Technical Indicators Explained: How to Choose and Audit a Signal

Start with a decision, then choose a measure

A technical indicator compresses price, volume, or both into a number or line. It does not create information beyond its inputs. Before adding one to a chart, write the question it should answer: Is the trend rising? Is momentum improving? Is volatility expanding? Is participation broadening? Different questions need different measurements. Three momentum oscillators with similar inputs may look like independent agreement while repeating the same evidence.

The current CMT Program Guide places moving averages, indicator construction, directional movement, and Bollinger Bands among Level I learning objectives. Level II extends interpretation to momentum, VWAP, Bollinger applications, and price confirmation. That verifies curriculum relevance, not the frequency or exact form of any future question. This legacy guide focuses on how to select and audit tools. For formulas and comparisons, see Technical Indicators for the CMT Exam.

Match the tool to the market question

An SMA gives equal weight to each observation. A four-close SMA for 100, 102, 104, 110 is (100 + 102 + 104 + 110)/4 = 104. An EMA gives more weight to recent closes. For an EMA with a four-period smoothing factor, alpha = 2/(4 + 1) = 0.4. If yesterday's EMA is 103 and today's close is 110, today's EMA is 0.4 × 110 + 0.6 × 103 = 105.8. The EMA reacts faster in this example, but faster reaction can mean more false turns in sideways trading. Always state which average and parameter produced a signal.

Separate observation from inference

Suppose a stock closes at 110, above a rising 20-period average of 105. The observation is that the current close is five points above that average. “The trend will continue” is an inference. Before acting, ask whether the 20-period average was chosen before seeing the chart, whether the instrument is range-bound, and whether the distance from the average is unusual for its recent volatility.

A momentum reading needs the same care. Consider five closes of 100, 103, 106, 105, and 109. Four-period price rate of change is (109/100 − 1) × 100 = 9%. If the next two closes are 110 and then 112, the four-period comparison for 112 is the earlier 106 close, giving (112/106 - 1) x 100 = about 5.66%. Price made a higher close, but the rate of change slowed. This is a divergence in this constructed sequence, not proof of a downturn. An analyst can require a support break or a change in trend structure before changing a position.

RSI readings above 70 and below 30 are commonly used reference levels, but they are not automatic sell and buy orders. An upward trend can sustain high RSI. A MACD line is conventionally a shorter EMA minus a longer EMA; its signal-line cross is sensitive to parameter choice and to sideways price action. Know what subtraction or smoothing produced the line before interpreting its sign.

A band example with explicit math

Take a four-observation population of closes: 98, 100, 102, 100. The mean is 100. Squared deviations are 4, 0, 4, and 0, so the population variance is 8/4 = 2 and standard deviation is about 1.414. A demonstration band at mean ± two standard deviations is approximately 97.17 to 102.83. If the next close is 103, it sits just above the previous band. That is a location statement, not a short signal. A trend can walk along a band; a low-volatility band can expand abruptly. Real platforms may use a rolling lookback and a specified standard-deviation convention, so reproduce the platform settings when comparing results.

One useful decision rule is: mark an “expansion watch” when price closes outside the band and the band's width rises; require a separate price-structure confirmation before acting. Invalidate the watch if price falls back into the prior range and volume or momentum fails to support the move. These are example research rules, not official CMT rules or investment advice.

Build a signal card

For any indicator, write six lines:

1. Input: prices, volume, adjustment policy, interval, and lookback.
2. Calculation: formula or platform implementation.
3. Observation: what the current value literally measures.
4. Context: trend, range, volatility, and nearby price levels.
5. Action condition: the event that changes your decision.
6. Failure condition: the event that tells you the interpretation was wrong.

A card prevents an indicator from turning into a decorative line. It also lets another analyst repeat your reasoning. For a long signal, a candidate might require the close above a prior high, a rising average, and a protective exit beneath a recent swing low. The exit is part of the rule, not an afterthought.

Common mistakes

A 20-day calculation on daily closes is not comparable with a 20-bar calculation on hourly data. Unadjusted stock splits can produce false indicator jumps. Comparing a new volatility reading with an arbitrary fixed threshold ignores changes in regime. A short and long average derived from the same prices are correlated; their agreement should not be counted as two independent facts. A “best” parameter found after testing many combinations may reflect the sample rather than a durable edge. See Statistical Analysis for the CMT Exam for the testing problem.

Self-check

A five-period SMA yesterday was 50. The oldest price in its window was 46, and today's new close is 56. What is today's SMA? Answer: The five-close sum was 5 × 50 = 250. Remove 46 and add 56 to get 260; 260/5 = 52. A rising SMA confirms that this smoothed value increased, but it does not tell you whether the new close is a good entry.

Next, select one chart, define a single question, and write a six-line signal card. Then compare your card with the relevant official learning objectives and test it on both a trending and a range-bound segment.