8.2 Technical Indicator Construction

A technical indicator is a derived series. You start with market data—price, volume, open interest, breadth, or a mix—and you apply a recipe (sums, ratios, smoothing, ranges) so that a relationship is easier to see than on the raw chart. The indicator is not the market. It is a lens. On CMT Level I, Internal and External Technical Indicators asks you to know what went into the lens, where it plots, and whether it is being used to follow a trend or to measure momentum—not to memorize every formula in a platform’s library. This OpenExamPrep section is independent study material for those construction topics.

What an indicator is

Construction always has four choices, even when a platform hides them:

1. Inputs. Close-only, high-low-close, typical price, volume, true range, or a prior indicator.
2. Window and smoother. Finite SMA or LWMA versus recursive Wilder or EMA; length n.
3. Transform. Difference, ratio, range location, cumulative signed volume, directional move.
4. Display and scale. Unbounded (MACD), bounded 0–100 (RSI, stochastic, ADX-style), or price units (overlays).

Change any choice and you have a different study, even if the name on the screen stays “RSI.” Exam items test whether you know the logic of the famous recipes, not whether you can expand every summation sign.

Internal indicators are built from the market’s own tape (price, volume, breadth). External indicators (sentiment surveys, fund flows, COT positioning) come from outside the price pane; those appear in the sentiment unit. This section’s building blocks are still internal: price, range, and volume assembled into studies you will also meet as “external” only when someone overlays a non-price series.

Two category framings (keep them separate)

Candidates mix two different taxonomies. The exam can use either. Teach both, and do not fuse them into one false dichotomy.

Framing 1 — Display: overlay versus oscillator

• Overlays sit on the price pane and share the price scale: moving averages, envelopes, Bollinger Bands, Parabolic SAR, sometimes volume-weighted average price (VWAP). You compare them directly with highs, lows, and closes.
• Oscillators (and most separate-pane studies) sit in a lower pane with their own scale: RSI, stochastic, MACD, many volume accumulators, ADX. You do not read “RSI = 70 dollars.” You read 70 on a 0–100 scale, or MACD in points of the EMA spread.

MACD is an oscillator even though it is built from moving averages. An SMA is an overlay even though it is an “indicator.” Display category is about where it plots, not about whether moving averages were used in the recipe.

Framing 2 — Role in time: lagging/trend-following versus leading/momentum

• Trend-following (often called lagging) studies confirm a path that is already in motion: moving-average direction and crossovers, DI+ / DI− dominance, a rising ADX. They are late on purpose. They earn their keep by staying with a persistent move and by keeping you off the opposite side.
• Momentum (often called leading) studies measure the speed or internal strength of the move: RSI, stochastic, rate of change, the MACD histogram. In a swing, momentum often crests or troughs before price prints its final extreme. That is all “leading” means here.

Critical caution: every one of these calculations uses past and present data. Nobody’s RSI is looking into the future. “Leading” is a statement about typical timing of turns in cyclical markets, not about peeking ahead. In a strong one-way trend, momentum can stay “overbought” or “oversold” for a long time; treating a high RSI as an automatic reversal is the classic misuse of a leading label.

Some tools wear both hats. MACD’s line-versus-signal cross is a trend-following rule; the histogram’s rise and fall is a momentum read. ADX is trend-following about strength and is silent on direction.

Momentum indicators at definition level

Do not dump platform code. Know the question each study asks.

Relative Strength Index (RSI) — Wilder. It asks: of the smoothed upward close-to-close changes and smoothed downward changes over n bars (conventionally 14), how dominant are the up days? The result is scaled 0 to 100. High readings mean recent closes have been persistently strong, not that a crash is due tomorrow. Construction logic: Wilder-smooth average gains and average losses, form a relative-strength ratio, then compress it onto 0–100. RSI is not the comparative relative strength of one stock versus an index; that later “RS” topic is a ratio of two prices.

Moving Average Convergence/Divergence (MACD) — Gerald Appel. It asks: what is the spread between a faster EMA and a slower EMA of price? The MACD line is that spread (common teaching lengths 12 and 26). A signal line is an EMA of the MACD line (common length 9). The histogram is MACD minus signal. When the spread widens in the direction of the trend, momentum is supporting the move; when the spread narrows, the trend is losing internal speed even if price has not yet reversed. MACD is unbounded. It can live in a lower pane as an oscillator while still being built from overlay-style EMAs.

Stochastic oscillator — George Lane. It asks: where is the close inside the high-low range of the last n bars (often 14)? %K is that location as a percent: (close − lowest low) / (highest high − lowest low) × 100. %D is a short smooth of %K. Readings near 100 mean the close is at the top of its recent range; near 0, at the bottom. Like RSI, that is range location, not a calendar forecast.

Contrast the three: RSI compares up versus down close changes (Wilder). MACD compares two exponential means of price. Stochastic compares close versus the recent high-low box. All three are momentum tools in ordinary display; none replaces a trend filter by itself.

Volume indicators: OBV and A/D

Price says what printed; volume says how much participation accompanied it.

On-balance volume (OBV) — Joseph Granville. Construction logic: keep a cumulative line. If today’s close is above yesterday’s close, add today’s volume. If the close is below, subtract today’s volume. Unchanged close, add nothing (or follow the platform’s tie rule). OBV asks whether volume is confirming the direction of closes. Rising price with falling OBV is a participation warning; rising OBV with sideways price can be accumulation language. There is no bounded 0–100 scale.

Accumulation/distribution (A/D) in the Chaikin-style line used in many textbooks does not sign the entire bar’s volume by close-to-close direction. It asks: where did this bar close inside its own high-low range? A close on the high assigns most or all of the volume as accumulation; a close on the low assigns it as distribution; a mid-range close assigns little net. Those signed pieces accumulate into a line. A/D can therefore rise on a down-close bar if the close was strong inside that bar’s range—a different story from OBV.

If the high equals the low, the range location is undefined; platforms skip or hold the line. That edge case is less important than the construction contrast: OBV signs volume by close versus prior close; A/D fractions volume by close versus this bar’s high and low.

Price-plus-volume hybrids

Hybrids scale volume by a price relationship instead of treating volume as a naked count. Examples at Level I definition level:

• A/D itself is already a hybrid (range location × volume).
• Chaikin Money Flow averages that A/D idea over a window (often 21 bars) and scales by total volume, producing an oscillator around zero.
• Price-volume trend and related money-flow studies multiply volume by the percent change in price (or by typical price) and accumulate.
• VWAP (and anchored VWAP) is a volume-weighted average price from a session or from a chosen anchor—more of an overlay benchmark than an oscillator, but still price and volume fused. Anchored VWAP is called out in the digital curriculum’s volume discussion; here, remember it as a hybrid overlay, not as RSI.

The construction idea to remember: price locates the vote; volume sizes the vote. Hybrids keep both.

DMI: three plots, interpretation, and uses

Wilder’s Directional Movement Index (DMI) is a trend system with three plots, conventionally Wilder-smoothed over 14 periods.

+DI (DI+) measures the share of true range that came from upward directional movement. −DI (DI−) measures the share that came from downward directional movement. ADX (Average Directional Index) is a Wilder smooth of DX, and DX is the normalized absolute difference between DI+ and DI−. In words: ADX asks how spread apart the two directional lines are, then smooths that spread. ADX does not carry a sign. A reading of 30 can be a powerful uptrend or a powerful downtrend.

Construction logic (not a formula dump):

• Compare this bar’s up-move (this high minus prior high) with this bar’s down-move (prior low minus this low).
• Award +DM only if the up-move is larger and positive; award −DM only if the down-move is larger and positive; inside bars can award neither.
• Smooth +DM, −DM, and true range with Wilder’s 1/14 factor.
• DI+ and DI− are those smoothed directional moves as a percent of ATR (smoothed true range).
• DX converts the DI gap into a 0–100 strength number; ADX smooths DX.

General interpretation

• DI+ above DI−: upside directional pressure has the edge.
• DI− above DI+: downside pressure has the edge.
• ADX rising from a low level (often taught as rising from below about 20): a trend is strengthening; use the DI lines for direction.
• ADX falling: directional spread is shrinking; conditions look more like a range or a tired trend.
• DI crossovers are directional signals; they fail more often when ADX is low.

Uses on Level I

• Trend filter: allow trend-following entries only when ADX is above a threshold (commonly taught around 20–25) so you are not firing MA crossovers in a coil.
• Direction: trade the side of the leading DI line while ADX confirms strength.
• Stand-aside: a collapsing ADX after a high reading warns that the trend-following edge is fading even if price has not yet fully reversed.

DMI is lagging by design. It will not catch the first bar of a new trend. Its job is to say whether a directional move has enough persistence to bother with. That is the same lag philosophy as a moving-average filter, built from true range and directional moves instead of from a mean of closes.

Read the three DMI plots as a team. DI+ versus DI− answers which side. ADX answers how much trend. Mixing those jobs—treating ADX like RSI overbought, or treating DI+ like a volume line—is the usual Level I error.