Strategy Review of the HeroRATs-Strategy by Chrilly Donninger
Using Volatility to Stay One Step Ahead of the Market
After the very positive feedback on my review of the Value & Opportunity BOSS-Score, I’m back with another strategy deep dive. This post is a bit longer, so please make sure it’s displayed correctly in your email client to get the full picture.
Most investors traditionally view volatility as an unavoidable cost of investing or a risk factor to be minimized. However, volatility can also be interpreted as an informative market signal, one that helps identify when risk-taking is rewarded and when capital preservation should take priority. This premise forms the foundation of the paper “How to Beat the Market with the Implied Volatility Term Structure: The HeroRATs Strategy” by Christian “Chrilly” Donninger.
Rather than relying on backward-looking indicators such as earnings growth, macroeconomic forecasts or complex econometric models, Donninger’s approach is built on the structure of implied volatility embedded in option prices. By analyzing the relationship between short-term and long-term implied volatility, the strategy seeks to identify changes in market regimes before they are fully reflected in equity prices.
Historically, this framework has delivered strong risk-adjusted performance, including materially lower drawdowns compared to a buy-and-hold investment in the S&P 500. Yet despite these results, neither the HeroRATs strategy nor the broader use of implied volatility term structure as a tactical allocation tool has achieved widespread adoption (to my knowledge).
In this post, I revisit the HeroRATs framework, evaluate how it has held up in recent years, and translate its core ideas into plain language by showing how volatility can guide timely shifts between risk-seeking and defensive positioning.
The Core Problem: Timing Risk Without Overengineering
At the heart of tactical asset allocation lies a persistent challenge: how to adjust portfolio risk dynamically without relying on overly complex or fragile models. Most forward-looking allocation frameworks rely on backward-looking momentum signals, such as reallocating assets based on recent performance.
This present meaningful limitations. Momentum-based strategies are straightforward and robust, yet inherently reactive. They tend to adjust exposure only after market regimes have already shifted, leaving portfolios vulnerable during sudden drawdowns or sharp reversals.
The HeroRATs strategy wants to fix this limitations by being:
Forward-looking (implied volatility)
Simple (rule-based switching)
Low turnover (minimal transaction costs).
Strategy Overview
The HeroRATs strategy is built around a 3-phase allocation framework that dynamically shifts capital between them. Specifically, the strategy allocates between risky and defensive assets.
The allocation decision is governed by the Implied Volatility Term Structure (IVTS), defined as the ratio of short-term implied volatility to long-term implied volatility.
IVTS = short-term implied volatility / long-term implied volatility
This ratio provides insight into the market’s perception of near-term versus longer-term risk.
When short-term implied volatility rises sharply relative to long-term volatility, the market is signaling acute, near-term uncertainty, conditions that historically precede equity sell-offs. Conversely, when the volatility curve is flat or downward sloping, risk perceptions are benign, and risky assets tend to deliver superior returns. Importantly, IVTS is not designed to forecast returns directly, but to identify shifts in market regimes.
To operationalize this insight, the HeroRATs framework classifies market conditions into three regimes. In low-IVTS environments, capital is fully allocated to VBK to capture small-cap growth premiums. During intermediate IVTS readings, the strategy adopts a balanced allocation between VBK and TLT to dampen volatility while retaining upside participation. In high-IVTS regimes, the portfolio moves entirely into TLT, prioritizing capital preservation. Although the signal is evaluated daily using closing prices, actual trading activity remains limited because regime transitions occur infrequently.
The following bullet points summarize the three regimes perfectly:
Calm Markets (Low IVTS)
Allocation: 100% VBK (Vanguard Small Cap Growth ETF)
Goal: Capture small-cap growth outperformance
Transitional Markets (Mid IVTS)
Allocation: 50% VBK / 50% TLT
Goal: Reduce drawdown while maintaining upside exposure
Turbulent Markets (High IVTS)
Allocation: 100% TLT
Goal: Capital preservation during crashes
The paper evaluates multiple IVTS configurations using CBOE volatility indices and VIX futures across different maturities, ranging from fast-reacting short-term measures to slower, more stable long-term combinations. The key trade-off is clear: faster IVTS signals provide earlier warnings of stress but are inherently more susceptible to noise.
One of the most important insights in the paper is not the volatility signal itself, but how that signal is cleaned up. Borrowing a concept from image and signal processing, the strategy applies a median filter to the IVTS in order to reduce noise and avoid costly overtrading.
Unfiltered IVTS data frequently exhibits sharp, short-lived distortions, including isolated one-day volatility spikes and brief episodes of market panic that quickly reverse. When acted upon directly, these distortions can trigger unnecessary regime changes, leading to whipsaw trades and avoidable transaction costs.
To address this issue, the strategy replaces single-day IVTS observations with a median of the previous five trading days. This simple modification has a profound impact. This approach filters out temporary panic spikes, significantly reduces false regime changes, and leads to smoother portfolio behavior. While the filter introduces a slight delay in recognizing real shifts in market conditions, the historical evidence shows that the benefits (lower drawdowns and more stable performance) clearly outweigh the cost of slower reaction.
My Analysis
I used SPY and QQQ as the risky assets for the analysis and TLT as the defensive one.
The strategy divides market conditions into three clear regimes. When IVTS is low, markets are considered calm, and the portfolio is fully invested in SPY or QQQ. When IVTS moves into a middle range, signaling rising uncertainty, the portfolio splits exposure evenly between equities and TLT. When IVTS is high, indicating elevated market stress, the portfolio shifts entirely into TLT.
The following bullet points summarize the three regimes:
Calm Markets (Low IVTS)
Allocation: 100% SPY or 100% QQQ
Transitional Markets (Mid IVTS)
Allocation: 50% SPY or QQQ / 50% TLT
Turbulent Markets (High IVTS)
Allocation: 100% TLT
Although Donninger tested six different versions of the implied volatility term structure, I simplify the approach by using VIX divided by VIX3M. This version of IVTS is easy to compute, widely available, but it reacts more slowly than Donninger’s original formulations.
To make the strategy actionable, IVTS values must be mapped to the three regimes using thresholds. A lower threshold separates calm from transitional markets, while an upper threshold identifies turbulent conditions. Donninger does not specify how his thresholds were chosen, but practical reasoning helps fill the gap. Since markets are calm most of the time, the strategy should spend the majority of its time in risk-on mode. Risk-off positioning should be infrequent and reserved for periods of genuine stress.
Because volatility measures tend to be positively skewed, IVTS is expected to follow a log-normal distribution. A simple and intuitive way to define thresholds is therefore to use percentiles of the IVTS distribution. The figure below illustrates how this percentile-based approach can be used to classify regimes.
As mentioned earlier, the risk-on regime naturally occupies the largest portion of the IVTS distribution, reflecting the fact that financial markets spend most of their time in relatively calm conditions. In contrast, the upper threshold is intentionally set to capture only the extreme tail of the distribution, corresponding to periods of elevated stress or outright crisis. This ensures that the strategy remains invested in risk assets during normal environments and shifts into defensive positioning only when volatility signals are truly exceptional.
The following figure illustrates the empirical distribution of IVTS without applying any filters.
As discussed earlier, the IVTS distribution closely resembles a lognormal shape, meaning that most observations are clustered at relatively low levels, with a long tail representing periods of elevated market stress. Based on this distribution, I determined the regime thresholds using a simple, intuitive approach rather than a fully optimized one. By visually inspecting the data, I set the lower threshold at the 88th percentile and the upper threshold at the 98th percentile.
While there is certainly room for further refinement and parameter optimization, the goal at this stage is not to fine-tune the model for maximum historical performance. Instead, following the 80/20 principle, these threshold choices are intended to capture the core behavior of the strategy with minimal complexity. If the approach is fundamentally sound, it should demonstrate its effectiveness even with rough, common-sense parameter settings.
Donninger emphasizes that signal filtering plays a critical role in making the strategy robust, particularly in reducing noise and avoiding unnecessary regime switches caused by short-lived volatility spikes. To account for this, the analysis includes multiple variations of the strategy. Specifically, I examine a version without any filtering, a version using a 5-day median filter, as proposed in the original paper. I additionally test a 10-day median filter for QQQ. This comparison helps illustrate how different levels of smoothing affect signal stability, responsiveness, and overall portfolio behavior.
The Results
I evaluated several variations of the HeroRATs strategy over the period from 2007 through 2024, a timeframe that includes multiple market regimes like, major crises and extended bull markets. Testing the strategy across such a diverse period helps assess not only its return potential, but also its behavior under very different volatility regimes. The outcomes of these tests are summarized in the following table.
Starting with returns, the unfiltered SPY-based strategy delivers a CAGR of 9.17%, which already comes close to buy-and-hold SPY at 10.30%, despite actively de-risking during turbulent periods. Once a 5-day median filter is applied, performance improves materially, with CAGR rising to 11.75%. This confirms Donninger’s original insight: removing short-term noise significantly enhances decision quality.
The effect becomes even more pronounced when using QQQ as the risk asset. Without filtering, the CAGR jumps to 14.31%, reflecting QQQ’s higher growth profile. Adding a 5-day median filter increases returns further to 16.27%, and extending the filter to 10 days pushes CAGR to 17.20%. The monotonic improvement suggests that smoother signals allow the strategy to stay invested longer during strong trends while still stepping aside during genuine stress events.
It is important to note that all reported results are shown before accounting for transaction costs. This distinction matters because the different strategy variants generate very different levels of trading activity. The unfiltered versions exhibit a high number of regime changes, as they react to short-lived volatility spikes and noise. In practice, this would translate into frequent portfolio rebalancing and, consequently, significantly higher transaction costs, which would further erode their realized returns.
By contrast, the filtered strategies and especially the 10-day median filter variant produce far fewer regime transitions. As a result, the performance gap between filtered and unfiltered approaches would likely widen even further once realistic transaction costs are taken into account.
What makes these results particularly compelling is that higher returns are not achieved through higher volatility. For SPY-based strategies, volatility remains almost unchanged when moving from no filter (15.11%) to the 5-day filter (15.12%). Yet returns improve substantially, indicating a clear gain in efficiency.
For QQQ-based strategies, volatility is naturally higher due to the underlying asset, but it remains remarkably stable across filtering choices, hovering around 18%. In contrast, buy-and-hold SPY exhibits the highest volatility in the period under review at 19.42%, despite delivering lower returns than most HeroRATs variants.
The Sharpe ratio provides the most concise summary of these improvements. Buy-and-hold SPY posts a Sharpe of 0.53, reflecting the well-known challenge of equity-only exposure across full market cycles.
By comparison:
SPY HeroRATs without filtering already improves Sharpe to 0.61
Adding a 5-day filter lifts it to 0.78
QQQ-based strategies range from 0.79 (no filter) to 0.94 (10-day filter)
The QQQ M10 filter variant stands out as the strongest overall performer, combining the highest return with a materially improved Sharpe ratio. This indicates that the additional smoothing is particularly beneficial for faster-moving, higher-volatility assets.
Key Takeaways
Median filtering is not optional — it is essential.
Performance improves consistently as noise is reduced.
Higher growth assets benefit more from filtering.
QQQ paired with longer median filters shows the strongest risk-adjusted results.
Risk-adjusted returns dominate buy-and-hold.
All filtered HeroRATs variants outperform SPY buy-and-hold on a risk-adjusted basis.
The results, particularly those of the QQQ strategy using the 10-day median filter, are striking. They demonstrate that the core design principles behind Donninger’s HeroRATs framework remain highly effective, even when applied to different assets and a different IVTS calculation method. Despite changes in market dynamics over time, the strategy’s underlying logic, using the implied volatility term structure as a regime signal and filtering out short-term noise, continues to deliver strong, risk-adjusted performance.
Further work
This analysis can be meaningfully extended in several ways. One obvious direction is to apply the framework to different risk assets, such as international equities or like in the original with small cap growth stocks. In addition, the current implementation relies on a relatively slow IVTS measure; experimenting with faster-reacting volatility ratios could improve the timeliness of regime detection, albeit with the need for careful filtering to manage noise. Exploring these variations would help further assess the robustness and adaptability of the strategy across different market environments.
Conclusion
By shifting the focus away from backward-looking return data and toward the forward-looking information embedded in the implied volatility term structure, the strategy reframes volatility from a mere nuisance into a practical decision-making tool. Rather than treating volatility as something to endure, it becomes a signal that helps distinguish environments in which risk-taking is likely to be rewarded from those in which caution is warranted.
What stands out most clearly from this review is not the optimization of any single parameter but the robustness of the underlying concept. Across different implementations (e.g. risk assets, filtering choices) the central insight remains consistent: markets tend to signal regime changes through the shape of the volatility curve well before those changes are fully reflected in equity prices. When this signal is properly filtered to remove short-term noise, the result is meaningfully improved risk-adjusted performance.
Equally important is what the HeroRATs framework does not rely on. It avoids macroeconomic forecasts, earnings models and complex optimization routines. Instead, it is built on a small set of transparent, rules-based decisions. A key lesson that emerges is that signal quality matters more than signal speed. The substantial improvement achieved through median filtering highlights that avoiding false alarms and unnecessary trades is often more valuable than reacting immediately to every market fluctuation. In real-world portfolios, fewer (higher-quality) decisions tend to outperform frequent (reactive) ones.
From a performance perspective, the strategy succeeds in raising long-term returns without a corresponding increase in volatility, leading to materially higher Sharpe ratios compared to a buy-and-hold equity strategy. This is precisely what a well-designed tactical overlay should aim to deliver: not perfect market timing, but a smoother investment journey across full market cycles and reduced exposure to severe drawdowns.
At a broader level, the continued effectiveness of the HeroRATs approach suggests that volatility-based regime detection remains underutilized by most investors. While momentum and trend-following strategies dominate the tactical allocation space, implied volatility offers a complementary and forward-looking perspective. For investors willing to look beyond traditional indicators, HeroRATs demonstrates that proactive risk management is possible without overengineering, excessive trading or reliance on fragile models.
Thank you for your interest in this post. If you liked it, I would really appreciate a like, as these are very motivating.
You are also welcome to share this post with others.
Disclaimer
The above article constitutes my or the authors’ personal views and is for entertainment purposes only. It is not to be construed as financial advice in any shape or form. Please do your own research and seek your own advice from a qualified financial advisor. I / The authors may from time to time hold positions in the aforementioned securities consistent with the views and opinions expressed in this article. The information provided in this article is not making promises, or guarantees regarding the accuracy of information supplied, nor that you guarantee for the completeness of the information here. The information in this article is opinion-based and that these opinions do not reflect the ideas, ideologies, or points of view of any organization the authors may be potentially affiliated with. The authors reserve the right to change the content of this blog or the above article. The performance represented is historical and that past performance is not a reliable indicator of future results and investors may not recover the full amount invested.