I perfectly understand that my readers ask for backtesting and performance statistics of the model. Let me explain, it is natural as most of you are trained in technical analysis and are used to manage tools that have as their main feature to let you scan it retroactively. It’a one dimensional analysis. Here, I don’t do one dimensional analysis. The model manages hundreds of dimensions. It’s so different that just the idea of designing a backtest tool that may have some meaning is depressive. Looking at the model as you look at a tech analysis chart is meaningless.
The model sees. It is instructed to recognize and it does it. The model may go deeper inside the hidden structure, through the numbers and NOT human opinions. The model is indipendent. Well, it surely depends on its design, and I’m working to make it evolve, but it is indipendent as it just correlates the numbers. I told you, it’s different.
Since mid March, I’m collecting some real time statistics, because I’m curious as you are about the smartness of the model: the chart here is the same day close forecast (magenta line) compared to the actual market close (blue line). This is not backtesting, this is real time audit with the model full steam, for latest eight months.
Some more statistics, on Profits and Losses,
(not considering open position)
- First Position: 12th February 2016
- Starting Index Value: 1857
- Number of Positions closed since then: 7
- Total profit (index points): 438
- Number of Positions with Profit: 6
- Number of Positions with Loss: 1 (-0.28%)
- Global Performance (Total Profit/Starting Value): +23.59%
This has been, in real time, the behavior of the model trading system embedded mainly in the Pos indicator, team working with the other indicators. The execution is always validated at the opening price of the day following the trigger generation (with a small realistic slippage).