The hypothesis:
Our inquiry into industrial research is poised to unravel intricate dimensions, with a focus on the following hypotheses:
a) Interrogating AI’s Frontier: This hypothesis delves into the intricate realm of artificial intelligence’s application, particularly deep neural networks, within the sphere of algorithmic trading. The dynamic interplay between conventional and cryptocurrency markets is scrutinized, aiming to ascertain the viability of these intricate methodologies. We want to answer a very simple but extremely challenging question: it is possible to apply the artificial intelligence methods our team has selected to predict the prices of at least some financial assets and obtain better results than popular benchmarks?
b) Architectural Insights: Our exploration extends to unveil the optimal neural network architectures for precise transaction prognosis and execution. The crux of our investigation lies in the meticulous assessment of these complex structures, pinpointing the configurations that harmonize with the nuanced landscape of trading.
c) Quantifying Strategic Efficiency: Amid our scholarly journey, we dissect investment strategies with a keen eye on their financial efficacy while concurrently navigating the intricate landscape of risk mitigation. The quest is to discern strategies that tread the delicate balance between financial gain and the mitigation of associated perils.
d) Navigating Multi-Criteria Complexity: Our scholarly expedition ventures into the uncharted domain of multi-criteria decision optimization. We explore the potential of a singular system orchestrating investment decisions across myriad markets, diversified assets, and heterogeneous transaction modalities, all while harmonizing complex matrices of profit and risk criteria.
Target assumptions:
- use of the XXX methods (name not revealed due to confidentiality reasons)
- win ratio within the range of: 60-70%
- max drawdown up to 25%.
Real results: SPY (S&P500)
- Annual Return: 4.791%
- Drawdown” 2.900%
- Sharpe Ratio: 1.251
- Loss Rate: 1%
- Win Rate: 99%
- Profit-Loss Ratio: 35.32
The main result:
- The research hypothesis has been positively verified. It is possible to use artificial intelligence methods that we selected to receive results outperforming the market.
- Some detailed results:
- In a few cases, the predictive power of our models was below acceptancy levels.
- In some other cases, we decided to accept lower win/loss rations, because of substantially lower than expected drawdown rates and higher win-rate ratio.
Further research:
- The predictive methods used are justified at this stage. The legitimacy of further implementation of the project was confirmed. The research should continue.
- The models should be refined in terms of their optimization, which should increase the profitability of transactions.