# Indicator Construction and Limitations Explain what an indicator measures and avoid double-counting related signals. Use this workbook alongside the course. Write your answers before opening the solutions. Practical work is self-reviewed; scored knowledge checks are in the Academy. ## 1. Smoothing and trend ### Simple and exponential averages A simple moving average assigns equal weight to a specified trailing window. An exponential average weights recent observations more heavily under its smoothing convention. Record the input price, window and initialization; the indicator name alone does not fully specify the calculation. ### Lag and responsiveness Smoothing reduces short-term variation at the cost of delayed response. A faster average follows noise more closely as well as responding sooner to changes. There is no parameter that eliminates the trade-off in every price regime. ### Warm-up observations An indicator needs enough past observations to initialize. Values near the start of a dataset can depend on the seed convention. Compare implementations only after matching warm-up, missing-data handling and the same source observations. ### Price crossings and whipsaws Price crossings of an average are events, not complete strategies. Repeated crossings in a range can create turnover and costs. Define eligibility, execution timing and exit rules before evaluating a crossing system. ### Worked example For closes 10, 12 and 14, a three-value simple average is 12. Adding 16 and dropping 10 gives 14. The average changes because the trailing window changes, not because future prices have been predicted. ### Independent exercise Calculate the next three-value average after a close of 11 follows 12, 14 and 16. State the retained observations. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 2. Momentum ### RSI inputs and interpretation RSI normalizes smoothed gains and losses into an oscillator. Its value describes recent directional movement under the chosen calculation; a high reading is not a contractual requirement for price to fall next. ### MACD components MACD compares moving averages and commonly includes a signal line and histogram. These components are transformations of related prices. A histogram change and an average crossing should not automatically be counted as independent evidence. ### Stochastic normalization A stochastic oscillator locates a close relative to a recent high-low range. A narrow range or an extreme close can produce a large normalized reading without a large absolute price change. Inspect the denominator and timeframe. ### Extreme readings in persistent trends Persistent trends can keep oscillators near extremes. The interpretation should depend on a tested rule and context rather than a universal overbought-equals-sell shortcut. Record failed reversals in any evaluation. ### Worked example A close is 19 within a lookback range from 10 to 20. Its unsmoothed range position is (19−10)/(20−10) × 100 = 90%. That number locates the close; it does not predict the next return. ### Independent exercise Calculate the range position of 11 in the same range and explain why neither result alone supplies a stop or position size. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 3. Volatility and volume ### ATR units True range considers the current high-low distance and gaps relative to the previous close; ATR smooths that quantity. It is measured in price units unless explicitly normalized. It measures movement magnitude, not a directional forecast. ### Bollinger Band construction Bollinger Bands combine a moving center with a dispersion measure under specified settings. A band touch is not proof of an imminent reversal, and a historical proportion inside bands does not automatically transfer to future market returns. ### VWAP input and session dependence VWAP weights prices by the volume measure available in the chosen session or anchor. The result depends on those inputs and reset boundaries. A tick-activity proxy should be labelled rather than represented as consolidated traded gold volume. ### Tick activity versus traded volume Volume can mean exchange contracts, units traded or provider tick activity. The measure's coverage limits the inference. Compare activity within a consistent feed and avoid claiming it describes all global market participation. ### Worked example A bar has high 105, low 103 and previous close 100. The three true-range candidates are 2, 5 and 3; true range is 5. Using high minus low alone would omit the gap component. ### Independent exercise Calculate true range for high 98, low 95 and previous close 100. Then state whether the result establishes a bullish or bearish forecast. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 4. Combination design ### Correlated indicators Indicators built from similar inputs can be highly dependent. Counting each as a separate vote may create a false impression of confirmation. Identify the underlying information—price trend, volatility or activity—before combining measurements. ### Parameter sensitivity Parameter sensitivity asks whether a conclusion survives reasonable neighboring settings. A single best historical parameter can reflect selection noise. Preserve all tested settings and compare stability rather than reporting only the winner. ### Timeframe alignment Timeframe alignment must respect information availability. A higher-timeframe indicator based on an unfinished bar can change before close. A historical test using its final value earlier introduces lookahead unless the evolving value was actually recorded. ### Rules for conflicting measurements Specify a rule for disagreement: no trade, priority to a particular horizon or another tested response. Choosing whichever indicator supports the desired trade after seeing prices makes the strategy impossible to evaluate honestly. ### Worked example A five-minute entry is justified with the final value of an hourly candle that closes forty minutes later. The historical record contains that final value, but the decision time did not. ### Independent exercise Rewrite the rule using only the last completed hourly value or explicitly recorded developing values. Explain how this affects reproducibility. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## Course project Compare several indicators derived from the same prices and document redundant information. ### Self-review rubric - Concepts and reasoning: 25% - Calculations, data and evidence: 30% - Process and risk controls: 25% - Limitations and communication: 20% Record one correction and one next practice task. This rubric is not automatically graded. ## Worked solutions ### Exercise 1 The retained window is 14, 16 and 11, whose mean is 41/3, approximately 13.67. Using all five values would calculate a different statistic from the specified trailing average. ### Exercise 2 The position is 10%. Both values are normalized descriptions. A stop requires an invalidation and execution assumption; size requires a budget, multiplier and costs. Those quantities cannot be inferred from the oscillator reading alone. ### Exercise 3 The candidates are 3, 2 and 5, so true range is 5. The magnitude calculation does not establish the next direction; it summarizes the movement including the gap. ### Exercise 4 Both alternatives can be specified, but they use different information. Store the observation timestamp and availability rule. A final hourly value cannot be silently substituted for a developing value that may later change. ## Further reading - https://www.tradingview.com/support/solutions/43000502338-relative-strength-index-rsi/ - https://www.tradingview.com/support/solutions/43000501823-average-true-range-atr/