New Insights into the Relative Strength Index: A Decade of Analysis
Written on
Chapter 1: Introduction to RSI Strategies
This article aims to evaluate the effectiveness of the Relative Strength Index (RSI) strategy by comparing it with two alternate versions over a ten-year span of intraday FX data. Theoretically, a widely-followed indicator like the RSI should carry more weight due to the number of traders acting on its signals. However, does this guarantee that the RSI will outperform its variations? Let's delve into how selective modifications and the addition of filters to the RSI can influence both predictability and profitability.
The Fibonacci Trading Book is now available! This comprehensive guide features Fibonacci-based trading techniques—tools, indicators, patterns, and strategies—to enhance your trading and analysis skills. For more information, see the PDF version below.
Intuition Behind the Research
Trading strategies are diverse, ranging from fundamental analysis to technical methods and hybrid approaches. We often face the question, "Which strategy is most effective and when?" Take the RSI, for instance, which relies on price momentum and is generally employed to generate contrarian signals during extreme readings. Determining the reliability of these signals is inherently challenging, making back-testing results essential for confirmation.
Default technical indicators seldom yield positive returns consistently. By applying filters and adjustments, we can enhance strategies to minimize poor signals and improve profitability, thus serving as a risk management tool to avoid low-confidence trades.
The objective of this study is to back-test a strategy involving three types of RSI to compare their performance. The strategy dictates that a long position should be initiated when RSI values are oversold, while a short position should be established when RSI values are overbought. An oversold condition exists when bearish momentum drives the RSI toward lower limits, whereas an overbought condition emerges when bullish momentum pushes it to upper limits.
This strategy will be tested on five major currency pairs. For a fair comparison, transaction costs will be excluded, as the goal is not to identify a deployable profitable strategy but to answer the following questions:
- Which RSI strategy performed best over the past decade?
- Which one exhibited a superior hit ratio?
The back-test will utilize hourly FX data from 01/01/2011. The hourly timeframe is chosen for its data richness and signal availability, allowing us to examine intraday strategies effectively. Additionally, lower timeframes offer a theoretical advantage in technical analysis, often overshadowed by fundamental analysis in higher timeframes due to other influencing variables.
We will evaluate three indicators, discussed in detail in subsequent sections: the standard 14-period RSI, a modified RSI using a simple moving average instead of a smoothed moving average, and a correlation-adjusted RSI (CARSI) that filters signals based on specific criteria. When back-testing a strategy, we must also consider exit techniques.
Our analysis will explore market reactions once the RSI signals overbought or oversold conditions, leading to two exit methods:
- 1-hour holding period: This method assumes that a quick market reaction follows an RSI signal, with the algorithm entering a trade and exiting at the next closing price after one hour. This is termed the fixed period strategy.
- Holding until a new signal: In this approach, a position is maintained until a subsequent signal arises. For example, if the RSI hits 30, we buy and hold until it reaches 70 (for a short position) or dips below 30 again (for a long position). This is referred to as the variable period strategy.
As risk management relates to individual trader preferences, we will take a neutral stance during our back-test. The performance metrics analyzed will include:
- Hit ratio: The proportion of profitable trades to total trades.
- Gross return: The overall return achieved from the strategy.
- Risk-reward ratio: The average gain compared to the average loss.
- Total trades: The aggregate number of closed positions, indicating trade frequency.
Chapter 2: Strategy Comparisons
The first strategy under consideration is the Relative Strength Index (RSI).
The RSI, developed by J. Welles Wilder Jr., is a widely-used technical indicator, primarily functioning as a contrarian tool where extreme values indicate potential market reactions. The calculation of the default 14-period RSI involves the following steps:
- Calculate the change in closing prices from prior periods.
- Separate positive net changes from negative ones.
- Compute a smoothed moving average for both positive and negative net changes.
- Divide the smoothed positive changes by the smoothed negative changes to derive the Relative Strength (RS).
- Apply the normalization formula to obtain the RSI.
Notably, Wilder's smoothed moving average is designed to be smoother than a standard moving average. The trading conditions dictate that:
- A long position is initiated when the RSI closes below 30, while the prior value was above 30.
- A short position is activated when the RSI closes above 70, with the previous value below 70.
The exit strategy will consist of two methods, as outlined earlier, with detailed results for five currency pairs presented in the next section.
The PDF link for the Fibonacci Trading Book is available below (include your email in the notes section):
The second strategy is the Modified RSI.
The modified RSI employs a simple moving average instead of the smoothed variant, resulting in a more responsive indicator that produces a higher signal frequency due to increased volatility. Although the natural thresholds for the modified RSI might extend beyond the typical 70 and 30, we will maintain these limits for consistency across back-tests.
The performance of the modified RSI will be assessed under the same conditions as the first strategy.
The third strategy is the Correlation-Adjusted RSI.
Creating the correlation-adjusted RSI involves several steps:
- Calculate the 14-period RSI on the market price.
- Determine a 3-period rolling correlation between price and RSI.
- Compute a 14-period RSI based on this correlation.
Correlation quantifies the linear relationship between variables, ranging from -1 to 1. Despite its limitations, this measure offers valuable statistical insights.
The correlation-adjusted RSI combines the standard 14-period RSI with a 14-period RSI calculated from price correlations, providing both signals and confirmation. The trading conditions are consistent with previous strategies, incorporating a filter based on correlation.
In the following section, we will analyze and compare performance metrics for all three strategies.
Chapter 3: Results and Performance Metrics
To recap the back-testing, our goal is to determine which of the three RSI strategies performed best on five major currency pairs since 2011 using hourly data. Findings indicated that the modified RSI (MRSI) yielded greater gross returns than both the standard and correlation-adjusted RSIs. The results were mixed between the regular RSI and CARSI, as outlined in the subsequent comparison tables.
Notably, the variable holding period consistently outperformed the fixed holding period. Throughout various years and currency pairs, the standard RSI's hit ratio did not surpass those of the MRSI or CARSI. Interestingly, in four out of five instances, CARSI exhibited a higher hit ratio than the regular RSI, suggesting the added value of the correlation filter. The modified RSI often demonstrated superior hit ratios compared to the other two methods.
For those interested in developing their algorithms, check out Lumiwealth, offering hands-on courses in algorithmic trading, blockchain, and machine learning.
In the variable period strategy, the modified RSI achieved the highest hit ratios overall, albeit with a lower risk-reward ratio. The correlation filter further enhanced performance, as CARSI's hit ratio surpassed that of the standard RSI, although the regular RSI's risk-reward ratio slightly exceeded that of CARSI.
In conclusion, our analysis of three distinct relative strength index strategies reveals that the modified RSI, utilizing a simple moving average, performed best over the last 11 years in the intraday context. The results suggest that traditional indicators like the 14-period RSI may not be adequately equipped to adapt to evolving market dynamics. Further research is necessary to substantiate the modified RSI's superiority over its standard counterpart, but current findings indicate it holds greater promise.
The first video titled "How To Use RSI For Multiple Time Frames" provides insights into utilizing the RSI across various timeframes, offering valuable strategies for traders.
The second video, "RSI Part 2 More Uses for RSI (Relative Strength Index) | Getting Started with Technical Analysis," explores additional applications of the RSI, enhancing understanding for those new to technical analysis.