Annotated Models
This site is best viewed in a browser that conforms to web standards.
The strategy model that NeuroStrategy learned for AAPL looks promising. The strategy showed weakness in the years 1996 and 1997. The losses during those phases, however, were not out of the ordinary and were easily absorbed. The training target of an investment development that is as linear as possible was achieved. The equity curve moves in close proximity around the line of regression.
Although the enormous gain during training would seem to suggest a case of overfitting, the curve in the period of production shows that the strategy really does hold this potential. The equity curve remains constant in its development and even lies above the regression line, that is, above the average gain in training.
What inspires confidence is the fact that NeuroStrategy anticipates the dramatic crash in 2000 and still yields an extraordinary return in spite of the very difficult situation during that year.
NeuroStrategy is already able to realize profits in the sideways movement at the beginning of the year. In the weeks prior to the crash, the strategy shows uncertainty. The crash of the stock is anticipated, however, and through the use of the short position, NeuroStrategy is able to realize enormous profits in the following days. During the subsequent phase of uncertainty, NeuroStrategy maintains course without wavering and achieves outstanding performance.
A look at the period of production confirms the confidence in the strategy model. It is ready for productive use.
A look at the signal chart reveals that the short position prior to the crash in week 39 was no coincidence. Already in week 38, the situation is judged to be critical. Accordingly, NeuroStrategy with foresight clearly signals short positions and profits from the enormous slump of the stock.
The model tends towards full investment. Such behavior is typical of models with an extremely high training performance that is confirmed in production. It means that NeuroStrategy is very sure of its recommendations and assesses the risk as low.
Because of these features, the model is suited both for the active trader as well as for users who normally operate without asset allocation. Since the model prefers full positions anyway, it is safe to react merely to the signal change from short to long and vice versa. This reduces the number of repositionings and hence the loss due to transaction fees. Even with an assumed broker's commission of 0.18%, NeuroStrategy generated a return of 63% within 300 days of production.