoon by reading your data provider’s documentation for the words “aggressor” or “initiator”, and it changes how much weight the rest of this lesson can carry for you. A classified delta and a flagged delta are not the same instrument.

That absorption is visible while it happens. It is not, and the name gives the game away: a level absorbed the selling only if it went on to hold, which you learn afterwards. In the moment you have a bar with a lot of volume and a delta, and the two readings above are both still open. Anything that presents absorption as a live signal has quietly moved the outcome into the definition.

That the two constraints make a footprint true. They make it possible. A reconciling footprint can still be misclassified, mislabelled or simply from a different instrument than you were told, and the arithmetic will not notice. What the constraints catch is fabrication, which is worth catching and is not the same as verification.

That the level holding tells you the level will hold next time. It tells you it held this time, which is one observation. Lesson 26’s table is the machinery for turning observations of that kind into a rate, and one is not a rate. This is the same discipline as everywhere else in the module and it does not relax because the evidence finally became a fact.

That where volume traded across a whole session is the same question. It is a different one and a good one: this lesson looks inside a single bar, and the distribution of volume across a session or a week has its own structure, its own vocabulary and its own failure modes. Lessons 29 and 30 take it up, in that order.

A footprint is not the truth behind the candle. It is a second measurement, taken by a different method, with its own error rate — and its value is precisely that it can disagree with the first one.

Problems

1. Audit five footprints. On your own platform, take five bars and for each one add the buy column and the sell column and compare the total to the bar’s volume, then check that no single row’s delta exceeds that row’s volume. It is a few minutes of arithmetic and should come out clean. If it does not, you have learned something considerably more useful than anything else in this lesson, and the thing to do is find out what your platform is actually plotting before you make another decision with it.
2. Find out how your delta is made. Open your data provider’s documentation and search it for “aggressor”, “initiator”, “side” or “tick rule”. You are looking for one of two answers: the venue publishes which side initiated, or your software is inferring it from the quote. Write down which. Everything you subsequently conclude from a delta inherits the answer, and the question is one this lesson cannot answer for you because it depends on where your data comes from.
3. Get the base rate the worked example needs. Take forty bars on the instrument and timeframe you actually trade — consecutive, not selected — and sort each into one of four boxes by two questions: did it close up or down, and was the net delta positive or negative. Two of the boxes are agreements and two are disagreements. The proportion in the disagreement boxes is how often the situation this lesson is about even arises. Then, for the disagreement bars only, count how many were followed by a higher close. That second figure is the only thing that makes bar B actionable, and nobody can hand it to you.

Where the problems above ask you to go and count something, How to Collect a Base Rate is the appendix that says how: define the observation, fix the criterion before you look, take consecutive cases rather than the memorable ones, and count into four cells.

Sources. Charles M. C. Lee and Mark J. Ready, “Inferring Trade Direction from Intraday Data” (Journal of Finance, 1991), the canonical rule for deciding which side of a trade was the aggressor when the venue does not say, and therefore the step that most delta figures pass through without the reader knowing. Katrina Ellis, Roni Michaely and Maureen O’Hara, “The Accuracy of Trade Classification Rules: Evidence from Nasdaq” (Journal of Financial and Quantitative Analysis, 2000), which checks that rule against data where the true side was known and reports where it fails: inside the spread, and in fast markets. David Easley, Marcos López de Prado and Maureen O’Hara, “Flow Toxicity and Liquidity in a High-Frequency World” (Review of Financial Studies, 2012), for what order-flow imbalance is good for when it is aggregated properly — over volume rather than over clock time — which is a different and more defensible use of the same raw material.

This lesson stayed inside one bar. The next two step back: where volume has traded across a session, why the busiest prices behave differently from the quiet ones, and what a distribution of volume can and cannot tell you about where price is likely to go.

Educational only. Trading involves substantial risk of loss. Not financial advice. Past performance does not guarantee future results.