Swedish

Screenshots

Technical Analysis with
Optimized Parameters

Working with optimized parameters provides the best conditions for high quality trading signals, but it is good to know the drawbacks and traps that exist. With Optimal Trader it is easy to conclude when an optimization for some reason has failed by analyzing the green and red color regions in the charts. In many cases it is possible to improve the settings and re-optimize the stock.

Do not use to short optimization periods.
The parameters run the risk of getting overfitted!

In the figure below a fund has been optimized with an optimization period of 100 days. The delay for a buy or sell order to be carried out is one day. Thus it takes one day for the Return Curve (blue) to become flat after a sell signal. In the same way it takes a day after a buy signal before the Return Curve again begins to follow the price movements.

In the figure below a sell signal is generated exactly one day before the price is beginning to fall. The latter buy signal is also generated exactly one day before the price is rising again. The fund will thus be sold exactly at the peak of the price and rebought exactly at the bottom of the price. This is a clear signal that the settings of the model (parameters) have been overadapted to these historical events. The solution in this case is to use a considerably longer optimization period.

Warning 1

 

The chart below shows that optimization has adjusted the parameters of the model to produce a sell signal exactly at the peak around 2004-12-01. This is a clear warning that curve fitting may have occurred.

Warning 2

The solution is to use the Limit Curve Fitting function.

Solution 2a

Another solution is to use a longer optimization period.

Solution 2b

Although it may seem that the solutions have produced inferior results, they are probably more stable and reliable than the result of the original optimization (Warning 2) .

 

Sometimes optimization finds a solution where it is required to be very active.

Warning 3

You can force Optimal Trader to find solutions with fewer trading occasions by setting a brokerage fee of about 0.2% - 2%. What happens is that solutions with fewer trading occasions are favored because optimization takes the fee into account at every trading occasion. Notice that the Return Curve also takes the brokerage fee into account so you can reset the brokerage fee to 0% after optimization for it to be shown correctly.

Solution 3

 

Between august 2004 and may 2006 the Stockholm Stock Exchange advanced constantly without any major declining periods. The chart below shows a Swedish Index fund optimized for this period.

Warning 4

Optimizing parameters for a period without any major declining periods may be dangerous because the models will then not have been trained to handle negative market movements. In this case it may be better to set the parameters manually or to use Optimal Traders default parameters which are developed to produce stable results for most market patterns. To use these parameters you have to reset all the settings for the model. Notice that other settings such as Brokerage Fee will also be reset.