Alpha Predator™: Parsing Performance
When the Alpha Predator™ was last observed in APM II, we studied how it utilizes machine learning to optimize it’s algorithms to fit market conditions — finding there were approximately 10 trillion parameter combinations necessary to optimize the underlying systematic models, necessitating 54 billion years to run through all these options manually. The Alpha Predator may be a cosmological entity, but being how this is longer than the age of the Universe and impossible to do manually in a single human lifetime, Bayesian optimization methods are used to optimize these various parameters within a matter of hours.
With a few months of live-testing Alpha Predator performance under our belts, it is an opportune time to discuss this performance, along with our backtested performance going back two years. Since 2017 and 2018 were vastly different years in the digital asset world, it’s interesting to see what we can learn about the Alpha Predator’s behavior during these two seemingly disparate periods.
Performance Recap
If cryptocurrency price information took a single shape during 2017 and 2018 it would be a beautiful parabola arching to the sky, with 2017 shooting all the way toward the moon, only to fall back down to Earth in 2018. Both bitcoin and ethereum are down 72% and 80%, respectively, from their parabolic peaks. According to Hedge Fund Research, the HFR Cryptocurrency Hedge Fund Index was down 69.8%, a sign that most funds within this space do not seem to be appropriately focusing on risk management hedging.
As such, we see tremendous opportunity in the digital asset world to employ a more traditional risk management approach to this voraciously volatile space. In our quest to continually strive for excellence, we recently reached a higher echelon of analysis, combining our traditional methods of equities risk management with Artificial Intelligence and machine learning to develop modern quantitative strategies designed to hedge portfolio risk in this new digital asset world. Ideally, during parabolic moves like we saw in 2017, our strategies seek to capture 80% of the market upside and 40% of the market downside. While most funds lost a significant amount of capital in 2018, the backtests of our strategies did an excellent job of managing downside risk, and we are happy to report that backtests of our three strategies were all up for 2018.
Strategy Overview
Behind the cosmological curtain of Alpha Predator, our quantitative team has developed a number of complementary strategies — all built to manage risk within the digital asset world. We currently have three quantitative models in production, with a number of others in various stages of development. The models currently in production are:
• Systematic — Bitcoin
• Systematic — Ethereum
• Token Rotation
Our strategies are designed from inception with the goal to take advantage of the significant alpha opportunities we see in digital asset markets. We believe the current market environment is ripe for alpha harvesting due to nascent infrastructure, fractured exchange marketplaces, and the ubiquity of illiquidity.
Systematic Overview
The AlphaPredator™ Systematic quantitative models (first studied in APM I) measure market direction, velocity, volatility, and price spreads to categorize and tailor algorithms to the current market environment. Details about how our models work at the code level can be found in our Tech Corner. The algorithms seek to add exposure when the market confirms an upward trend. Similarly, when the trend plateaus or reverses, the model is designed to trim exposure. The models focus on entering positions when it calculates the probability of profit is above average and removing exposure when the algorithms deem the probability of profit to be waning or non-existent. The general goal of our systematic strategies are to provide between 70–80% of the market upside, 40–50% of the downside, and approximately half the underlying asset’s volatility. The models seek to maximize the risk/reward ratio by measuring Sharpe, Sortino, and downside risk metrics — all in real time.
Systematic Bitcoin Performance
Our Systematic-Bitcoin performance metrics shown below represent live-test results from October 15th, 2018 through December 31st, 2018, and backtest results from January 1st, 2017 through October 15th, 2018, where applicable.
Systematic Ethereum Performance
Our Systematic-ETH performance metrics represent live-test results from December 13, 2018 through December 31st, 2018, and backtest results from January 1st, 2017 through December 13th, 2018, where applicable.
Token Rotation Overview and Performance
The AlphaPredator Token Rotation strategy aims to add exposure when the market is in an upward trending environment. The model analyzes the top 15 digital assets by market capitalization and invests in the four best-performing assets, as measured by the velocity and performance of those assets against each other. However, if certain digital assets appear to be reversing their trend negatively, the model seeks to protect downside risk by rotating out of those assets into cash. At any given time the model could be invested 0–100%.
While the systematic strategies focus on limiting risk based on the underlying asset, the token rotation strategy is a risk-seeker when the model calculates there to be a high probability of an upward trending market. The long term goal of the Token Rotation model is to target 120% of the upside of the HFR cryptocurrency index, 60% of the downside, and approximately 60% of the volatility.
Quick Note about Token Rotation: Given the paltry volumes in altcoins when compared to the large-cap cryptocurrencies like Bitcoin, Ethereum, and XRP, there are severe limitations in the amount of capital that can be allocated to the strategy.
Diversification to Lower Correlation
As shown in the backtest statistics above, our strategies kept upside pace in 2017 and protected downside risk in 2018. Each strategy is currently run individually to track performance, but these strategies are complementary to each other and therefore when run simultaneously they should balance risk and lower correlations to the overall markets. Utilizing modern portfolio theory we can tailor an optimal combination of risk reducer strategies (market neutral) and return enhancer strategies (Systematic and Token Rotation). The resulting optimally blended strategy should maximize the Sharpe and Sortino ratios, limit downside risk, and contain a return stream that is less volatile than would be achieved through traditional market strategies.
What’s Next?
Our quantitative team is constantly working on optimizing our algorithms — incorporating new signals and developing infrastructure — all with the goal of creating scalable solutions that appropriately manage risk.
Below is a small sample of what our team is currently working on:
1. Systematic Strategy — Upgrading from v1 to v1.5: Current model (v1.0) allocates 100% long or 100% cash. The upcoming upgrade will incorporate numerous signals, legging in and out of positions, with an emphasis on seeking to enhance upside capture and limit downside risk.
2. Token Rotation Strategy — Creating Scale: The current Token Rotation model is limited in capital due to the rotational aspect of allocations to smaller market capitalization digital assets. We are scaling the capital management by incorporating a multi-factor model that allocates based on long and short term signals, with the goal of slowing down the portfolio turnover and increasing the amount of capital that can be allocated to the strategy.
3. Market Neutral Strategy — Profitability and Scalability: While we haven’t talked much about our market neutral strategies in this update or published numbers for this strategy yet, our team is working on building and enhancing the performance of the system versus total return performance. We have a number of strategies in various stages of development, including exchange arbitrage, statistical arbitrage, stable coin arbitrage, and market making. Over the next few months, we will roll out these strategies after they have passed through our rigorous testing environment.
4. Shorting: Our current models focus on taking long positions or entering into cash. That was intentional, as we made it a priority to focus on lowering volatility and adding exposure via long positions. In most scenarios, we view shorting as a position, not a hedge, which would add volatility to the portfolio, contradicting our goal. Once we finalize our long-biased positions, we will look to add algorithms that aim to capture performance via downside moves through short positions.
Conclusion
As the sun sets on yet another Alpha Predator paper, the ongoing self-optimization continues. With machine learning algorithms working diligently between the bones of the predator, our strategies get that much tighter and more efficient — good for getting into those crevices of alpha inaccessible to most. We’ll continue optimizing, and hope you continue reading.
This publication was developed by Blockforce Labs, LLC. Blockforce Labs, LLC, is a research company and is not a registered investment advisor and is not acting as a broker-dealer under any federal or state securities laws. This publication is for illustrative purposes only and contains information regarding the theoretical (back-tested and live-tested) performance of trading strategies. The theoretical results presented are not actual results, and actual results may differ materially than those presented herein. Past performance does not guarantee future results. This publication is not an offer to sell or the solicitation of an offer to purchase any interest in any investment. It is not possible to directly invest in a strategy. The strategies discussed above are unmanaged and do not incur management fees, transaction costs, taxes or other expenses associated with investable products. The theoretical results presented herein are presented “gross of fees,” that is without any of the potential fees deducted that would be charged by a manager or adviser utilizing the trading strategy. Theoretical results do not reflect the separate advisory fee that may be charged by an investment adviser (for example, an advisory fee of 1% compounded over a 10-year period would reduce a 10% annual return to an 8.9% return). The results reflect performance of a strategy not offered to investors and do NOT represent returns that any investor actually attained.
Theoretical performance results may have inherent limitations, and actual returns from live investment may differ materially from the hypothetical returns presented. While this information has been prepared in good faith, there are inherent limitations that recipients must consider carefully. Theoretical performance results do not present the impact that material economic and market factors might have on a trading strategy or an investment adviser’s decision-making process if anadviser were actually managing investor funds. Different market and/or economic conditions, different allocation sizes, different feesor different assumptions may yield different results. Back-tested performance is developed with the benefit of hindsight and has inherent limitations. Back-test results are calculated by the retroactive application of a model constructed on the basis of historical data and based on assumptions integral to the model which may or may not be testable and are subject to losses. General assumptions include: the assets recommended by the model would have been able to be purchased and the markets were sufficiently liquid to permit all trading. Changes in these assumptions may have a material impact on the back-tested returns presented. Certain assumptions have been made for modeling purposes and are unlikely to be realized. No representations and warranties are made as to the reasonableness of the assumptions. This information is provided for illustrative purposes only.
This publication does not constitute a recommendation of any investment strategy or product for a particular investor, or take into account the unique financial circumstance of any individual. Investors should always consult a financial professional before making any investment decisions. All investments carry a certain degree of risk, including complete loss of principal.