MM4 Model Design
The model is an autonomous Stock Portfolio Trading model.
It utilizes high probability trading patterns to enter the market and smart risk/reward parameters to exit the market.
Finally, the model only trades a portion of the entire portfolio. The model compares historical performance among the stocks and only trades the top-tier performers.
The model can operate independently for years. It is designed to minimize the decisions needed to make it profitable.
Part 1 - High Probability Entries
The entry system is based on high probability price patterns. In the beginning, MM4 began as a forecasting model designed to forecast the direction of individual stocks on a daily, weekly, and monthly time periods. A key goal was to be 66 to 70% accurate on average for forecasting all the stocks for all three time periods.
The goal was achieved, but then another goal was ignited. Can the model be made to trade stocks?
Relative to trading, forecasting is much easier. Trading is far more complex involving when to exit with profits or losses, and even what to trade is important. Part 2 of the building the model was tasked with answering these questions
Part 2 - Smart Risk Reward
In part 2 of the model development, the question of when to exit with a profit or loss was very difficult. To solve this problem, a page from the Generic Algorithm playbook was employed.
Genetic algorithms are used to find optimal solutions for problems with wide or complex parameter ranges. This sounded like our exit problem, so our model runs an exhaustive number of exit scenarios for every possible entry. Over 41,000 scenarios are run for each trading opportunity and two algorithms are used to determine how to adjust parameters for trading going forward.
The algorithms are balanced to be aggressive enough to achieve substantial profits while keeping risk and losses low.
In the backtesting from 2010 to 2016, the model adjusts itself quarterly without human interaction.
Part 3 - Stock Selection & Evaluation
Prior to this evaluation stage, the model calculates the best trading patterns and parameters for each stock that will yield a substantial return. However, not every stock can be traded at once, so the model has an algorithm that chooses the best of the best stocks for each trading cycle.
The stock evaluation model compares the performance of each stock and selects the top stocks to trade.
