Part XI · Swing trading · Chapter 11F

Multiple timeframes

Elder's triple screen

Where we are

Chapters 11C to 11E built the instruments, from indicators and chart patterns to the stage of a trend, each reading one chart at one scale. This chapter makes them check each other across scales with Elder's triple screen, which takes a trade only when three timeframes agree.

What you need first

Bars, and the simple and exponential moving averages that smooth them (Chapter 11A).
The relative strength index, MACD and the average true range (Chapter 11C).
The slope of the 30-bar moving average, and the four stages of a trend it helps classify (Chapter 11E).
The rule that a lower layer of the hierarchy never overrides a higher one (Chapter 11B).

1 · Three screens, three jobs

The Dealer Mechanics Lab calls Alexander Elder's triple screen, formalized in his Trading for a Living (1993), the cleanest formalization of multi-timeframe discipline. At any moment three timeframes are at work, each with one job: the higher timeframe (HTF) sets the direction, the intermediate timeframe (ITF) signals the timing, and the lower timeframe (LTF) triggers the entry. All three must agree.

A timeframe is the length of the bars a chart is drawn in. A bar, as Chapter 11A defined it, is one period of the chart, with its open, high, low and close. A weekly bar condenses five daily bars: it opens at Monday's open, closes at Friday's close, and takes the week's highest high and lowest low as its own. For daily swing trading the HTF is typically weekly, the ITF daily and the LTF intraday. For intraday trading the whole ladder moves down a step, so that the HTF is daily.

In plain terms

Ask three questions, each from a different distance, and act only when all three answer yes. From far away you ask which way the tide is running, from closer in whether a wave is coming in now, and up close whether this is the moment to step in. The order matters, because a perfect moment to step into an ebbing tide is still a mistake.

2 · The higher timeframe sets the direction

The HTF reads the slope of the 30-bar simple moving average (SMA). Writing $C_t$ for the close of bar $t$ , the SMA of the last 30 closes and its slope over the last $k$ bars are

M_t = \frac{1}{30}\sum_{i=0}^{29} C_{t-i}

(11F.1)

\text{slope}_t = \frac{M_t - M_{t-k}}{k}

(11F.2)

A positive slope permits bullish trades only and a negative slope bearish trades only, while a flat slope permits no trade at all.

The rule does not say how flat is flat. Chapter 11E's stage classifier gives a usable answer: take $k = 5$ , divide the slope by the 14-bar average true range (ATR), and call anything between −0.05 and +0.05 flat. The ATR is the typical size of a bar, so dividing by it lets one threshold serve any price level and any timeframe.

Worked example: Reading the weekly screen

Step 1: the slope. Suppose SPY's 30-week SMA stands at 560.0 today and stood at 552.0 five weeks ago. By (11F.2) the slope is $(560.0 - 552.0)/5 = 1.6$ a week.

Step 2: normalize it. With a weekly ATR of 14.0, the normalized slope is $1.6/14.0 = 0.11$ . That is above +0.05, so the direction is up and only long trades are allowed.

3 · The intermediate timeframe opens the window

The ITF watches the histogram of moving average convergence divergence (MACD), typically on daily bars when the HTF is weekly. Write $E^{(n)}_t$ for the $n$ -bar exponential moving average (EMA) of Chapter 11A, which moves a fraction $\alpha = 2/(n+1)$ of the way toward each new close. The MACD line is the gap between a fast and a slow EMA:

\text{MACD}_t = E^{(12)}_t - E^{(26)}_t

(11F.3)

The histogram is the gap between the MACD line and its signal line, which is the 9-bar EMA of the MACD line, with $\alpha = 2/10 = 0.2$ :

\text{hist}_t = \text{MACD}_t - \text{signal}_t

(11F.4)

The MACD line is positive whenever the fast average sits above the slow one. The histogram asks whether that gap is wider than its own recent average, so it crosses zero exactly when the MACD line crosses its signal line.

A cross above zero opens the bullish window and a cross below zero closes it, so the ITF picks out the moment within the direction the HTF allows.

Worked example: A window opens

Step 1: the MACD line. Today SPY's 12-day EMA is 603.20 and its 26-day EMA is 601.80, so MACD is 1.40.

Step 2: the signal line. Moving a fifth of the way from yesterday's 1.10 toward 1.40 gives $1.10 + 0.2 \times (1.40 - 1.10) = 1.16$ .

Step 3: the histogram. By (11F.4) it is $1.40 - 1.16 = +0.24$ , up from −0.05 yesterday. It crossed zero today, so the bullish window has just opened.

4 · The lower timeframe pulls the trigger

The LTF, typically drawn in intraday bars, waits for two things at once: a pullback of the price to its 21-bar EMA, and the relative strength index (RSI) in the band from 40 to 65. The pullback to the moving average supplies the location. The RSI band confirms that momentum has cooled without breaking, which makes the pair a trigger.

RSI compares recent gains with recent losses. With $\bar{U}$ and $\bar{D}$ the average gain and the average loss over $n$ bars, each smoothed in Wilder's way (Chapter 11C), it is

\text{RSI} = 100 - \frac{100}{1 + \bar{U}/\bar{D}}

(11F.5)

Below 40 the pullback has broken momentum rather than cooled it, while above 65 momentum has not cooled at all. The rule names neither an RSI period nor a distance that counts as touching the average. The figure below supplies both: the 14-bar RSI, and a close within 0.5% of the 21-EMA.

Worked example: A trigger on the hourly chart

Step 1: the location. The hourly 21-EMA stands at 599.74 when SPY closes the hour at 600.40. With $\alpha = 2/22 = 0.091$ , the EMA moves to $599.74 + 0.091 \times 0.66 = 599.80$ , so the close sits 0.10% above it.

Step 2: the momentum. With 14-hour averages of 0.42 for the gains and 0.38 for the losses, (11F.5) gives an RSI of $100 - 100/(1 + 1.105) = 52.5$ , inside the band.

Step 3: the verdict. The weekly slope is up and the daily window is open, so all three screens agree and you enter.

5 · When the screens disagree

Most of the time the screens disagree. That is the point. What to do about a disagreement is settled by a mechanical decision tree, not by judgment in the moment:

Higher	Intermediate	Lower	Action
Flat	Any	Any	No trade. Wait for a direction: the higher screens dominate
Up	Bearish window	Any	No trade. The bullish window is closed, so wait for it to reopen
Up	Bullish window	Waiting	A setup, not an entry. Wait for the lower screen to retrace or confirm
Up	Bullish window	Trigger	Enter. Size by Chapter 13C, and invalidate on a lower-screen close below the 21-EMA
Down	Any	Any	Mirror the rules if shorting is in your mandate. If it is not, no trade

The table is Chapter 11B's hierarchy at small scale, in which a signal on a lower screen never overrides a contradicting higher one. Invalidation is the condition, fixed before entry, that proves a trade wrong and closes it.

6 · Three screens on one tape

The figure draws one synthetic tape of 350 bars at three aggregations, each panel spanning the same stretch of time. The lowest panel shows the raw bars, while the middle one merges every 5 bars into one and the top one every 25, leaving the top panel only 14 bars. So the top screen uses a 10-bar SMA rather than a 30-bar one and reads its slope over the last four bars. It calls the direction up when that slope is above +0.05 a bar, in the tape's price units, down below −0.05 and flat between.

The middle screen reads the sign of its MACD histogram, with the standard 12, 26 and 9 bars. The lowest triggers when the close is within 0.5% of the 21-EMA and the 14-bar RSI is strictly between 40 and 65. A slider picks the tape by its seed, from 1 to 30 and starting at 7, while readouts give each screen's state, a verdict and the numbers behind them. Start at seed 7, then try seeds 9, 18 and 25.

tape seed #7

HTF · direction = UP

ITF · window = BEAR window

LTF · trigger = waiting

One or more screens disagree. No trade. Wait for confluence.

HTF slope +6.35 a bar; ITF MACD histogram −1.13; LTF close 1.43% from its 21-EMA, RSI 43.9.

Three screens on one tape. The same synthetic 350-bar tape at three aggregations, each panel spanning the same stretch of time: 25 bars to one (HTF), 5 to one (ITF) and the raw bars (LTF). The HTF reads direction from the slope of its 10-bar SMA over the last four bars, the ITF reads the sign of its MACD histogram, and the LTF triggers when the close is within 0.5% of the 21-EMA with RSI between 40 and 65. Re-roll the seed: on most tapes at least one screen disagrees.

At seed 7 the top screen is up, with a slope of +6.35 a bar, while the middle screen's histogram of −1.13 marks a bearish window. The lowest screen is waiting, with the close 1.43% from its EMA and an RSI of 43.9. The verdict is no trade until the screens agree, the second row of the decision tree.

Seed 9 is the only one of the 30 tapes on which all three screens agree. Its verdict says enter, "with explicit invalidation per Chapter 13C". The lowest screen triggers inside a closed window at seed 18 and with the window open but the direction down at seed 25. The figure trades only long, so both are no-trades. Seed 19 is the one flat tape, with a slope of +0.04 a bar.

7 · Why the discipline matters

The bias toward action pushes you to trade on a single screen: a daily breakout is exciting, so you act before checking the weekly chart. But the weekly trend was rolling over, which means you bought into distribution disguised as a breakout, the Stage 3 top of Chapter 11E. The triple screen exists specifically to slow you down at the moment you most want to act.

The Dealer Mechanics Lab judges the selectivity this discipline brings to be worth more than any stack of indicators. It presents the triple screen as a discipline rather than as a tested edge, and it cites no study of the rule's returns.

Familiar ground

Requiring all three screens is serial testing, in which a case is called positive only when every test in the series is positive. Specificity rises and sensitivity falls: you miss some good trades in order to avoid many bad ones. The gain in specificity is largest when the tests err independently of each other. Three screens computed from one price series are not independent, since they share the multicollinearity of Chapter 11C. So they filter less than multiplying their false-positive rates would suggest.

In practice

Fix the three timeframes in writing before the session, as Chapter 11B asks you to state the horizon of every trade. Weekly, daily and hourly bars suit a daily swing trader.
Mind the notional. A close below the hourly 21-EMA makes a tight stop, which Chapter 13C's loss budget turns into a large position. With a $200,000 account, a 1% risk budget and a stop $1.60 below a $600.40 entry, the budget allows 1,250 shares: $750,500 of SPY, about 3.75 times the account.
Cap it by buying power. Under Reg-T margin an overnight stock position can be at most twice the equity (Chapter 2E), which caps this one at about 666 shares risking $1,066 at the stop.
An hourly stop does not guard the night. A stop-loss order fills at the first price available after an overnight gap, however far beyond the stop that is (Chapter 5F).

When it fails

A flat higher screen. When the slope hovers near the edge of the flat band, the direction flickers between allowed and forbidden, and the rule whipsaws.
Lag. Every screen is an average of past prices. A sharp reversal therefore shows first on the lower screen and last on the higher one, by which time the move is old.
Gaps. A shock can jump the price through the lower screen's invalidation overnight, leaving a loss larger than the one you sized for.
Shared inputs. Agreement among three screens built from one price series is weaker evidence than agreement among three independent tests.

Decision rule

Direction: trade only in the direction the higher screen's slope points and do nothing while it is flat.
Timing: enter only while the intermediate window is open, from a histogram cross above zero until the next cross below it.
Entry: wait for the lower screen's trigger, a pullback to the 21-EMA with RSI between 40 and 65.
Size and exit: size from the loss budget of Chapter 13C, and close the trade on a lower-screen close below the 21-EMA.
Downtrends: mirror the rules only if shorting is in your mandate, and otherwise stand aside.

Common mistake: trading one screen

The breakout on the daily chart may be real and is still only one screen, so read the weekly slope and the daily window before acting. A breakout against a rolling-over weekly trend is the distribution trap described above.

Common mistake: shopping for a timeframe

If the weekly chart says no, a monthly or a four-day chart may say yes. Switching timeframes until one agrees turns the rule into a search, with the multiple-comparisons problem of Chapter 1D. Choose the three timeframes before the session, and keep them whatever the charts say.

What to carry forward

Three timeframes, three jobs: the higher sets the direction, the intermediate opens the window and the lower triggers the entry. All three must agree.
Each screen reads one number: the 30-bar SMA's slope for the direction, the MACD histogram's sign for the window, and a pullback to the 21-EMA with RSI between 40 and 65 for the trigger.
Most of the time the screens disagree, which means no trade, because a higher screen always overrides a lower one.
The rule's value is selectivity: it slows you down at the moment you most want to act.

8 · Check yourself

Question 1

The weekly slope is up and the daily MACD histogram is −0.3 and falling, while on the hourly chart SPY has pulled back to its 21-EMA with RSI at 50. What do you do?

Answer

Nothing yet, because the intermediate window is closed although the lower screen has triggered: the second row of the decision tree. Wait for the histogram to cross back above zero and take the next trigger after that.

Question 2

A stock's 30-week SMA is 412.0 today and was 411.0 five weeks ago, with a weekly ATR of 9.0. Is the higher screen up?

Answer

The slope is $(412.0 - 411.0)/5 = 0.2$ a week, and divided by the ATR it is $0.2/9.0 = 0.022$ . That is inside the flat band from −0.05 to +0.05, so the direction is flat and there is no trade, although the average is still rising.

Question 3

With a 12-day EMA of 101.5, a 26-day EMA of 101.9 and yesterday's signal line at −0.50, is the daily window open?

Answer

MACD is $101.5 - 101.9 = -0.4$ , which moves the signal line to $-0.50 + 0.2 \times (-0.4 + 0.50) = -0.48$ and makes the histogram $-0.4 - (-0.48) = +0.08$ . The window is open although the MACD line is still negative, because the histogram measures the line against its own average, not against zero.

Question 4

Across the figure's 30 tapes, the top screen reads up on 19, the middle window is open on 20 and the lowest screen triggers on 5. If the screens were independent, on how many tapes would all three agree? The figure finds one: why might the two numbers differ?

Answer

Under independence, $30 \times \tfrac{19}{30} \times \tfrac{20}{30} \times \tfrac{5}{30} = 2.1$ tapes would agree, against the single tape, seed 9, that does. Thirty tapes are too few to tell 1 from 2.1 apart. Nor are the screens independent, since all three are computed from the same prices.