Part XI · Swing trading · Chapter 11E

Stage analysis

Weinstein's four phases

Where we are

Chapter 11D read chart patterns as priors on order flow. This chapter classifies the trend itself into one of four stages, following Stan Weinstein, and turns the classification into rules a computer can run bar by bar. The stage is the main input of the structural layer of Chapter 11B. The Dealer Mechanics Lab finds its highest-edge moments at the transitions between stages.

What you need first

Simple and exponential moving averages (Chapter 11A).
Average true range and the average directional index (Chapter 11C).
Accumulation and distribution, and the patterns that show them (Chapter 11D).
Gamma exposure and the gamma flip (Chapters 8A and 8C), and the Edge Book's lower opinion of them (Chapter 10H).

1 · Four stages of a trend

Stan Weinstein's four-stage framework, formalized in Secrets for Profiting in Bull and Bear Markets (1988), is in the Dealer Mechanics Lab's judgment the most durable operating heuristic in swing trading. It compresses everything an institution does, from accumulation through markup and distribution to markdown, into a small finite-state machine. That is a system always in exactly one of a few states, with rules for moving from one to the next, and here each state is a Weinstein stage.

In plain terms

A stock's life on a chart runs in a cycle. It goes quiet and sideways while large buyers accumulate, rises while they mark it up, stalls while they hand it off, and falls once nobody is left to buy. Stage analysis names the four phases and gives you one job in each: wait, buy the dips, protect your gains or stay out.

Stage	Name	Stance	Action	What it looks like
1	Basing	Neutral	PATIENCE	A flat 30-bar moving average (MA), volume drying up, range-bound price
2	Markup	Long	BUY PULLBACKS	A rising 30-bar MA, price above it and ADX above 18
3	Distribution	Reduce	TIGHTEN STOPS	A flat MA after an extended advance
4	Markdown	Short or cash	DO NOT BUY DIPS	A falling 30-bar MA, price below it and ADX above 18

Each stage comes with its own discipline:

Stage 1, basing. Wait for the first Stage 2 breakout above the ceiling of the base, and do not anticipate it.
Stage 2, markup. The Dealer Mechanics Lab calls this stage the one "where 90% of swing money is made". Pullbacks to the 10-bar or 21-bar EMA are gifts: buy them, with a trailing stop below the rising MA.
Stage 3, distribution. With institutions handing their positions off to retail buyers, reduce your size and take the first close below the rolling MA as the exit.
Stage 4, markdown. Bounces fail at the MA, so wait for it to flatten before you look for a way back in.

Stage analysis reads a trend from the slope of a moving average and the side of it the price is on. The trend following of Chapter 10D reads a trend from the sign of a 12-month return. Both condition on a trailing measure of trend, one to time a single chart and the other to size a whole portfolio.

2 · The classifier

The classifier turns the four descriptions into rules built on three inputs:

The slope of the 30-bar SMA, divided by the ATR so that one threshold serves any asset.
The price relative to the MA: whether the close sits above or below it.
ADX as a trend-strength gate, which the rules ask to exceed 18.

Write $M_t$ for the 30-bar SMA and $\text{ATR}_t$ for the 14-bar ATR of Chapter 11C. The ATR-normalized slope is the average's rise per bar over the last five bars, measured in ATRs:

\tilde{s}_t = \frac{M_t - M_{t-5}}{5\,\text{ATR}_t}

(11E.1)

A threshold of 0.05 on $\tilde{s}_t$ means the average climbs one ATR every $1/0.05 = 20$ bars. Dividing by the ATR is what lets the same number describe SPX at 6,000 and a $50 stock. The third stage also uses the slope of the 10-bar EMA, $(E_t - E_{t-5})/5$ .

The source states four rules, tried in order, the first match winning:

s = (SMA30[t] − SMA30[t−5]) / 5 / ATR14[t]

Stage 2:  s > +0.05   and price > SMA30   and ADX > 18
Stage 4:  s < −0.05   and price < SMA30   and ADX > 18
Stage 3:  |s| < 0.05  and price > SMA30   and EMA10 slope < 0
Stage 1:  |s| < 0.05  (otherwise)

Those four rules leave some bars unassigned: a clear slope whose price or ADX test fails. The figure below, like the Dealer Mechanics Lab's own widget, assigns those by the sign of the slope and the side of the price:

slope up,   price above SMA30:  Stage 2
slope up,   price below SMA30:  Stage 1
slope down, price above SMA30:  Stage 3
slope down, price below SMA30:  Stage 4

A consequence follows that the source does not state: a bar that fails only the ADX test lands in the same stage through these fallback rules as it would with ADX above 18. In the figure's rules the ADX gate therefore never changes a classification. It would matter only if the source's "otherwise" sent such bars to Stage 1, which the figure does not do.

Worked example: Classifying three bars

Step 1: a markup. SPX's 30-bar SMA is 5,850, against 5,835 five bars ago, and with a 14-bar ATR of 45, (11E.1) gives $\tilde{s} = 15/(5 \times 45) = 0.067$ , above +0.05. With the price at 5,905, above the SMA, and ADX at 22, the first rule gives Stage 2.

Step 2: a flattening. Weeks later the SMA is 5,871, against 5,868 five bars ago, so $\tilde{s} = 3/225 = 0.013$ : flat. The price, 5,880, is still above the SMA, but the 10-bar EMA has fallen from 5,905 to 5,890 in five bars, a slope of −3 a bar, so the third rule gives Stage 3.

Step 3: a fallback. A stock's SMA rises 0.4 a bar with an ATR of 5, so $\tilde{s} = 0.08$ . The price is above the SMA, but ADX is only 15. The first rule fails on ADX, leaving no rule of the source that matches. The fallback for a rising slope with the price above gives Stage 2 anyway.

3 · Try the stage classifier

The figure below classifies a synthetic 350-bar tape, generated from a seed, that passes through a base, a markup, a pullback, a continuation, a distribution, a markdown and a new base. It draws the price with its long moving average and shades each bar's background with the stage the rules assign. One slider sets the long MA period from 15 to 60, starting at 30, and another the seed from 1 to 30, starting at 7, while the ATR, ADX and 10-bar EMA keep their fixed periods. Readouts give the last bar's stage and the action that goes with it.

The first long-period-plus-10 bars carry no stage, since the rules need that much history. Start at the defaults and find where the classifier first calls the markup and where it first sees the top, then set the period to 15 and to 60.

long MA period 30

tape seed #7

last bar = Stage 1 · Basing

action = PATIENCE

The classifier, bar by bar. A synthetic 350-bar tape, generated from a seed with a base, a markup, a pullback, a continuation, a distribution, a markdown and a new base. The price is drawn with its long moving average, and each bar's background is the stage the classifier's rules assign to it. Change the long-MA period or the seed to test how robust the classification is, and watch for flicker between adjacent stages at the transitions.

At the defaults the classifier calls the markup Stage 2 from bar 40, the first bar it classifies. It reads the sharp pullback that follows as Stage 4 for a while, then returns to Stage 2 for the continuation. It calls most of the distribution Stage 2 as well, because a 30-bar average keeps rising while a top forms: stage analysis sees a top late. Near bar 240 the background flips between Stage 1 and Stage 2 several times before settling into Stage 1 and then Stage 4. The last bar reads Stage 1, basing, with the action PATIENCE. At a period of 15 the stage changes about 47 times, against about 20 at the default and 17 at 60.

4 · Stages and dealer gamma

The Dealer Mechanics Lab reads stage analysis and the GEX regime of Part VIII as two views of the same underlying reality. The overlap, it argues, is no coincidence: drift supported by rising dealer gamma is exactly what institutional accumulation looks like on the tape.

Stage	Typical GEX regime	Path	Best strategy
1	Near zero or modestly positive	Range-bound, with low ATR and low ADX	Sell premium in range trades; avoid directional trades
2	Positive, above the flip	Trending up, with mean-reverting pullbacks to the MA	Buy pullbacks; trail stops below the rising MA
3	Compressing toward zero	A widening range, rising ATR, falling ADX	Tighten stops; reduce size; the first break of the MA is the exit
4	Negative, below the flip	Trending down, with sharp counter-trend bounces that fail	Do not buy dips; bounces are short opportunities for the brave

The source cites no study for the correspondence. On its GEX side, the Edge Book keeps only the sign, as low-confidence context (Chapter 10H). Selling premium for range trades in Stage 1 is also a short-volatility trade of the kind the Edge Book allows only as monthly cash-secured put writing on XSP (Chapters 10C and 10I).

5 · The transition is the trade

The highest-edge moments are transitions, not the middles of stages: Stage 1 to Stage 2, basing to markup, is the breakout, and Stage 3 to Stage 4, distribution to markdown, is the breakdown. The middle of Stage 2 is where money comes most easily, while its boundaries are the most important places to read carefully.

When the classifier flickers between adjacent stages at a boundary, that is the signal, not the noise. When a stage classification is unstable across days, the regime is genuinely in transition, which your sizing should reflect. One way to put a number on it is to count the changes of stage over the last ten bars and scale the position down as the count rises.

Worked example: Sizing through a flicker

Step 1: the count. Over the last ten bars the classifier has read 2, 2, 1, 1, 2, 1, 2, 2, 1, 1, changing stage five times.

Step 2: a rule set in advance. Suppose your written rule is full size with no change in ten bars, half size with one or two, and none with three or more. Five changes put this chart at zero size, which keeps you out until the regime settles.

The thresholds in that rule are an illustration, not a tested setting. What matters is that you fix them before you look at the charts they judge.

Familiar ground

Stage analysis is disease staging for a trend: each stage carries its own treatment, with criteria explicit enough that two readers assign the same stage. The lesson of stage migration carries over too: change the criteria, as a new imaging method does, and borderline cases move between stages. That migration can improve the average outcome in both stages without any patient doing better. Tuning the classifier's moving-average period until each stage's backtest looks good does the same, so fix the period before you test.

In practice

Classify before you look for a setup: the stage decides which setups you may consider, as the structural layer of Chapter 11B.
In Stage 2, buy pullbacks to the 10-bar or 21-bar EMA, with a trailing stop below the rising 30-bar MA, and size from the distance to that stop (Chapter 11C).
In Stage 3, cut size and exit on the first close below the MA. In Stage 4, stay flat or short, if shorting is in your mandate.
Expect lag: the classifier recognizes a top only after the 30-bar average flattens, so part of every decline arrives before Stage 3 appears.

When it fails

The classifier is late by construction, since a 30-bar average turns only after the price does. At the defaults of the figure it reads most of the distribution as Stage 2. It flickers at every boundary, while a longer period that damps the flicker makes it later still. In the figure's rules the ADX gate never changes a stage, so a weak, trendless drift can read as Stage 2. And the overlap with GEX regimes is the Dealer Mechanics Lab's reading, not a tested result.

Decision rule

Stage 1: wait, and act only on a breakout above the ceiling of the base.
Stage 2: long setups only, entered on pullbacks, with a stop trailed below the rising MA.
Stage 3: reduce size, tighten stops, and exit on the first close below the MA.
Stage 4: no dip-buying, and no re-entry until the MA flattens.
At a boundary: if the stage has changed more than once in recent bars, size down by a rule you wrote in advance.

Common mistake: buying the first bounce in Stage 4

A sharp rally inside a falling trend looks like a bottom. In Stage 4, bounces fail at the moving average, which is why the rule waits for the average to flatten, not for the price to jump.

Common mistake: anticipating the breakout

Buying a Stage 1 base before it breaks out saves a few points on a winner and pays for every base that never breaks out. The source's advice is plain: wait for the first Stage 2 breakout, and do not anticipate.

What to carry forward

Weinstein's four stages, basing, markup, distribution and markdown, each carry one job: wait, buy pullbacks, tighten stops, stay out.
The classifier reads the ATR-normalized slope of the 30-bar SMA, the price's side of it, ADX and the 10-bar EMA's slope. A slope of 0.05 means one ATR every 20 bars.
In the figure's rules the fallbacks make the ADX gate inert, and the classifier sees tops late.
Transitions are the trade, with flicker at a boundary as information: size down while the stage is unstable.
The overlap with GEX regimes is the Dealer Mechanics Lab's reading, which the Edge Book discounts on the GEX side.

6 · Check yourself

Question 1

A stock's 30-bar SMA is 102.0, against 101.0 five bars ago, and its ATR is 2.5, with the price at 104 and ADX at 25. What stage is it in?

Answer

The slope is $1.0/5 = 0.2$ a bar, so (11E.1) gives $\tilde{s} = 0.2/2.5 = 0.08$ , above +0.05. With the price above the SMA and ADX above 18, it is Stage 2.

Question 2

With the same slope, the price drops to 100. What stage is the stock in now?

Answer

The slope still clears +0.05, but with the price below the SMA none of the source's four rules matches. The fallback for a rising slope with the price below gives Stage 1.

Question 3

The normalized slope is −0.03, the price is above the SMA, and the 10-bar EMA is rising: is this Stage 3 or Stage 1?

Answer

Stage 1. The slope is flat, since $|{-0.03}| < 0.05$ , and the price is above the SMA. Stage 3 also needs a falling 10-bar EMA, which this chart lacks.

Question 4

What does a normalized slope of +0.10 mean in plain terms?

Answer

The 30-bar average rises 0.10 ATRs a bar, one full ATR every 10 bars: twice the Stage 2 threshold.

Question 5

In the last ten bars the stage has read 2, 2, 2, 2, 2, 2, 3, 2, 3, 3. What does the chapter say to do?

Answer

The stage changed three times as the markup gave way to distribution, so the regime is in transition. Cut the size by the rule you set in advance. The Stage 3 reading also calls for tighter stops and an exit on the first close below the MA.