Introduction

‍

In the upcoming three articles, we'll explore a key challenge for portfolio managers: managing risk and enhancing performance while minimizing rebalancing actions. Many asset managers face constraints due to portfolio size or regulatory requirements, aiming to maintain stable portfolios that withstand market downturns and maximize long-term returns without frequent rebalancing.

‍

In this article, we'll show that using a basic Relative Strength Index (RSI) indicator for rebalancing, without frequency constraints, can provide more value than a fixed schedule. However, this method often leads to increased rebalancing actions due to the indicator's variability, requiring smoothing to reduce frequency and make it feasible in a portfolio context. Unfortunately, this smoothing can significantly reduce the RSI's effectiveness, impacting the long-term strategy. This presents a trade-off: highly reactive indicators may require frequent rebalancing, while smoother ones can support more sustainable schedules but often result in weaker performance.

‍

The second article will discuss a simple solution from Alquant to optimize rebalancing frequency in an equity-cash portfolio, and the third will introduce a low-frequency rebalancing approach for an equity-only portfolio.

‍

Rebalancing based on a risk indicator

‍

Most asset managers adhere to a fixed rebalancing schedule, a passive approach that neither actively manages risk nor adds significant value. A more dynamic alternative is ad-hoc rebalancing, triggered by specific conditions indicated by a risk management tool or indicator.

This section will demonstrate that using a simple risk indicator for ad-hoc rebalancing can improve long-term performance, albeit requiring more frequent adjustments. We test a very simple strategy based on the Relative Strength Index (RSI), adjusting equity exposure according to RSI values.

‍

1. Low RSI: Indicates an oversold market or major drawdown. It’s prudent to avoid market exposure.
2. High RSI: Suggests an overbought market, signaling a potential correction. Again, avoiding equity exposure is wise.
3. Mid-range RSI: Equity exposure is maintained.

‍‍

Our investment universe comprises US Equity (S&P 500 TR index) and USD cash. The strategy adjusts exposure based on RSI values. We will apply a stepwise strategy according to RSI values, as described below in Figure 1.

‍

Fig 1: Description of our RSI strategy and its equity exposure depending on RSI value

‍

We compare this strategy with a 50%-50% equity-cash portfolio rebalanced biannually, including 0.05% transaction costs.

Fig 2: Performance of a simple RSI strategy compared with a 50% equity - 50% Cash benchmark, from 2008-01-15 to 2024-05-20

‍

The RSI strategy outperforms the equally weighted portfolio during bull markets while maintaining comparable performance during major drawdowns like the 2008 financial crisis and the COVID-19 pandemic.

‍

Fig 3: Statistics of our simple RSI strategy compared with a 50% equity - 50% Cash benchmark, from 2008-01-15 to 2024-05-20

‍

The RSI strategy significantly boosts returns and improves risk-adjusted performance, achieving an annualized Alpha of 5.1% and an information ratio of 0.54. However, it requires more frequent rebalancing - averaging over once a month, impractical for institutional asset managers. To mitigate this, one could smooth the RSI signal to stabilize it and thus reduce the rebalancing frequency. We will explore this potential solution in the next section.

‍

Smoothing a signal to reduce rebalancing frequency

‍

To smooth a signal effectively while maintaining the ability to detect significant changes or regime shifts, several techniques are available. Among common approaches, we can find:

‍

1. ‍Moving Average: The moving average is a straightforward way to smooth out noise in a signal. You can choose a simple moving average (SMA) or an exponential moving average (EMA) which gives more weight to recent data points.‍
2. Low-Pass Filters: Low-pass filters are effective for smoothing signals while preserving sharp step changes. ‍
3. Butterworth Filter: This filter is ideal for data that has high-frequency noise, but you want to keep sudden changes, like shifts in regimes.‍
4. Median Filter: A median filter is particularly effective in removing “spike” noise and is better than averaging filters when dealing with outliers because it uses the median of the data points in the window instead of the mean.‍
5. Savitzky-Golay Filter: The Savitzky-Golay filter smooths the data by fitting successive subsets of adjacent data points with a low-degree polynomial. It tends to preserve features of the distribution such as relative maxima, minima, and width, which are usually flattened by other types of filters.‍
6. Wavelet Denoising: Wavelet denoising is suitable for non-stationary signals where frequency characteristics change over time. It allows you to decompose the signal into components at different frequencies and smooth out the components that are considered noise.

‍

The success of any smoothing technique heavily depends on precisely tuning parameters like window size or cutoff frequency to suit the specific characteristics of your data. However, tuning these parameters carries the risk of overfitting or data mining, which is a significant drawback of traditional smoothing methods. Moreover, it is essential to validate the chosen smoothing technique by visually inspecting the results to confirm that it maintains the critical features of the data. This validation step is crucial to ensure that, despite the smoothing, significant changes and trends within the data are still accurately represented.

Using a 20-day rolling average to smooth the RSI signal, we aim to match the benchmark's semi-annual rebalancing schedule.

‍

Fig 4: Performance of a smoothed RSI strategy compared with a 50% equity - 50% Cash benchmark, from 2008-01-15 to 2024-05-20, including 0.05% transaction costs.

‍

The smoothed RSI strategy reduces rebalancing frequency to about twice per year. However, it decreases overall performance and is less effective during major market downturns, as seen in 2008, 2020, and 2022, resulting in drawdowns that don't align as closely with the benchmarks.

‍

Fig 5: Statistics of a smoothed RSI strategy compared with a 50% equity - 50% Cash benchmark, from 2008-01-15 to 2024-05-20, including 0.05% transaction costs.

‍

Overall, the smoothed strategy now achieves a risk-adjusted performance comparable to a 50%-50% equity-cash portfolio, indicating a profound alteration in the value provided by our indicator.

In this article, we explored dynamic risk management strategies, transitioning from fixed to ad-hoc rebalancing based on a risk indicator. While signal smoothing reduces rebalancing frequencies, it can reduce the indicator's effectiveness. The upcoming articles will present Alquant's solutions to maintain low rebalancing frequency while enhancing risk-adjusted and absolute performance. These solutions employ the well-known principle of diversification, combining multiple indicators to reduce noise, instead of implementing a smoothing technique. As we will explore, this method not only helps mitigate noise but also retains most of the original indicators' added value.

‍

Disclaimer

‍

‍This content is advertising material. This content as well as all information displayed on any of Alquant’s websites does not constitute investment advice or recommendation, and shall not be construed as a solicitation or an offer for sale or purchase of any products, to effect any transactions or to conclude any legal act of any kind whatsoever. Past performance is not a guide to future performance.