A most important goal for CyberTrade is to maximise the trading Profit and Loss (PnL) , while managing risk, to produce Alpha.
The CyberTrade roadmap includes work on utilising Deep Reinforcement Learning, using neural networks
Alpha is often considered to be the 'active return' on an investment as compared to the HODL return.
The Artificial Nural Network (ANN) is used to convert the current 'state' (based on data and trading indicators) to get a result (Buy, Sell, Hold)
The Alpha component is monitored until it falls, at which time, a new trading strategy is considered in an attempt to create new alpha.
Using Gamification, each trade can be considered to be a 'game' in which a maximum reward is sought, as defined by a reward function.
Machine Learning (aka Predictive analytics) is a powerful component of of Artificial intelligence.
ML Algorithms enable computers to perform aspects of trading and technical analysis, including:-
1. Managing the investment portfolio
2. Analysing data and indicators to perform automated trading
3. Recognising trading chart patterns (Wedges & triangles, flags & pennants, head & shoulders, cup & handle, tops & bottoms)
4. Recognising Candlestick patterns (Doji, marubozu, hammer/hanging Man, Inverted hammer/shooting stars, three white soldiers/three black crows, morning star/evening star).
ML is well suited to trading because it is good at processing large amounts of data in a relatively short time.
The initial ML steps are; Create hypothesis, test, set goal, decide the ML algorithms and split data into training, testing and validation test data-sets.
Once the set-up is complete, the machine learning model is tuned (Hyper parameters) using the training data-set then trained on a walk-forward basis to test its predictions and goal-seeking ability.
A successful ML algorithm may be integrated into financial systems to manage position, risk etc.
CyberTrade sees the creating of new ML processes as a continuous activity,
Game planning can use machine learning (ML) to create the trading strategy.
A key element of this is knowing how to select the best set of indicators to use in particular market conditions.
ML provides a very powerful technique, known as 'Ensemble learning', by which it becomes possible to combine multiple, uncorrelated indicators to create a single, more robust trading signal.
Overall, several ML techniques can be used to help produce the game plan.
ML can be used to quickly recognise sets of indicators that give particularly strong signals in particular situations. This avoids the time-consuming, manual trial-and-error approach to finding good indicator combinations.
For example, Game Planning can use multiple rules, in combination, to create a Trading strategy, based on, say, 3 indicators and 4 trading rules.
AI in general and ML in particular is used to automate trading and to maximise rewards according to the reward function.
The game planning module uses Market analysis, indicator groups and rules to come up with the game plan.
The Trading strategy executes the game plan using AI to out-perform the market and produce alpha.