Trading Indicators and Tools

Indicator Duplication: How to Tell When Several Indicators Measure the Same Thing

Learn how correlation, signal agreement and ablation tests reveal duplicate trading indicators and help build a clearer decision system.

Indicator Duplication: How to Tell When Several Indicators Measure the Same Thing

Indicator duplication occurs when several indicators derive such similar information from the same market data that they repeatedly confirm a conclusion already reached. You cannot reliably identify it from the appearance of the lines or the indicator names alone. You need to examine their input data, calculation logic, response speed and actual effect on trading decisions.

In practice, three tests are needed: mapping indicator functions, measuring agreement between their values and signals across different market conditions, and comparing the system with and without each indicator. The aim is not to reduce the number of indicators at any cost. Every indicator retained should have a clear and distinct role.

What Indicator Duplication Actually Means

The fact that two indicators use price does not necessarily mean they duplicate each other. Duplication arises when they repeatedly represent the same information dimension within a decision system. For example, several indicators may describe short-term price momentum in different ways. If all three authorise a trade at almost the same time, three confirmations may simply be one confirmation presented in three visual forms.

It is useful to distinguish four types of duplication:

  •          Input-data duplication. Indicators use the same price field, period or very similar windows.
  •          Mathematical duplication. Different formulas produce closely related numerical series.
  •          Signal duplication. The values look different, but the buy, sell or filter signals coincide.
  •          Decision duplication. Two indicators perform the same role in a strategy—for example, both confirm momentum—and the second one does not change any final decision.

The last type is the most important. Mathematically correlated indicators can be useful if one identifies the market regime while the other refines entry timing. Conversely, indicators with low overall correlation may duplicate each other precisely during the rare moments when the strategy opens a position.

Why Multiple Confirmations Can Create False Confidence

Technical analysis indicators are mostly transformations of price, volume or both. In the official TA-Lib function list, indicators are grouped into categories such as momentum, overlap studies, volatility and volume. These categories help define indicator roles, but they do not by themselves guarantee independent information.

For example, a moving average and MACD do not use the same formula, but both relate to smoothed price dynamics. RSI and the stochastic oscillator use different calculations, yet both can respond to short-term momentum and end up serving a similar filtering role. Three similar momentum confirmations do not make one market observation three times more independent.

Duplication creates several practical problems:

  •          one market characteristic receives disproportionate weight;
  •          the system becomes more complex, but not necessarily more informative;
  •          fewer trades occur because several similar filters must agree;
  •          a seemingly precise combination of indicators may be fitted to historical data;
  •          when signals conflict, it is unclear which indicator takes priority.

This does not mean that only one indicator may be used from each category. A different calculation period, timeframe or point of use can provide a distinct function. However, that distinction must be demonstrated through system rules and data, not indicator names.

Start with a Role Map, Not a Correlation Table

Before conducting a statistical test, write down the question each indicator is intended to answer. If an indicator does not have one clear question, its place in the system is already open to doubt.

Indicator role

Question it should answer

Possible output

Market regime

Is the market trending or ranging?

trend, range, uncertain

Direction

In which direction are trades allowed?

long, short, neutral

Momentum

Is the movement currently accelerating or weakening?

confirms, does not confirm

Volatility

How wide are current price fluctuations?

number or regime

Entry timing

Has the trigger occurred at this specific moment?

yes, no

Risk parameter

What distance or position size should be used?

number

Then document each indicator’s source data, period, timeframe, normalisation and exact rule. Writing “RSI confirms the trade” is not enough. Specify which value is used, whether the rule evaluates a level or a level crossing, and whether the decision is made only after the candle closes.

This table often reveals duplication before any calculations are performed. If three indicators have “confirms momentum” in the role field, there is no need to judge which looks better. Test whether the second and third indicators change the final decision.

How to Measure Agreement Between Values and Signals

1. Compare Like-for-Like Observations

All data series must relate to the same instrument, timeframe and timestamps. You must also decide whether only closed candles will be used. Otherwise, correlation may be distorted by missing data, different trading hours or indicator values that are still changing during an active candle.

Indicators with different scales can be transformed into a comparable form—for example, by standardising them or converting them into directional states. Retain the original values so that the transformation does not conceal a meaningful difference.

2. Calculate Correlation, but Do Not Make It the Only Criterion

The Pearson correlation coefficient ranges from −1 to +1 and describes the direction and strength of a linear relationship. NIST also explains an important limitation: low correlation does not by itself prove that no relationship exists between the variables, because the relationship may be non-linear.

Alongside Pearson correlation, you can therefore examine rank correlation and agreement between indicator directions. Do not use a single universal threshold to delete indicators automatically. The result must be assessed together with the indicator’s role and the strategy’s decisions.

Correlation should also be examined in a rolling window. A full-period average can hide a situation in which indicators nearly duplicate one another during trends but behave differently in ranges. Changes in market regime often determine whether a seemingly independent filter genuinely adds information.

3. Measure the Final Signals

If the strategy converts an indicator into a binary condition, compare the conditions themselves as well as the continuous values. For each candle period, you can record whether:

  •          both indicators allow the trade;
  •          both reject it;
  •          the first allows it and the second rejects it;
  •          the second allows it and the first rejects it.

Pay particular attention to disagreements. If they almost never affect an entry, exit, position size or decision to skip a trade, the second indicator may be only a visual confirmation. Conversely, if disagreements are concentrated in a particular market regime, this may indicate a separate function—provided it protects against a specific system risk.

4. Test Time Lags

Two indicators may measure a similar phenomenon but respond with different delays. Contemporaneous correlation does not always reveal this. Compare whether the current value of one indicator matches the other indicator’s value one or more candles later.

A lagged version is not automatically redundant. A slower indicator may act as a regime filter, while a faster one serves as an entry trigger. These roles must, however, be defined in advance. Otherwise, the delay may merely provide late but psychologically reassuring confirmation.

The Ablation Test: Does the Indicator Actually Change the System?

The most reliable practical method is an ablation test. Create a baseline version containing all the rules, then create variants in which each indicator is removed individually. Do not change the other conditions, so it remains clear which change caused the difference.

In addition to final profit, compare:

  •          the number of trades and their distribution over time;
  •          maximum drawdown;
  •          average result per trade;
  •          the distribution of losses and profits;
  •          results across different market regimes;
  •          the effect of trading costs and slippage;
  •          which specific trades were added or skipped.

If removing an indicator barely changes the set of trades, it is probably not an independent decision component. If the trades do change, determine whether the effect remains stable outside the data used to develop the rules. A small improvement in historical results is not sufficient justification for additional complexity.

The test should be conducted both on the data used to develop the system and in a separate out-of-sample period. It should later be followed by a forward test using unchanged rules. This does not guarantee future performance, but it reduces the risk of choosing an indicator merely because it fitted one historical period well.

A Practical Example with Four Indicators

Suppose the chart contains two moving averages, MACD, RSI and ATR. At first, the system may appear to contain five separate confirmations. The role map presents a different picture:

  •          the position of the moving averages determines the trend direction;
  •          MACD also describes the relationship between smoothed price series and is used to confirm direction;
  •          RSI evaluates momentum and is used as an entry filter;
  •          ATR describes the range of price fluctuations and determines the stop-loss distance.

The first step is to test the overlap between the roles of the moving averages and MACD. If MACD rejects only the trades already rejected by the relationship between the averages, it is not an additional vote. RSI may differ in its formula, but it must also be tested at the decision level. The function of ATR is different because, in this system, it does not confirm direction but parametrises risk.

This example does not imply that MACD or RSI should always be removed. The conclusion is narrower: an indicator’s value is determined by its function within specific rules, not by its popularity or position on the chart panel.

When Similar Indicators Are Still Justified

Two related indicators may remain in a system if they have separate responsibilities. For example, a longer-period indicator on a higher timeframe may determine the permitted direction, while a faster indicator on a lower timeframe identifies the entry point. They may be mathematically related yet affect different stages of the decision.

A backup control can also be a valid exception. In systematic execution, one calculation may be used to detect a data or implementation error rather than provide an additional view of the market. In that case, it is a technical control mechanism, and its result should not be counted as an independent trading confirmation.

To justify retaining an indicator, you should be able to answer three questions: which unique decision does it change, under what conditions is it valuable, and what happens when it is removed? If the answers cannot be tested, the complexity probably exceeds the benefit.

Limitations and Risks

Correlation and historical testing do not solve every problem. Relationships between indicators change with the instrument, timeframe and market regime. Signals that are independent in one period may begin moving together in another. An audit is therefore not a one-off certificate. It should be repeated periodically using a predefined methodology.

Repainting risk must also be checked. TradingView documentation uses “repainting” to describe situations in which a script behaves differently on historical and real-time data. If an indicator uses an unclosed candle, incomplete higher-timeframe data or future information, historical signal agreement may be misleading. The comparison must reflect how the signal would actually have been available at the time of the decision.

Other important limitations include trading costs, slippage, an insufficient number of observations and testing multiple variants until the desired result is found. The more combinations tested, the greater the risk of mistaking random historical coincidences for a stable edge.

This audit evaluates information overlap; it does not predict market direction. Even a well-differentiated set of indicators does not guarantee profit or replace controls for position size, loss limits and order-execution risk.

A Short Indicator Audit Checklist

Before retaining an indicator in the system, check:

  1. Does the indicator have one clearly defined role?
  2. Which data, period and timeframe does it use?
  3. Does another indicator already answer the same question?
  4. Is the correlation between values stable, or does it change across market regimes?
  5. How often do the indicators’ final signals disagree?
  6. Do these disagreements actually change the trade?
  7. What happens to the set of trades and risk when the indicator is removed?
  8. Does the result persist in out-of-sample and forward testing?
  9. Does the calculation involve repainting or the use of future data?
  10. Is there a clear, documented benefit to the additional complexity?

In a good set of indicators, simultaneous agreement between lines is not an end in itself. Each component answers a different necessary question, and its effect can be tested. If removing an indicator does not change decisions or risk, a simpler system will generally be easier to execute, analyse and maintain.