FROM THE VAULT: Regime Aware Asset Allocation

Introducing FROM THE VAULT, an inside look at our hedge fund methodologies — written by our digital assets experts and analysts.

Markets go through different phases, or "regimes," each with its unique characteristics. By identifying these regimes, we can tailor our investment strategies to better align with current conditions. To do so in an objective and systematic way we can apply advanced unsupervised machine learning algorithms.

What is Regime Aware Asset Allocation?

Unsupervised Machine Learning approach, with the core idea of dividing its history into similar buckets. These algorithms cluster the history into homogenous groups through the lens of so called “features”.

Most commonly used features are market returns and volatility. By using these 2 features we can divide the past into 4 different regimes:

Bullish Calm: Positive returns, low volatility.

Bullish Volatile: Positive returns, high volatility.

Bearish Calm: Negative returns, low volatility.

Bearish Volatile: Negative returns, high volatility.

But why stop there? We can also incorporate alternative data like macroeconomic indicators to get a feel of the current macro regime or with digital assets we have the luxury of having access to on-chain metrics to uncover deeper insights.

Methodology

Here is how it works:

1. Regime Identification: Using algorithms like Hidden Markov Models (HMM), we classify historical data into distinct regimes.

2. Regime Forecasting (Optional): We can either use yesterdays regime as prediction of todays regime or build an extra model that tries to predict future regimes.

3. Portfolio Model Construction: Different regimes call for different strategies. In a calm, bullish regime, a straightforward long strategy might work best. In volatile times, we might switch to hedging or mean reversion tactics.

Practical Example

In this example we only use price data to categorize the market into 4 different regimes. To keep it simple, we use yesterdays regime to predict tomorrows regime and we simply go long during regime 0 and are flat otherwise.

This gets us the following performance metrics:

This simple RAAA strategy improves all of the above metrics compared to buy and hold. With only halve the downside it delivers higher total returns, providing overall higher risk adjusted returns.

Challenges and Limitations

No approach is without its hurdles. Regime-aware asset allocation faces challenges like:

Instable States: Resulting in high turnover.

Misclassification Risks: Incorrectly identifying a regime can lead to suboptimal decisions. Data Limitations: The accuracy of our models hinges on the quality and quantity of data available.

Many Pitfalls: Especially when wanting to deploy these models in live production.

Conclusion

Regime-aware asset allocation can be a powerful way to leverage machine learning to make sense of the market. However, they need to be treated carefully as there are many pitfalls when trying to use them live. Also there is no silver bullet it is just one other way to look at the market. The power of systematic approaches comes from combining many different models to counter single model failure.

At Hartmann Capital, we pride ourselves on integrating the latest financial advancements into our decision making. By ensembling many quantitative models with our human expertise we aim to provide our clients with a sophisticated, dynamic investment approach that is robusts and outperforms.

Join us on this exciting journey. At Hartmann Capital, we don’t just invest in the future; we also integrate the most powerful advancements into our decision-making process.