Technical analysis tools and indicators: which can you test?

Contents

1. What are the main families of technical indicators?
2. How does an indicator become an entry and exit rule?
3. Is there a best indicator for trading?
4. Why can't you judge an indicator by looking at a chart?
5. How do you combine indicators without redundant signals?
6. What can an indicator test tell you, and what can't it?
7. Frequently asked questions
8. References

What are the main families of technical indicators?

Most technical indicators fall into four families: trend, momentum, volatility and volume. No indicator is best in general. You can test any of them once you write exact entry and exit rules with parameters, include trading costs, and check the result on data the rules never saw.

A technical indicator is a calculation on past prices or volume that is drawn on a chart, such as an average of the last 20 closes. Each family answers a different question about the market. The table lists common members of each family and the job each family usually does inside a rule.

Family	Question it answers	Common examples	Typical job in a rule
Trend	Which way is price moving, and how strongly?	Moving averages, ADX, Donchian channel, Supertrend, MACD	Filter: only take long trades while the trend points up
Momentum	How fast did price change recently?	RSI, Stochastic, Williams %R, CCI, rate of change	Timing: enter after a short pullback or a burst of strength
Volatility	How large are the moves?	ATR, Bollinger Bands, Keltner channel, standard deviation	Risk: set the stop distance, position size or breakout width
Volume	How much trading took place?	OBV, VWAP, Chaikin money flow, volume profile	Confirmation or a reference price within the session

Some indicators sit in two families. MACD is built from moving averages, so it measures trend, but traders also read it as momentum. Other technical analysis tools, such as trend lines, Fibonacci levels and chart patterns, are usually drawn by hand. They can be tested too, but only after you write down exact rules that a computer can follow without your judgment.

Each indicator also has typical strengths and failure points on short timeframes. Our guide to best indicators for day trading: what each does and where it fails covers those one by one. This article covers how to test any of them.

How does an indicator become an entry and exit rule?

An indicator becomes testable when you write it as a rule with exact conditions, an exact action and exact numbers. "RSI looks oversold" cannot be tested. "Buy at the next bar's open when the 14-bar RSI closes below 25" can be tested. A parameter is a number in the rule that you choose, such as the RSI length or the level of 25.

1. Choose the indicator and its length. For example, a 14-bar RSI.
2. Write the condition. Use a level ("closes below 25"), a crossing ("fast average crosses above slow average") or a comparison ("close above the upper band").
3. Write the action and the fill price. If the condition is checked at the bar's close, fill at the next bar's open. A rule that fills at the same close it used to decide trades at a price you could not have traded.
4. Write the exit. Use an opposite signal, a fixed number of bars, a stop based on ATR, or a profit target.
5. Set a range for each parameter. For example, RSI length from 5 to 30.
6. Add costs and the trading session. Commission, slippage and the hours the rule may trade belong in the first test, not in the last.

The number of variants grows fast. Testing 270 variants by hand, each with costs and an out-of-sample check, takes far longer than most traders can spend. For that reason, systematic traders let software write and run the rules. In AlphaPrime (Valynox), strategy generation turns research ideas into testable trading rules. You set the rules and indicators, the parameter ranges, the entry and exit conditions and the search constraints.

Is there a best indicator for trading?

No single indicator is best, because a test result belongs to a whole rule on one market, one timeframe and one cost setting. It does not belong to the indicator. The same RSI can lead to a useful rule on one contract and a losing rule on another. The table shows the factors that change the result.

Factor	Why it changes the result	What to check
Market	Some markets trend for long periods; others move back and forth in a range. A rule that buys weakness does well in a range and can lose for months in a strong trend.	Test each market separately before you trust a rule on several markets.
Timeframe	Shorter bars produce more trades with smaller moves per trade, so costs take a larger share of each trade.	Compare the average trade with the round-trip cost.
Costs	A small edge per trade disappears after commission and slippage.	Rerun the test with double the slippage.
Period	High-volatility and low-volatility years reward different rules.	Look at results year by year, not only the total.
Role in the rule	An indicator used as a filter behaves differently from the same indicator used as the entry trigger.	Test each role as its own variant.

The cost check matters most on short timeframes. Our article on trading costs in backtests: commission, slippage and the average trade shows how to set them. Running every indicator variant through backtesting with fees and slippage in AlphaPrime keeps the cost check the same across short-timeframe rules.

Why can't you judge an indicator by looking at a chart?

Judging by eye fails because people remember the signals that worked and skip the ones that did not. It is like judging a weather forecaster only by the sunny days you remember. A chart example in a book or video is usually chosen because the signal worked. A rule tested on every bar counts every signal, the good and the bad.

Testing every signal is still not enough if you choose the parameters on the same data. In-sample data is the part of the history used to build and tune the rule. Out-of-sample data is history that the rule did not see while you built it and tuned its parameters. A rule that keeps a reasonable part of its in-sample result on out-of-sample data is more believable than one that does not. Walk-forward analysis repeats this split many times across the history.

Data snooping is the error of testing many variants on the same data and then treating the best one as if it were the only test. White (2000) proposed a test that asks whether the best of many models still beats a benchmark once the size of the search is counted. Sullivan, Timmermann and White (1999) applied this idea to a large set of technical trading rules. The best rule from their original sample did not keep its performance in the later period they tested. The Sharpe ratio is the average return divided by how much returns swing up and down. Harvey and Liu (2015) show how to reduce a backtest's Sharpe ratio to account for the number of tests behind it.

Check	Good sign	Bad sign
Out-of-sample result	Keeps a reasonable share of the in-sample result	Flat or negative while in-sample looked strong
Neighbouring parameters	Length 12, 14 and 16 give similar results	Only one exact value works
Number of trades	Enough trades that one or two big trades do not decide the total	A handful of trades, or the profit comes from one trade
Costs	Still positive with double slippage	Turns negative with a small cost increase

How do you combine indicators without redundant signals?

Combine indicators from different families, because indicators from the same family mostly repeat each other. A redundant signal is a second condition that is almost always true when the first one is true, so it adds a parameter but no new information. RSI and Stochastic are both momentum calculations on the same prices. Asking both is like asking two friends who read the same newspaper.

1. Give each indicator one job. A common structure uses a trend indicator for direction, a momentum indicator for timing and a volatility indicator for the stop or size.
2. Measure the overlap. Count the bars where condition A is true. Then count how many of those bars also have condition B true. If B is almost always true together with A, B is redundant.
3. Test with and without the extra condition. Keep it only if the out-of-sample result improves, not only the in-sample result.
4. Watch the trade count. Each filter removes trades. Too few trades leave a result that luck can explain.
5. Count the parameters. Each new indicator multiplies the variants, as in the example above, and gives the search more room to fit noise.

What can an indicator test tell you, and what can't it?

An indicator test tells you how one exact rule behaved on past data after costs, and on data it did not see. It does not tell you how the rule will behave next year. Treat a passed test as a reason to keep studying the rule, not as proof that it works.

• It can show whether a rule had an edge after commission and slippage, whether that edge held on out-of-sample data, and whether nearby parameter values behave the same way.
• It can show whether an added indicator changed the result or only added a parameter.
• It cannot show that the market will keep behaving as it did. Markets change, and a rule that held for years can stop working.
• It cannot rate the indicator itself. A failed RSI rule says that this rule failed on this market. It does not say that RSI is useless everywhere.
• It cannot undo a large search. If you tested hundreds of variants and kept the best, part of its result is luck, even after an out-of-sample check.

AlphaPrime's strategy validation runs these checks with the same standards for every candidate. It offers Basic, Standard and Extensive validation, walk-forward analysis and thresholds that you set.

What this means if you use AlphaPrime: write your indicator ideas as rules with parameter ranges, let AlphaPrime evaluate every variant with fees, slippage and next-bar execution, and keep only the rules that hold up on data they never saw.

Frequently asked questions

What are the most popular technical analysis indicators?

Moving averages, RSI, MACD, Stochastic, Bollinger Bands, ATR and VWAP are among the most widely used. They cover the trend, momentum, volatility and volume families. Popularity shows how often traders use an indicator, not whether a rule built on it makes money after costs.

Which technical indicator is the most accurate?

No indicator is accurate in general, because an indicator only describes past prices. What people call accuracy is usually the win rate of a rule. A win rate means little without the size of wins and losses and the costs. Compare rules by their result after costs on out-of-sample data.

Are free technical analysis tools good enough?

Most charting platforms include the standard indicators, and the calculations are well known, so the indicator itself is rarely the problem. The harder part is the test. You need exact rules, realistic costs and a check on data the rules did not see.

How many indicators should a trading strategy use?

Use as few as the idea needs, with each indicator doing a different job. A common structure uses one indicator for direction, one for timing and one for risk. Every extra indicator adds parameters and removes trades, so keep it only if it improves the out-of-sample result.

Can technical indicators predict price?

Indicators do not predict. They summarise past prices and volume in a form that a rule can use. A test can show that a rule had a statistical edge in past data, but it cannot promise that the edge will continue.

References

• Harvey, C. R., & Liu, Y. (2015). Backtesting. The Journal of Portfolio Management, 42(1), 13–28.
• Sullivan, R., Timmermann, A., & White, H. (1999). Data-Snooping, Technical Trading Rule Performance, and the Bootstrap. The Journal of Finance, 54(5), 1647–1691.
• White, H. (2000). A Reality Check for Data Snooping. Econometrica, 68(5), 1097–1126.