We may permit the system to select individual stocks on a pro-forma basis, using the neural networks which continue their learning process into the future. The results can be seen in the following chart:
The stock selection process utilizes neural networks that were created between 2004 and 2007 for 1000 individual tickers. The model was constrained to meet liquidity requirement of more than 500,000 shares daily trading volume. The networks were allowed to continue to learn from daily market data from 2007 to the present. The system will trade the same equity multiple times if the favorable pattern surfaces on more than one occasion.
The pro-forma annualized return for the CONCERT AI – Liquid is 27.8%, suggesting that allowing the model to get ‘smarter’ over time has tangible benefits.
The following chart compares CONCERT AI S&P, CONCERT AI Liquid, and the benchmark S&P 500 statistics: