AI TRADING MODELS
AlphaCube: On-premises AI/ML automated investing solution
Algorithmic trading benefits and problems
Financial markets have always attracted the strongest minds of this world because they have high liquidity, sometimes seem predictable, demonstrate high volatility, and the potential to earn on them. The price knows everything, each of its movements are due to certain actions of the market maker, institutional participants, or a large number of companies and individual traders. An imbalance of supply and demand generates price movements on which companies and traders are trying to generate profit.
All trading consists of two simple trading operations: BUY and SELL. After the BUY operation is completed, the trader is in a LONG position, earns on rising prices, then closes the position with the opposite SELL order.
The opposite is another type of position - SHORT. In this case, the broker or the exchange gives a temporary trading asset as a loan, the trader sells it, after which he is in a SHORT position and receives profit from the price drop. Then the trader closes the position with a BUY order, returns the loan, and receives a price difference.
Algorithmic trading has many obvious advantages. Such as automatic risk control, strict decision-making system, testing on the historical data and analysis of transactions, the absence of a human factor. However, despite all these advantages, algorithmic trading has a number of difficulties that all experts face. These difficulties include:
The complexity of the technical process.
The lack of universal software that is able to implement various ideas of analysts.
Inaccuracy and slowness of technical analysis indicators.
The difficulty of building effective and profitable strategies.
Permanent market variability, that makes it extremely hard to keep up with, constantly changing trading strategies.
The difficulty of validating strategies and checking for overtraining/overfitting. In other words, it is difficult to make sure that the result on historical data is not a coincidence and the strategy will be profitable in the future.
Our team has managed to solve all the problems of algorithmic trading that we know. We have developed such a concept for creating trading strategies that very easily matches with the task of automation and application of AI/ML solutions.
First of all, our algorithms are based on the experience of trading and automation of various trading ideas. We have worked with different analysis methods: indicators, classical technical analysis, pattern analysis, VSA, order book analysis, psychological models, genetic algorithms, neural networks, and other AI/ML methods.
We are sure that an efficient strategy is similar to a living organism, which is mutated and adapted. The strategy should consist of various methods of analysis, simultaneously solving many problems of recognition and adaptation. The market has a number of features and rules: trends are formed, there are various scenarios of volatility changing, performing various reactions of participants to overbought/oversold assets, manipulations by a market maker, and much more. Understanding all these and analyzing is the only way to make an efficient and profitable strategy.
Our specialists have extensive experience in algorithmic trading, as well as experience in the professional use of mathematics. We know the weaknesses of the “dry” mathematical approach, we are familiar with the problem of “lagging” indicators of technical analysis, the difficulty of recognizing visual patterns, and we know how to solve such problems.
Vision and goals
The project’s goal is the automatic generation of a diversified portfolio of investment strategies and their execution on world stock, FX and cryptocurrency markets. We are developing a fully automated solution that continuously learns and adapts to the market by leveraging data-driven technologies to find new strategies for profitable trades. It is the only possible way to automatically generate significant profit, in a market where multiple machine learning algorithms and AI operate.
Strategies automatically execute LONG and SHORT trades across multiple markets at scale, by leveraging machine learning, algorithmic trading, and AI.
In addition to this goal, we use our already developed models and experience in programming, analysis, and machine learning to create the most effective and safe portfolio of automated strategies.
Any world market
We currently have integration with three key global markets:
USA stocks exchanges.
It is possible to connect with any other world market, exchange, and brokerage.
Different trading strategies
Diversification with different trading strategies can guarantee high reliability of strategies portfolio.
Ability to adapt to the market and find new effective strategies without human interaction.
Best profit/risk ratio
Ability to choose any specifications of trading strategies and to choose the most profitable and most secure strategies.
Using advanced market analysis methods and performing the calculation and caching of necessary market data. Running machine learning algorithms that adapt to the market and find a way to earn by executing trades. Performing simulation and validation to ensure the reliability of detected trading strategies.
The project provides the possibility to generate a portfolio of strategies for each time interval and further simulate trading for next time interval. Such simulation with profitable trading results guarantees that generated data-driven strategies portfolio will also perform profitable trading in the future.
Why AlphaCube strategies is better in comparison with developed in-house?
We use about 200 different models as the basis for creating investment strategies. Different combinations of these models allow you to get different types of strategies: mean reversion, trend following, pattern trading, scalping, and many others. These strategies will be created for different world markets/exchanges, for different trading instruments, with a different approach to maximize diversification and reliability. Our technology will create from one to one hundred billion strategies for an individual client and select the best strategies among them.
The track record on live account
In order to validate our approach, simplified strategies were developed in a manual mode according to our methodology. The portfolio of series of trading strategies was launched in live trading on the cryptocurrency Binance exchange.
Portfolio of strategies generated 16.93% of profit during the 3 months test period. With average profitability of 5.64% per month and estimated yearly profitability of 93% (using compound interest).
With our data-driven algorithm, it is possible to find different types of investment strategies. We know how to train and adapt our system for various types of strategies that we actively implemented earlier with classical quantitative analysis.
50-100% annually, by using intraday and scalping strategies. Based on the trading stocks market and Forex.
20-50% annually, by using long-term strategies. Based on the trading stocks market and Forex.
10-20% monthly, by using aggressive scalping strategies. Based on trading cryptocurrencies.
These values are based on estimations and may differ significantly in practice. Trading results may be different due to the volatility of markets and may also depend on unpredictable fundamental factors. Any investment and trading in world markets may lead to losses.